Everyone gets the framing wrong from the first conversation. They say things like, “I need to automate my follow-ups,” or “I want a sequence that goes out to leads.” The unit of work they’re picturing is the message. And once that’s the frame, every decision that follows is about the message. The cadence. The wording. The subject line. The trigger.

That framing leads to bad workflows, even when the messages themselves are fine. Because the actual problem in most small businesses isn’t that the right email didn’t get sent. It’s that the business owner forgot something important about a customer. The birthday. The follow-up after a hard appointment. The check-in two weeks after the consultation. The reminder about the document the client said they’d send. The owner remembered some of these things. They missed others. The miss is what cost the relationship.

Subscribe now

If you’re a small business owner, your real bottleneck isn’t writing capacity. It’s memory. You don’t have enough RAM to hold every promise, every personal detail, every “I’ll get back to you next month” floating around your business. Nobody does. The clients you lose, you lose to the clients who get your attention more reliably.

What good automation actually does is offload memory. The system remembers when something happened, who was involved, what context surrounded it, and surfaces that at the right moment so a human can act. The message that gets sent at the end of that chain is a side effect, not the point. The point is that you didn’t forget.

Once you start thinking about it this way, the design of the workflow changes. You stop optimising for “did the email send” and start optimising for “did the right thing get noticed in time to do something about it.” Those are different success metrics. The first one is fully automated and forgettable. The second one requires the system to flag things to a human and let the human decide.

A retention reminder for a hairdresser shouldn’t be “send a rebook SMS at week 6.” It should be “if a client who normally comes every 6 weeks hasn’t booked by week 7, surface them to the owner with a one-line note about what they had done last time, and let the owner write a quick text in their own voice.” The output looks similar from the outside. The thing being automated is completely different.

The conventional view has a point. For high-volume, low-relationship businesses, fully automating the message is fine. An e-commerce store with 10,000 customers can’t have a human in the loop for every birthday email. The automated message is the only way it scales. But for the typical small service business with hundreds, not thousands, of customers, fully automating the message is the wrong target. There aren’t enough customers to justify the cold tone, and there are too many for human memory alone. Memory automation is the middle path that fits.

The reason this reframe matters is that most small business owners look at automation and decide it’s not for them, because they’ve seen what fully automated outreach looks like and they don’t want to do that to their clients. They’re right to be hesitant. What they don’t realise is that they’re allowed to use the same tools to do a completely different thing. They can use Claude and a CRM and Zapier not to send messages, but to remember on their behalf. To flag the client who’s been quiet. To pull the loan details before a renewal conversation. To draft a message that the owner reviews in 30 seconds and sends with their own thumb.

That’s a fundamentally different relationship to the technology. It’s not “the bot does the work.” It’s “the bot keeps the threads alive so I can do the work.” Which is closer to what small business owners actually want, and almost never what they get sold.

So when you’re scoping an automation project, the first question isn’t “what should we send.” It’s “what are you forgetting that you don’t want to forget anymore.” Start there and the rest of the design becomes obvious.

What looks like automation, in the businesses that get it right, is mostly just memory that doesn’t fail. The sending is incidental.

Subscribe now

It’s the message that goes out to a customer who’s been silent for around 10 days. Not a chase. Not an offer. Not a reminder. Just a short note that says, in effect, hey, we noticed you’ve been a bit quiet, no pressure, just letting you know we’re here.

It outperforms almost every other touch in the workflow, and the reason is psychological. By day 10, the customer hasn’t booked, hasn’t replied, hasn’t done the thing. They are starting to feel a small but real amount of guilt or awkwardness about it. They might want to come back. They might not even know yet. What they need, at exactly that moment, is permission to come back without having to explain themselves. The 10-day message gives them that. Nothing else in the workflow does.

The mistake almost everyone makes is treating this message like a marketing touch. They tack on a discount. They include a CTA. They optimise the subject line for opens. All of which kills it. The moment the message reads like a campaign, the customer sees it for what it is: a system trying to reactivate them. And reactivation is exactly the dynamic they’re trying to avoid. They don’t want to be reactivated. They want to feel remembered, which is the opposite thing.

The other mistake is sending it too early. At 3 days, the customer hasn’t yet started feeling like they’ve gone quiet. The message lands as a nudge, not a check-in, and it adds pressure. At 7 days, same thing, just slightly less so. At 10 days, the window opens. After 14 days, it starts to close again, because by then the customer has either moved on or rebooked elsewhere. There’s a specific moment, and it’s narrow.

Here’s roughly what the message should look like:

“Hey {name}, just noticed it’s been a little while. No reason to message other than to say we’re still here if you want to come back. Hope everything’s going well.”

That’s it. No offer. No link. No suggested booking time. The signature is the business owner’s first name, not “the team at X.” It can be sent by SMS or email. Email actually works better for this specific message because the customer can sit with it for a moment before responding, where SMS pressures an immediate reply.

What changes when this message is in place is harder to describe than to feel. In a typical setup with regular customers who’ve started drifting, a meaningful slice of them will respond, sometimes within minutes. Not because of anything clever in the message. Because they were already half-thinking about coming back, and someone gave them a low-pressure reason to act on it. A trainer I’ve spoken with says she gets her best reactivation rates from this single touch, beating birthday messages, beating reminders, beating offers.

What most people don’t notice is that even the customers who don’t reply benefit from this message. The ones who are genuinely gone read it and feel slightly warmer about the business. They might recommend it to someone else later. They might come back in six months. The 10-day message doesn’t have to convert immediately to do work. It puts a small deposit in the relationship that pays out invisibly over time.

So if you’re setting up retention automation for any service business (fitness, beauty, dental, allied health, anything where customers come back regularly), this is the touch to start with. Not the elaborate sequence. Not the 5-step nurture. Just one short, warm, no-ask message at day 10.

The broader lesson is that the highest-performing automations are often the ones that look like they’re doing the least. A reminder asks for something. An offer asks for something. The 10-day check-in just exists. That’s why it works.

Subscribe now

Not because the automation is bad. Because the tone is wrong. And tone is the entire relationship.

Here’s the common belief: small businesses lose customers because they don’t follow up enough. They forget to send reminders. They forget birthdays. They drop the ball between visits. So the answer is obvious. Automate the follow-ups. Set up sequences. Make sure no one falls through the cracks.

That logic is right in the abstract and wrong in practice. Because what most businesses actually automate isn’t the follow-up. It’s the feeling of being followed up by a brand instead of by a person. And those are completely different experiences for the customer on the other end.

Subscribe now

I see this most clearly in the businesses that get it wrong. The hairdresser whose rebooking SMS says “Hi {first_name}, it’s time to book your next appointment!” with a generic offer attached. The personal trainer whose missed-session email opens with “We noticed you haven’t been in.” The accountant whose tax-time reminder reads like it was written by a marketing department, because it was. Every single one of those messages is technically a follow-up. Every single one of them tells the customer: you are a record in a system, not a person we know.

And here’s the cost. A customer who feels remembered will come back. A customer who feels managed won’t. The automation, by being slightly off in tone, doesn’t just fail to bring them back. It actively pushes them out, because the message is a small reminder, every time it arrives, that the relationship they thought they had isn’t what they thought it was.

The reason this happens is structural. Most automation is built by someone who is thinking about the workflow, not about the moment the message arrives. They’re thinking about triggers and conditions and segments. They’re not thinking about what it feels like to be the person who opens that text on a Tuesday morning. Those are different skills. The automation works fine. The message lands wrong.

What works instead is automation built around tone first, workflow second. The triggers can be the same. The timing can be the same. But the actual words have to sound like they came from a human who actually knows the customer, even when they didn’t write that particular message. That means short. Specific. Casual. Sometimes a bit imperfect on purpose. No “Dear valued customer.” No exclamation marks at the end of every line. No offers tucked in awkwardly. Just a brief note that sounds like a text from someone who remembered them.

This is one of the few areas where AI actually helps a lot, because Claude or a similar model can write the message in the business owner’s voice, pulling in the specifics about that customer, and it lands closer to how a real text would read than any generic template ever did. But only if the person setting it up sweats over the voice the same way they’d sweat over the menu of a restaurant. The voice is the product.

The conventional view has a point, though. Some businesses do need more follow-up, not less. A solo accountant trying to do everything by memory will lose clients to the firm down the road that actually sends reminders. The answer isn’t to skip the automation. The answer is to be careful about what the automation says when it arrives.

I think about this in terms of a simple test. Read the message out loud. If it sounds like something the owner would actually say to a customer they ran into at the supermarket, it works. If it sounds like a CRM, it’s already failed, regardless of how clever the trigger logic is behind it.

The thing nobody talks about is how badly mistuned automation poisons the perception of the business. A customer who gets one robotic message will not consciously think “the tone of that automation was bad.” They’ll just feel slightly worse about the business. They’ll book less often. They’ll mention the brand less to friends. The decay is invisible until you wonder why your retention is slowly leaking.

So if you’re setting this up for your business, the question to ask first isn’t what to automate. It’s: when this message arrives in someone’s inbox at 7pm on a Wednesday, will it make them smile or make them sigh? That’s the only thing that actually matters.

Most of what’s labelled “AI for small business” is solving the wrong problem with the wrong tone. The relationship was already there. The automation just needs to not break it.

Subscribe now

It doesn’t chase.

Subscribe now

The default move when a client goes quiet is to chase them. They’ve missed two sessions. They haven’t replied to the booking link. The trainer’s instinct is to nudge: “Hey, are you still in? Want to lock something in for next week?” Reasonable. Polite. And it almost always fails.

It fails because by the time a client has gone quiet, they’re already a bit embarrassed. Maybe they were sick. Maybe work got intense. Maybe they fell off their routine and they don’t know how to come back without explaining themselves. A booking nudge forces them to either lie (”yes, locking in next week!” and then disappearing again) or confront the awkwardness. So they pick a third option: they ignore the message and quietly become a former client.

The version of this workflow that works does the opposite. It checks in. It says, in effect: I’m not asking you to book. I’m just letting you know I noticed you’ve been off, and I’m here when you’re ready. That’s a completely different message. One asks the client to commit. The other lets them off the hook.

Here’s the flow:

Client misses two sessions in a row, with no rebook —> system flags them as “paused” (not “lost”) —> 7 days later: a short check-in message goes out (”Hey, hope you’re doing alright. No pressure to book anything, just wanted to check in. Here whenever you’re ready.”) —> if they respond: trainer reads the reply and decides whether to offer a session or just keep the door open —> if they don’t respond for another 21 days: a second, even lighter message goes out, this time with a small unrelated value drop (a recovery tip, a stretch sequence, no ask)

That’s it. Four steps. The whole thing depends on the wording of message one and the discipline to not turn it into a booking attempt.

Where this breaks: trainers who can’t help themselves and tack on “btw, I’ve got Tuesday at 6 free if you want it.” The moment that line goes in, the message stops being a check-in and starts being a chase, and the client reads it as a chase. The whole psychology resets. They feel pressured, they feel they need to either commit or explain, and they go quiet again. Permanently this time.

The other place it breaks is when the message sounds like a template. Personal training is a high-trust, high-relationship business. A client can tell within two seconds whether their trainer actually wrote a message or whether a system spat one out. The version that works uses the client’s name, references something specific from their last session if you can (”hope the shoulder’s been feeling better”), and reads like a text from a friend. The AI can draft these, but a human has to read each one before it sends. Sometimes you want to tweak it. Sometimes you want to skip it because you already saw the client at a school pickup and you know what’s going on.

The third thing that separates the working version from the broken one is the framing internally. Trainers who set this up call it the “pause workflow.” Not the “retention sequence.” Not the “win-back funnel.” Pause, because the assumption baked into the system is that the client is still your client. They’ve just paused. The language you use about your own clients shapes the language you write to them.

Why does this matter more than a booking reminder? Because in fitness specifically, drop-off isn’t a contract problem, it’s an emotional one. Clients quit because they feel bad about themselves, not because they’re unhappy with the trainer. A workflow that respects that, that holds the door open without forcing them to walk through it, gets a meaningful number of those clients back. Not all of them. Some are gone for good. But a chunk of the people who would otherwise quietly disappear will reappear two months later because of one short message that didn’t ask anything of them.

The interesting thing is how counterintuitive this feels at first. Most automation tutorials teach you to nudge harder, follow up more aggressively, never let a lead go cold. In fitness, that approach is the thing actively killing your retention.

So if you’re building this for a trainer, the question isn’t “what’s the best follow-up cadence.” It’s: what’s the lightest possible touch that still keeps the relationship warm? Start there. Build out from there. And resist every urge to add a booking link.

Worth thinking about whether that same instinct applies in your business too.

Subscribe now

Six weeks later, you haven’t seen them since session two.

Not because the program was wrong. Not because they didn’t want to get better. They just lost momentum, skipped a session, then felt awkward coming back. Life got in the way. Nobody followed up.

Sound familiar?

Subscribe now

Drop-off is the most common problem in private physiotherapy. Patients don’t complete their programs. They improve a little, stop hurting as much, and quietly disappear. The clinic loses the revenue. The patient loses the outcome. And nobody wins.

Most clinics don’t have a system to catch this. The physio is fully booked between appointments. There’s no admin staff to chase every patient who hasn’t rebooked. So the follow-up doesn’t happen, and patients drift away.

We helped a physio clinic fix this. A patient retention workflow that sends the right message at the right moment: after each appointment, between sessions, and when someone goes quiet.

Here’s what the manual process looked like before:

Appointment finished —> patient leaves with verbal instructions —> no contact until next booking —> gap grows —> patient doesn’t rebook —> physio notices weeks later —> too late to re-engage

Here’s what it looks like now:

Appointment completed —> post-visit message sent within 2 hours (”Here are today’s exercises, see you next [day]”) —> 48 hours before next session: reminder sent —> session missed with no rebook: check-in sent (”Hope you’re doing okay, want to get back on track?”) —> 10 days no contact: gentle re-engagement (”Your program is still here when you’re ready”) —> program milestone reached: encouragement message sent —> discharge: short feedback request and maintenance plan reminder

Each message is short. Conversational. Written in the clinic’s voice. Not a generic reminder from a booking system.

Does the automation replace the clinical relationship?

No. The assessment, the hands-on treatment, the judgment calls about when to push and when to ease off — that’s the physio’s job and it stays that way. But between sessions, most patients hear nothing. That silence is where the drop-off happens. This workflow fills it.

Four things the system handles:

Post-appointment instructions. Within a couple of hours of each session, the patient gets a short message: what they worked on today, what to do before next time, any precautions. It reads like a note from their physio, not a form letter. Patients consistently say this is the thing that keeps them doing their home exercises.

Session reminders. A reminder goes out 48 hours before the next booking. Simple. Not flashy. It reduces no-shows and gives the patient a natural nudge to stay on schedule.

The drop-off catch. If a patient misses a session and doesn’t rebook within a few days, a message goes out. Not chasing. Not sales-y. Just: “Hope you’re doing okay, your spot’s here when you’re ready.” Most patients who receive this message rebook. They were just waiting for a reason to come back.

Program milestone recognition. When a patient hits a milestone in their program (halfway through, final session booked), they get a short encouraging message. Small thing. Meaningful.

The stack: a clinic management system like Cliniko or Nookal for appointment data, Zapier to trigger messages at each stage, Claude to draft the personalised post-appointment instructions and check-ins, SMS delivery through Twilio or email through Gmail. If the clinic already uses Cliniko, this can be connected within a few days.

The result for the clinic we worked with: 3 physios, around 80 active patients at any one time. Before the workflow, roughly 35% of patients didn’t complete their prescribed programs. After six months, that was down to 19%. Average sessions per patient went up. Revenue per patient went up. And the physios reported spending less time wondering which patients had fallen off the radar, because the system was watching for them.

The thing about physio specifically is that patients often stop coming right when they feel a bit better but before they’re actually fixed. They think they’ve recovered. They haven’t. The 10-day re-engagement message catches a lot of those patients. It’s not pushy. It’s just present.

If you run a physio clinic and you know patients are leaving before they should, this is worth your time.

If you want it built for your clinic, let’s talk. And subscribe — I find the workflows, you focus on the patients.

Subscribe now

You’re still with a client. Your paralegal has left for the weekend. The message sits in the inbox.

By Monday morning, that person has already called two other firms.

Sound familiar?

Subscribe now

This is the most expensive problem in small law firms. Not the billing. Not the admin. The leads that go cold before anyone gets to them.

Most law firms are slow on first response not because they don’t care, but because the people who handle enquiries are the same people doing the actual legal work. When a lawyer is in court or deep in a contract, an enquiry waits. And in professional services, the firm that responds first usually gets the client.

We helped a small law firm fix this. An intake and lead nurturing workflow that responds within minutes, qualifies the lead automatically, and routes it to the right lawyer before anyone has to open their inbox.

Here’s what the manual process looked like before:

Enquiry submitted —> sits in inbox —> lawyer sees it when free —> writes a reply —> sends intake form manually —> waits —> chases if no response —> books consult or loses the lead entirely

Here’s what it looks like now:

Enquiry submitted (web form, email, or phone callback request) —> immediate acknowledgment sent (”We’ve received your enquiry, here’s what happens next”) —> short intake questionnaire sent automatically —> 48 hours no response: polite follow-up sent —> questionnaire completed: AI summarises key details and routes to the right practice area —> lawyer receives a 5-minute brief on the prospective client —> consultation booked via calendar link —> reminder sent 24 hours before the call

The first message goes out in minutes, regardless of when the enquiry arrives. Friday evening, public holiday, 11pm on a Sunday. The prospective client hears from the firm before they’ve had time to start searching for someone else.

Does this mean a chatbot is handling legal advice?

No. The intake form asks basic questions: what type of matter, rough timeline, what outcome they’re hoping for. Nothing legally sensitive. The AI drafts the acknowledgment and summarises the intake responses for the lawyer. The lawyer reviews the brief and decides whether to take the consult. All the actual legal judgment stays where it belongs.

Three parts the system handles:

Immediate acknowledgment. The moment an enquiry comes in, a response goes out. It confirms receipt, explains the next step, and sets expectations for when the lawyer will be in touch. Professional. Warm. Written in the firm’s voice.

Intake and qualification. A short form collects the basics: practice area, type of matter, urgency. When it comes back, Claude summarises it for the lawyer in plain English. “Prospective client seeking advice on a commercial lease dispute, needs resolution within 3 months, has already attempted to negotiate directly.” Two minutes to read, not twenty.

Routing and booking. Based on the practice area flagged in the intake, the enquiry routes to the right lawyer. A calendar link goes to the client. The consultation gets booked without a single back-and-forth email.

The stack: Tally or Typeform for the intake form, Zapier to trigger messages at each stage, Claude to draft acknowledgments and summarise intake responses, Calendly for booking, Gmail or Outlook for delivery. Most small firms can have this running in under a week.

The result for the firm we worked with: around 25 enquiries a month. Before the system, roughly 40% converted to a booked consultation. After six months, that was 64%. The work didn’t change. The speed of first contact did.

There’s something worth saying about the weekend enquiry specifically. It sounds like a small thing. But a prospective client who reaches out on a Saturday and hears back within minutes forms a very different impression of the firm than one who waits until Tuesday. That impression carries into the consult. Into the engagement. It matters more than most lawyers realise.

If you run a small firm and you know enquiries are slipping through between Friday afternoon and Monday morning, this is worth looking at.

If you want it set up, let’s talk. And subscribe — I find the workflows, you focus on the clients.

Subscribe now

And then three days later they call asking if anyone looked at it.

Not because you didn’t fix it. Because you forgot to tell them.

Sound familiar?

Subscribe now

This is the most common complaint about small IT support companies. Not that the work is bad. That nobody communicates while it’s happening. The client submits a request and then waits in silence, not knowing if anyone saw it, not knowing when it will be done, not knowing it has been resolved until something stops working again.

We helped an IT support company fix this. A ticket workflow that handles all the communication automatically, from first contact to confirmed resolution, without anyone having to remember to send an update.

Here’s what the manual process looked like before:

Client emails or calls —> technician adds it to a mental queue —> fixes it when time allows —> sends a quick reply if they remember —> client follows up because they heard nothing —> technician explains it was sorted last Tuesday

Here’s what it looks like now:

Client submits request (email, web form, or portal) —> ticket created automatically with a reference number —> acknowledgment sent within minutes (”We’ve got your request, reference #1042, we’ll be in touch”) —> 24 hours later: update sent if still open (”Still working on it, expected resolution by end of day tomorrow”) —> technician marks issue resolved —> resolution email drafted and sent (”Issue resolved: plain-English summary of what was done”) —> 3 days later: follow-up sent (”Everything still running smoothly?”) —> no issues reported: ticket closed cleanly —> resolution summary added to internal knowledge base

Each update sounds like a person wrote it. Because technically, the template was written by a person once. After that, the system fills in the specifics: ticket number, client name, issue description, resolution notes.

Does this replace the technician?

No. The diagnostic work, the actual fix — that still needs a human. But the communication around the fix was eating hours every week. Status update calls, chaser emails, tracking down which tickets were open. That’s what this removes.

Three parts the system handles:

Intake and acknowledgment. The moment a request comes in, the client knows it has been received. Reference number included. This alone eliminates most of the “did anyone get my email?” calls.

In-progress updates. If a ticket is still open after 24 hours, an update goes out automatically. Not a detailed technical note. Just: we’re on it, here’s roughly when to expect resolution. Simple. Professional.

Resolution and follow-up. When the technician marks a ticket resolved, a plain-English summary goes to the client. Three days later, a short check-in: is everything working as expected? If there’s an issue, the client replies and a new ticket opens. If not, it closes cleanly.

The stack: Freshdesk or a simple Airtable ticketing setup for intake, Zapier to trigger messages at each stage, Claude to draft the resolution summaries and knowledge base entries from the technician’s notes, email delivery through Gmail or Outlook. For companies already using a helpdesk tool, most of this plugs in directly.

The result for the company we worked with: about 45 active clients, a mix of small businesses. Before the system, they were fielding 10 to 15 status-update calls per week. Clients calling not because something was wrong, but because they hadn’t heard anything. After setup, that dropped to 2 or 3. The team had more time for actual support work. And client satisfaction scores in their quarterly reviews went up, even though the technical quality of the work hadn’t changed at all.

That’s the thing about IT support. The clients don’t always know if the technical work is good or not. They know how it felt. And what it usually feels like is silence while you wait and hope someone is working on it.

This workflow changes that without the technicians having to do anything differently. They fix things. The system handles the communication around it.

If you run an IT support business and your inbox is full of clients asking for updates on things you’ve already fixed, this is for you.

If you want it set up, let’s talk. And subscribe — I find the workflows, you focus on the problems.

Subscribe now

And then they leave.

Most of them never come back.

Not because the coffee was bad. Not because the service was off. They just didn’t think about you again. The café near their office started a loyalty card. A friend recommended somewhere new. Life moved on.

Sound familiar?

Subscribe now

Here’s the thing about cafés: you get a lot of foot traffic. But foot traffic doesn’t automatically become regulars. A regular is someone who has chosen you over and over. And that choice doesn’t happen by accident. It happens because something kept reminding them you exist.

Most café owners don’t have a system for that. They rely on the coffee being good, the staff being friendly, and hoping people come back. That works for some. It doesn’t work for most.

We helped a café fix this. A loyalty and re-engagement workflow that starts the moment someone scans their first stamp.

Here’s what it looked like before:

Customer walks in —> orders —> leaves —> no follow-up —> maybe returns, maybe doesn’t

Here’s what it looks like now:

First visit: customer scans QR code at counter —> visit logged with name, email, birthday —> 3 days later: short message sent (”Loved having you in. Here’s 10% off your next visit”) —> birthday approaching: message goes out 7 days before (”Free flat white on your birthday”) —> 30 days no visit: win-back message sent (”It’s been a while. Your next coffee is on us.”) —> loyalty card complete: reward triggered automatically

Each message is short. Casual. The kind of thing a good barista would say if they had time to write it down.

Does this feel intrusive?

Not if you get the tone right. The mistake most small businesses make is writing these messages like an email marketing campaign. Café customers don’t want that. They want to feel remembered, not marketed to. The messages here are written to sound like they came from the café itself: warm, brief, low-pressure.

Three touchpoints the system handles:

The 3-day come-back. The simplest one. A short message saying “glad you stopped in, here’s something if you want to come back.” No hard sell. Most people who visited once and liked it will respond to a gentle nudge. This is that nudge.

The birthday message. Collected at first check-in. Goes out 7 days before the date. One free drink offer. It costs the café almost nothing. It creates a memory.

The 30-day win-back. If someone visited once and hasn’t been back in 30 days, they’ve probably forgotten you exist. This message is a specific reason to return: a free coffee, a new menu item, a simple “we miss you.” Something concrete.

The stack: a loyalty app like Stamp Me, or a QR check-in built in Airtable. Zapier triggers messages at each interval. Claude drafts the messages in the café’s voice. Email or SMS delivery through Klaviyo or Twilio. If the café already has a loyalty programme, most of this can be plugged in without starting from scratch.

The result for the café we worked with: around 400 unique visitors a month. Before the workflow, about 24% returned within 30 days. After three months with the system running, that was 38%. That sounds like a small shift. It isn’t. At a café, repeat visits compound. A customer who comes back twice a month is worth 10 to 15 times what a one-time visitor is worth across a year.

Worth noting: the 3-day message does the most work. Cafés are higher frequency than most service businesses, but the emotional connection is lighter. You’re not their accountant. You’re not their physio. You’re the place they start their morning. The 3-day message leans into that: the lightest possible nudge at exactly the right moment.

If you run a café and first-time visitors keep walking out without coming back, this is worth looking at.

If you want it set up for your café, let’s talk. And subscribe — I find the workflows, you focus on the coffee.

Subscribe now

And then three years pass. Their fixed rate expires. And they call a different broker.

Not because they were unhappy. Not because you did anything wrong. But because someone else got to them first. The bank sent a letter. A competitor ran a comparison ad. And when the client started thinking about renewal, they didn’t think to call you. They went with whoever was in front of them.

Sound familiar?

Subscribe now

The renewal window is the most predictable opportunity in mortgage broking. You know exactly when every client’s rate expires. You know what they’re paying. You know their situation. All of that is in your CRM.

What most brokers don’t have is a system that uses that information automatically.

We helped a mortgage brokerage fix this. A renewal pipeline that starts reaching out 12 months before expiry, warms up the relationship over time, and drafts a personalised proposal when it matters most — without the broker having to remember to do any of it manually.

Here’s what the manual process looked like before:

Broker checks spreadsheet occasionally —> identifies clients expiring soon —> writes individual emails when they have time —> follows up if they remember —> misses the ones who slipped through —> client refinances elsewhere

Here’s what it looks like now:

Mortgage settled —> expiry date logged —> 12 months out: warm check-in sent (”hope you’ve settled in well, let us know if anything changes”) —> 6 months out: rate environment update sent (”rates have been moving, worth a chat about your options”) —> 3 months out: personalised renewal proposal drafted and sent (”your fixed rate expires in 3 months, let’s review before you roll to variable”) —> no response: follow-up sent 2 weeks later —> client responds: call booked

Each message is personalised to the client: their loan amount, their lender, their current rate, their situation. Not a generic newsletter. A message that reads like the broker actually remembered them.

Does this replace the broker relationship?

No. The actual renewal conversation, the rate comparison, the recommendation — that still needs the broker. But most clients never even get to that conversation because nobody reached out in time. This workflow fixes that.

Three stages the system handles:

12-month check-in. Not about the mortgage at all. A short, warm message checking in on how the client is settling in. Building relationship capital before it’s needed. The broker’s name is in the client’s inbox before the renewal is even a thought.

6-month rate review. A light, useful message. “Rates have been moving in the last six months. Worth a quick chat to make sure you’re still in a good position.” Informative. Low-pressure. Opens the door without pushing.

3-month renewal proposal. The most important message. Personalised to the client’s exact situation: “Your fixed rate of X% with [lender] expires on [date]. Before it rolls to the variable rate, let’s look at what options are available.” Includes a clear call to action. The AI drafts this from the client’s loan data. The broker reviews and sends.

The stack: a CRM or Airtable with mortgage expiry dates, Zapier to trigger messages at each interval, Claude to draft personalised messages from loan data, and email delivery through Gmail or Outlook. If the brokerage already has a CRM with loan records, this can be running within a week.

The result for the brokerage we worked with: they had 140 active clients. Before this workflow, the principal was manually checking renewals every few months and catching maybe half of them in time. After setup, every expiry was tracked, every message went out on schedule, and renewal retention improved from around 55% to 78% in the first year.

That is a material difference in revenue. Not from new clients. Just from not losing the ones they already had.

There’s something worth noting about the 12-month check-in specifically. It changes the whole dynamic. When the broker makes contact 12 months before renewal, it’s not a sales call. It’s a relationship touchpoint. By the time the 3-month proposal lands, the client already feels looked after. Comparison shopping tends to happen before you’ve been in touch. This workflow makes sure you’re in touch first.

If you’re a mortgage broker and you know you have renewals slipping through, this is worth a look.

If you want it built for your brokerage, let’s talk. And subscribe — I find the workflows, you focus on the clients.

Subscribe now

And at some point during every single week, at least one client calls to ask what’s happening on their job.

It’s not always because something has gone wrong. Most of the time it’s just anxiety. They’ve committed a significant amount of money. The site is a mess of equipment and workers they don’t understand. They haven’t heard from you in a while. So they call.

Sound familiar?

Subscribe now

The problem isn’t the clients. The problem is the silence. When people don’t know what’s happening, they imagine the worst. An update, even a short one, changes that completely.

We helped a construction company fix this. A communication workflow where the site manager logs what happened in a quick form, AI drafts a professional update, the owner reviews it in about 30 seconds, and it goes to the client before they even think to pick up the phone.

Here’s what the manual process looked like before:

Site manager tells owner verbally what happened on site —> owner sits down to write client update when they have time —> update goes out late or not at all —> client calls to ask what’s happening —> owner interrupts their day to take the call —> repeat across every active project

Here’s what it looks like now:

Milestone hit on site —> site manager fills in a 5-minute form (what was completed, what’s next, any issues or delays) —> AI drafts a professional update in the company’s tone from the form —> owner receives draft for review —> approves or edits in 30 seconds —> sent to client via email —> end of week: summary of all milestones automatically compiled and sent

The owner still controls every message. They’re just not writing it from scratch anymore.

Does this replace the owner’s judgment on sensitive issues?

No. A structural problem, a delay caused by bad weather, a dispute about a variation — those need a real conversation. But about 80% of client communication is routine progress updates. This workflow handles the routine.

Three things the system handles:

Milestone updates. After significant work is completed (foundations poured, framing up, roofing done, fit-out started), the site manager logs 3 to 5 bullet points into a simple form: what was done, what comes next, anything the client needs to know. The AI turns it into a proper client email. Clear, professional, specific to their project. The owner reads it, approves it, done.

Weekly summaries. At the end of each week, the system compiles everything logged for that project and sends a brief summary to the client: what happened this week, what’s planned for next week. Clients on active builds get one consistent email a week without anyone having to write it from scratch.

Issue flagging. If the site manager logs a delay or a problem, the form triggers a different path. The owner gets notified immediately and the draft message is flagged for careful review before sending. Nothing sensitive goes out automatically.

The stack: Tally or Google Forms for the site manager log, Zapier to route submissions to Claude, Claude drafts the update using project and client context, and the owner approves via a simple link or reply before delivery through Gmail or Outlook. Most builders are up and running in about a week. No new software the client needs to install.

The result for the company we worked with: the owner was spending close to 2 hours per week writing client updates across 8 active projects. That dropped to about 15 minutes of reviewing drafts. Inbound “where are we at?” calls dropped by more than half in the first month.

One client told them they felt like the most organised builder they’d ever worked with. Nothing about the actual build had changed. Just the communication.

That’s the thing with this kind of work. The quality of what you’re building isn’t always what clients remember. It’s how they felt throughout the process. Anxious? Ignored? Or informed and reassured all the way through?

This workflow is a straightforward answer to that.

If you run a construction company and your phone rings too much because clients don’t know what’s happening, this is worth looking at.

If you want it built for your company, let’s talk. And subscribe — I find the workflows, you focus on the build.

Subscribe now

But somewhere between the morning appointments, the surgeries, the phone calls, and the front-desk chaos, tracking which patients are due for their annual checkup has fallen to a system of manual checks and good intentions.

Sound familiar?

Subscribe now

Vet clinics are interesting because the client actually wants to be reminded. Pet owners don’t avoid the vet the way people sometimes avoid the dentist. They just forget. Life gets busy. The vaccine card goes in a drawer. And by the time they remember, Biscuit is 14 months overdue for her booster.

That gap between forgetting and acting is exactly where this workflow sits.

We helped a vet clinic automate two things: recall communications for annual checkups and vaccine due dates, and post-visit care instructions so the printed sheet that nobody reads can finally retire.

Here’s what the manual process looked like before:

Nurse manually checks records each week —> identifies who’s due in the next month —> sends bulk recall email or postcards —> follows up by phone if no response —> patient comes in (or doesn’t)

Here’s what it looks like now:

Pet visits —> next due dates logged in the system (annual exam, core vaccines, parasite prevention) —> 4 weeks before due: personalised SMS sent to owner with pet’s name and booking link —> 1 week before due: second reminder —> no response: flagged for a brief staff call —> pet books in —> appointment ends: care summary sent via SMS within 2 hours —> medication prescribed: daily reminder SMS until course is complete

The nurse goes from manually chasing appointments to reviewing a short list of patients who genuinely need a call.

Does this replace the relationship between vet and client?

No. The consultation, the diagnosis, the reassurance when a pet is unwell — none of that is replaceable. But the administrative layer around it doesn’t need a human for every step. The reminder that Luna’s rabies booster is due in 28 days can run on its own.

Three things the system handles:

Annual and vaccine recall. The system knows each pet’s due dates. Four weeks out, a message goes to the owner: “Hi [Name], just a reminder that [Pet]’s annual health check is coming up in about a month. We’d love to see them — book here: [link].” A week before the due date, a follow-up goes out. It mentions the pet’s name. It’s not a generic reminder. It reads like it came from the clinic.

Post-visit care instructions. The vet records what care was given and any instructions after the appointment. Two hours later, the owner gets a personalised SMS or email: what to watch for, how to give the medication, when to come back if things don’t improve. No more relying on the printed sheet that ends up in the car. The instructions are readable on a phone and don’t get lost.

Medication reminders. If a medication course was prescribed, the system sends a daily reminder until the course is complete: “Just a nudge: time for [Pet]’s antibiotic with food.” Simple, but it makes a real difference in compliance, especially for older pets on long-term medication.

The stack: Ezyvet or VetLink as the practice management system (both support API access), Zapier to route trigger events, Claude to draft personalised messages from appointment records, and Twilio for SMS delivery. A booking link is included in each recall reminder. Setup takes about a week for a typical small clinic.

The result for the clinic we worked with: vaccine and checkup compliance for active patients improved by about 22 percentage points in the first six months. That’s not from new clients. That’s existing clients who previously slipped through the cracks.

And the post-visit care messages started getting replies. Owners asking follow-up questions they forgot to ask in the consultation room. Which is exactly the kind of thing that builds trust and keeps clients coming back.

Pet owners are genuinely grateful when you remind them. That’s the part worth sitting with. They’re not annoyed by a message saying their dog’s booster is coming up. They’re relieved. Because they care about their animal and they don’t want to drop the ball.

This workflow just makes sure neither side does.

If you run a vet clinic and you’re still chasing recalls manually, this is worth looking at.

If you want it built for your clinic, let’s talk. And subscribe — I find the workflows, you focus on the animals.

Subscribe now

You already know which clients are going to be a problem. The same ones as last year. And the year before that.

They won’t send their documents until you ask. And they won’t respond to the first ask. Or the second. You’ll be emailing them individually, checking your sent folder, following up again, watching the deadline creep closer while their bank statements sit in a shoebox somewhere.

Sound familiar?

Subscribe now

This is one of those problems that accountants have accepted as just part of the job. But it’s not really an accounting problem. It’s a communication timing problem. And that is exactly the kind of problem that automation handles well.

We helped an accounting practice fix this. A document collection workflow that starts running six weeks before the tax deadline, escalates automatically for clients who haven’t responded, and gives the accountant a live view of who’s submitted and who still needs a nudge.

Here’s what the manual process looked like before:

Accountant sends a bulk email to all clients —> waits a week —> manually checks who hasn’t replied —> writes individual follow-ups —> waits —> chases again —> two weeks out: sends another round —> deadline week: phone calls to the stragglers

Here’s what it looks like now:

Tax deadline set in the system —> 6 weeks out: personalised document request sent to each client (lists exactly what’s needed for their situation) —> client submits: marked complete, removed from the sequence —> 4 weeks out: second nudge to everyone still pending —> 2 weeks out: escalation message with deadline clearly stated —> 1 week out: final reminder —> dashboard shows submission status across all clients in real time

The accountant goes from writing and tracking 80 individual email threads to reviewing a dashboard and calling the handful who genuinely need a conversation.

Does it work for every client?

No. Some clients need a phone call regardless of how many emails they receive. But those were always the ones who needed calls. The automation handles the other 60 or 70. It frees the accountant to focus on the 10 who genuinely need personal attention instead of spending equal effort on all 80.

Three things the system handles:

Personalised document requests. The first message isn’t a generic “please send your documents.” It’s specific to each client: “Hi [Name], we’re getting ready for your 2024 tax return. For your situation, we’ll need your PAYG summary, any investment income statements, and receipts for deductions. Here’s the upload link.” Reviewing and sending takes about 30 seconds per client. Previously it took 5 to 10 minutes.

Escalating reminders. The tone shifts naturally across the sequence. The first message is informative and light. The second is a gentle nudge. The third is clear about the deadline and what happens if documents aren’t received in time. None of them feel like automated chasing. The language stays warm throughout.

Status dashboard. The accountant opens one view and sees every client: submitted, pending, or overdue. No searching through email threads. No trying to remember who was chased last week. Clients who submit get marked done automatically when they use the upload link.

The stack: Airtable to hold the client list and track submission status, Zapier to trigger messages at each interval, Claude to draft personalised emails from client data, and a shared Google Drive or Dropbox folder for document uploads. Most practices can plug this into their existing email setup. Setup took about a week for the practice we worked with.

The result: the practice had 80 individual tax clients. Before this workflow, the principal was spending 3 to 4 hours per week during tax season on document chasing alone. After setup, that dropped to about 30 minutes of reviewing the dashboard and a handful of calls.

That’s time that goes back into actual accounting work. Or into taking on more clients. Or into just not being stressed from October to April.

There’s something else worth mentioning. The clients noticed the difference too. Not that they were receiving automated emails, but that the communication was clearer. They knew exactly what was needed, when it was needed, and where to send it. A lot of the chasing wasn’t really clients being difficult. It was clients not knowing precisely what to send. The system fixed that from the start.

If you run an accounting practice and you’re still manually chasing the same documents from the same clients every year, this is worth looking at.

If you want it built for your practice, let’s talk. And subscribe — I find the workflows, you focus on the numbers.

Subscribe now

Not dramatically. There’s no cancellation. No complaint. They just stop coming back. Six weeks go by. Eight weeks. You see their name in the system and realise you haven’t seen them in months.

Sound familiar?

Subscribe now

Salon churn is quiet. That’s what makes it hard. A gym client cancels their membership — you know they’re gone. A salon client just doesn’t rebook. There’s no notification. No flag. Just an empty slot where they used to be.

And by the time you notice, they’ve probably found someone else.

We helped a salon fix this. A retention workflow that sends rebooking reminders automatically, personalises birthday offers, and runs a win-back sequence for clients who’ve gone quiet — without the owner or receptionist having to track any of it manually.

Here’s what the manual process looked like before:

Receptionist checks who hasn’t rebooked —> tries to remember specific clients —> sends a follow-up if they have time —> maybe gets a response —> repeat inconsistently

Here’s what it looks like now:

Client visits —> appointment logged —> 4 to 6 week timer starts (based on service type) —> AI drafts rebooking reminder —> sent via SMS —> no response after 7 days: follow-up sent —> client still quiet at 10 weeks: win-back sequence triggers —> birthday in 7 days: birthday offer drafted and sent separately

Three sequences the system handles:

Rebooking reminders. Four weeks after a colour appointment. Five to six weeks after a cut. The message is warm and personal: “Hi [Name], it’s been about a month since your last visit — just wanted to check in. Ready to rebook?” It comes from the salon, not a generic booking app. Setup takes about 30 seconds per client profile, and after that it runs on its own.

Birthday offers. The client’s birthday is already in the system. Seven days before it, an offer goes out based on what they usually book. Not “Happy birthday, here’s 10% off everything.” Something like: “Your birthday is coming up. We’d love to treat you to a complimentary blow-dry with your next colour.” More personal, better result.

Win-back sequences. Client goes quiet for 8 to 10 weeks with no rebook and no response to the reminder. A two-step sequence starts. A casual check-in first. A small offer a week later if there’s still no reply. The tone matters here. It doesn’t feel like chasing. It feels like the salon noticed they were gone.

Does this replace the personal relationship?

No. A good stylist still knows their clients. But the relationship is only as good as the follow-through. Most salons lose clients not because the service was bad but because nobody reached out at the right moment.

The stack: Timely, Fresha, or Vagaro as the booking system (all three support these triggers), Zapier to route the timing events, Claude to draft personalised messages from client history, and Twilio for SMS delivery. Most salons are up and running in about a week.

The result for the salon we worked with: rebooking rate for existing clients went from around 45% to 68% in the first 90 days. That’s not from bringing in new clients. That’s from not losing the ones they already had.

The birthday campaign had the highest response rate of anything we tested. Clients replied personally. Some came in the same week. A few mentioned the message specifically when they walked in.

That tells you something. Clients want to feel remembered. They just need the salon to have a system that actually remembers.

If you run a salon and you’re watching clients drift away without knowing why, this is worth a look.

If you want it built for your salon, let’s talk. And subscribe, I find the workflows, you focus on the clients.

Subscribe now

Each one gets a personalized letter. Different holdings, different goals, different risk tolerance, different things that happened in the market that are relevant to them specifically.

You used to write all 80 yourself. Over about a week and a half.

Sound familiar?

Subscribe now

Quarterly client updates are one of the biggest time drains in independent financial advisory. It’s not something you can easily outsource because the content is client-specific, has to sound like you, and sits in a regulated environment. So it lands on the advisor’s desk every 3 months like clockwork.

That is exactly what we helped one independent advisor fix. A workflow that pulls the relevant data for each client, drafts a plain-English summary in the advisor’s voice, and turns what used to take 10+ days into a 2-day review process.

Here’s what the manual process looked like:

Quarter ends —> advisor opens each client’s portfolio data one by one —> writes a performance summary, market context, and personalized notes for each relationship —> formats it —> sends —> repeat 80 times

Here’s what it looks like now:

Quarter ends —> portfolio data for all clients exported from the platform —> each client record fed into AI alongside their goals profile —> AI drafts all 80 letters in one batch —> advisor reviews each in 5-10 minutes, adds any personal notes, makes edits —> finalized and sent

The compliance piece is worth addressing directly. The drafts don’t go out without review. Every letter gets advisor sign-off before it’s sent. The AI is writing the first draft, not the final product. That’s an important distinction for anyone working in a regulated environment.

Does the AI know each client personally?

No. That’s why the goals profile matters. Before running this workflow, you build a short profile for each client: their goals, their risk tolerance, any notes from the last conversation, anything specific to their situation. The AI writes from that profile. Without it, the drafts are generic. With it, they feel specific.

Three things this setup handles:

Performance summary. The AI reads the quarter’s numbers and writes a plain-English paragraph: the portfolio was up X%, the main drivers were Y and Z, here’s how that compares to the benchmark. No jargon. Language the client actually understands.

Market context. A brief note on what happened in the market and how it’s relevant to this client’s specific situation. The advisor writes one market overview paragraph once per quarter, and the AI weaves it differently into each letter based on the client’s holdings and risk profile.

Personalized goals check-in. Based on the client’s profile, the letter connects performance back to their goals. Something like: “You’re still on track to reach your target by 2031 based on current projections” or “Given the rate changes this quarter, it might be worth revisiting your bond allocation when we speak.” This is the part that makes the letter feel like it was written for them.

The stack: portfolio management software with a data export function (most platforms support this), a structured client profile template in Google Sheets or Notion, Claude for drafting with a consistent prompt, and Google Docs for the advisor’s review and final formatting. No special integrations needed. The batch runs once a quarter.

The results for the advisor we worked with: quarterly updates went from a 10-day writing project to a 2-day review process. And honestly, the letters got better. When you’re not writing 80 from scratch, you have the mental space to actually improve each one.

There’s a version of this where every letter still feels hand-crafted. That version is possible. It just doesn’t take 10 days anymore.

If you run an advisory practice and want to build this workflow, let’s talk. And subscribe — I find the workflows, you focus on the clients.

Subscribe now

And every morning, you open your inbox to find the same messages.

“Where is my order?”
“Can I return this?”
“Does this come in another size?”
“I never got a confirmation email.”
“How long does shipping take?”

Sound familiar?

Subscribe now

It’s not hundreds. It’s maybe 30 or 40 on a busy day. But they land at all hours. And answering them takes time you don’t have. Because you’re also managing suppliers, updating listings, running ads, and trying to actually grow the business.

The worst part is that most of the answers already exist somewhere. In your Shopify dashboard. In the FAQ page nobody reads. In the return policy you wrote six months ago.

The information is there. What doesn’t exist is the system to surface it automatically.

That is exactly what we helped one e-commerce store fix. A support workflow that handles order status lookups, return requests, and frequently asked questions without the owner having to type a single word.

Here is what the manual process looked like before:

Customer sends message —> owner checks inbox —> finds order in Shopify —> copies tracking link —> types reply —> repeat 40 times a day

Here is what it looks like now:

Customer sends message —> chatbot receives it —> classifies the intent (order status, return, FAQ, or other) —> pulls order data from Shopify API if needed —> drafts response from store knowledge base —> sends reply automatically —> complex or sensitive issues flagged for owner review

The owner now handles about 8-10 messages a day instead of 40. The rest are dealt with.

Does it handle everything?

No. Disputes, damage claims, situations where a customer is genuinely upset and needs a real conversation — those still go to a human. But those aren’t the majority. About 80% of incoming support is routine. That is what this workflow handles.

Three things the system covers:

Order status. Customer asks where their order is. The chatbot pulls the order from Shopify using the customer’s email or order number, grabs the tracking link, and replies with it. No human involved. Response time: under 30 seconds.

Return requests. Customer wants to return something. The chatbot confirms eligibility based on the store’s return policy rules, sends the return instructions, and logs the request. If it is outside policy or involves a damaged item, it flags for owner review.

FAQs. Shipping times, size guides, product questions. The store’s existing FAQ content and product descriptions go into a knowledge base. The chatbot draws from that to answer. If it doesn’t know, it says so and escalates. It doesn’t guess.

The stack: Tidio or a similar chatbot widget on the storefront, connected to Shopify via API for order lookups, with Claude handling response drafting from a knowledge base built from store content. Zapier or a direct integration routes flagged tickets to email or Slack for owner review. Setup takes about a week for a typical Shopify store.

The result for the store we worked with: support time went from roughly 2 hours a day to about 20-25 minutes. The owner stopped answering the same questions every morning. Customers got faster responses. Often immediately, even at 2am.

That last part matters more than it sounds. A customer asking where their order is at 11pm on a Sunday is not going to wait until Monday morning happily. An instant, accurate reply keeps the trust. A long silence doesn’t.

E-commerce support is one of those problems that gets worse as the business grows. The more orders you fulfill, the more messages come in. This workflow scales with you. The chatbot handles the 40th query the same as the first.

If you run a store and you’re still answering the same five questions every day, this is worth looking at.

If you want it built for your store, let’s talk. And subscribe — I find the workflows, you focus on growing the business.

Subscribe now

Daily updates to parents. Absence follow-ups for the kids who didn’t show up. Reminders about next week’s event. A note to the family whose child had a rough day.

Most of it lands on the director or the room lead. Typed out one by one. Before 10am.

Sound familiar?

Subscribe now

Childcare is one of those industries where parent communication is everything, and where nobody has built great tools to make it easier. The big childcare apps help a little, but they still require someone to sit down and write. And in a small centre, that someone is usually the person who also has to manage staff, handle arrivals, and actually be present with the children.

That is exactly what we helped one childcare centre fix. A communication workflow that handles the daily update, flags absences automatically, sends event reminders on schedule, and drafts the routine messages so the director can focus on the actual running of the centre.

Here’s what the manual process looked like:

Director or room lead sits down each morning —> writes updates for each age group —> sends individually or through the centre’s app —> checks who hasn’t arrived and follows up with those families —> writes upcoming event reminder —> sends to relevant families —> repeat the next day

Here’s what it looks like now:

End of day: staff fills in a quick daily log (what the group did, any notable moments, meals, naps — about 5 minutes) —> AI drafts a warm parent update from the log —> Director reviews, edits if needed, and sends via the centre’s existing app —> Absent children: system flags them at 9:30am if no check-in recorded —> AI drafts a personal follow-up to that family automatically —> Event reminders: built from a calendar, AI drafts and sends 7 days and 1 day before each event

Does this replace the human touch in childcare communication?

No. The sensitive conversations — an injury, a developmental concern, a difficult family situation — those always need a real person. But about 70% of daily parent communication is routine. This workflow handles the routine so staff have time for the conversations that actually matter.

Three things this setup handles:

Daily group updates. Staff write a short log (3-5 bullet points about the day). The AI turns it into a readable parent message: “The toddler room had a great day today. We spent time in the garden, worked on our colour recognition activity, and the group absolutely loved lunch.” The log takes 5 minutes. Reviewing and sending the message takes about 30 seconds.

Absence follow-ups. By 9:30am, if a child hasn’t arrived and no absence has been recorded, an AI-drafted message goes out to the family: “Hi [Name], we noticed [Child] hasn’t arrived yet today. Just checking in — everything okay? Let us know if they’ll be out today.” Personal-feeling, not a generic form. Zero staff time.

Event reminders. The centre’s event calendar connects to the workflow. One week and one day before any event, a reminder goes automatically to the relevant families. The AI drafts it from the event details. Nobody has to remember to send it.

The stack: a daily log form via Tally or Google Forms, Zapier to route to Claude for drafting, and delivery through Storypark, Kinderloop, or the centre’s existing email setup. For most centres this integrates with whatever communication tool they already use. Setup took about a week.

The result for the centre we worked with: the director went from spending 45-60 minutes on morning communications to about 10-15. Room leads stopped being pulled away from the children to type updates.

And the parents noticed. Updates came out more consistently. Messages felt warmer. Not because a robot wrote them, but because the staff writing the logs could actually focus on what happened that day instead of how to phrase it.

Childcare is one of the most underserved spaces when it comes to operational tools. This is a small change with a real impact on people who are already stretched thin.

If you run a centre or know someone who does, this workflow is worth looking at.

If you want it built for your centre, let’s talk. And subscribe — I find the workflows, you focus on the kids.

Subscribe now

You know this because it’s in your practice management software. It’s been there for months. And nothing has happened with it.

Sound familiar?

Subscribe now

This is the quiet revenue leak most dental practices don’t talk about. Not the patients who cancel and reschedule. The ones who drift. They came in 8 months ago, meant to book their 6-month recall, and life got in the way. Now they’re not on your schedule and nobody has reached out.

That is exactly what we helped one dental practice fix. A recall automation that runs in the background, sends reminders at the right intervals, handles no-show follow-ups, and fills the schedule without the front desk having to work through a list of 200 names by hand.

Here’s what the process looked like before:

Front desk pulls overdue recall list —> calls patients one by one —> leaves voicemails most don’t return —> sends a few reminder postcards —> books whoever responds —> forgets about the rest —> list keeps growing

Here’s what it looks like now:

Patient completes visit —> practice software marks 6-month recall due date —> At 5 months: first reminder SMS sent automatically —> At 6 months: second reminder if still not booked —> At 7 months: final nudge with a direct online booking link —> No-show on appointment: reschedule prompt sent automatically the next morning —> Everything tracked in a simple dashboard

Does this replace the front desk?

No. They still handle the conversations, the insurance questions, the patients who call back with specific needs. This handles the outreach — the part that used to mean 2-3 hours a week working through a recall list that never seemed to get shorter.

Three things this setup handles:

Recall reminders. Triggered automatically from the recall due date in the patient record. Three touches over two months: a friendly first reminder, a follow-up if they haven’t booked, and a final nudge with the booking link. Patients who book after the first message don’t get the follow-ups. No spam.

No-show recovery. When a patient misses an appointment, an automated message goes out the next morning: “We missed you yesterday. Would you like to reschedule? Here’s a link to find a time that works.” Most practices rely on staff to catch this manually. It usually doesn’t happen consistently.

Long-lapsed reactivation. Anyone who hasn’t been in for over a year gets a separate sequence. A simple “we haven’t seen you in a while” message, written to feel personal. The sending is automated. The writing still needs a human touch, but you do it once and it runs.

The stack: most dental practice management software (Dentrix, Eaglesoft, Carestream) can export recall data or connect via integration. Zapier handles the trigger logic. Messages go out via Twilio for SMS, or through the practice’s existing email system. Online booking connects through Jane App or whatever the practice already uses. Setup took about a week, mostly spent connecting the software and writing the three message templates.

The results: within 6 months, the practice’s recall fill rate went from 61% to 84%. That’s 23 more patients per month coming in for hygiene appointments they would have otherwise missed. At an average hygiene visit value of $200, that’s roughly $4,600 per month in recovered revenue.

And the front desk spends about 45 minutes less per day on recall calls.

The data was always there. The software had every recall due date sitting in it. This workflow just made it do something with that information.

If you want to build this for your practice, let’s talk. And subscribe — I find the workflows, you focus on the patients.

Subscribe now

You meant to go through them. But you were also running the business, managing the team, and handling a client situation that came up at the worst possible time. And somewhere in there, the applications just piled up.

Three weeks later, you hired someone through a referral. And the 80 people who applied? You never got back to them.

Sound familiar?

Subscribe now

This is one of those problems that feels too embarrassing to talk about. Small businesses ghost candidates all the time. Not because the owners are bad people. Because they’re understaffed and the hiring process was never designed to handle more than a referral or two.

That is exactly what we helped one business fix. A workflow that acknowledges every application, screens candidates against a criteria doc, and books interviews for the right people — without the owner touching it until someone is actually worth talking to.

Here’s what the old process looked like:

Post job on Indeed or LinkedIn —> applications come in via email —> owner skims when time allows —> some get replies, most don’t —> interviews booked manually over several email threads —> good candidates accept other offers while waiting

Here’s what it looks like now:

Application submitted via form —> auto-acknowledgement sent within minutes —> AI reads the application against a criteria doc and scores it: strong, maybe, or no —> Strong candidates move to a review folder —> Clear mismatches get a polite decline automatically —> Owner reviews the shortlist (12-15 people instead of 80) —> Calendly link sent to shortlisted candidates automatically —> Interview booked without any back-and-forth

What does the AI actually do here?

It reads each application and compares it to a criteria doc you write once. Something like: must have 2+ years in this role, must be based locally, nice to have experience with a specific tool, disqualify if no relevant background. The AI scores each application against that list and categorizes them. You review the ones that made it through.

Does it get it right every time?

No. Someone who looks weak on paper might be great in person. Edge cases exist. The owner still reviews the shortlist. But instead of reading 80 applications, they’re reading 12. And every single applicant has heard from you.

The rejection part matters more than people think. A polite decline, sent within a few days, is a genuinely positive experience compared to silence. Candidates remember that. And if you hire regularly, your reputation as an employer lives partly in how you treat people who don’t get the job.

Three things this setup handles:

Immediate acknowledgement. Every application gets a response within minutes. Not a canned non-answer. A real message that sets expectations: when they’ll hear back, what the next step is. It stops the anxiety, reduces the follow-up emails, and signals that your company is organized.

AI screening. The criteria doc is the key piece. You write it once before you post the job, refine it after the first hire, and it gets better over time. The AI reads each application, checks it against the doc, and categorizes it. Strong candidates get flagged. Clear mismatches get a polite decline. Maybes land in a separate bucket for a human to decide.

Interview scheduling. Shortlisted candidates get a Calendly link automatically. No email chains to find a time. They book directly, the owner gets the invite, done. Average time from application to booked interview: 2-3 days instead of 2-3 weeks.

The stack: a Tally or Google Form for applications, Zapier to route submissions, Claude to read and score against the criteria doc, Gmail for sending responses, and Calendly for scheduling. No special HR software. Setup took about two days.

The result for the business we worked with: faster hires, fewer dropped conversations with strong candidates, and no more LinkedIn messages from applicants wondering if their application was even received.

And honestly, the biggest change was psychological. The owner stopped dreading the hiring process. Because the inbox wasn’t waiting for them anymore.

If you want this built before your next hire, let’s talk. And subscribe — I find the workflows, you focus on building your team.

Subscribe now

And you never hear from them again.

Sound familiar?

The average restaurant loses 60% of first-time customers. Not because the food was bad. Because nobody followed up.

This is the gap. Most restaurants are so focused on getting people in the door that they do nothing after the door closes behind them. And the ones who left quietly unhappy? They go home, open Google, and leave you a 1-star review while the frustration is still fresh.

Subscribe now

That is exactly what we helped one restaurant owner fix. A post-visit workflow that catches unhappy guests before they write that review, sends happy guests a reason to come back, and does all of it automatically.

Here’s what happened before:

Guest finishes meal —> leaves —> owner hopes they come back —> occasionally checks Google Reviews —> discovers a 1-star review from three days ago —> nothing left to do but respond publicly and move on

Here’s what it looks like now:

Guest pays —> receipt system triggers a short SMS within the hour: “How was your visit today?” with a 2-question form —> AI reads the response and classifies it: happy, neutral, or unhappy —> Unhappy guests get a personal message from the owner within minutes + a make-it-right offer —> Happy guests get a thank-you and a return offer (10% off next visit) —> All responses log automatically to a shared dashboard

The whole thing runs without anyone touching it. The owner checks the dashboard once a day.

Does the AI write everything perfectly?

No. Recovery messages sometimes need a personal tweak. But the draft is there, the response is fast, and that’s what matters. Most unhappy guests just want to feel heard. Getting a message within an hour of leaving does that.

Here’s the thing most people don’t know: the majority of unhappy restaurant guests don’t complain in person. They leave. And then they vent online. A single follow-up message, sent quickly, changes that for a lot of them. Not all. But enough.

Three things this setup does:

Feedback collection. A short form linked in the follow-up SMS. Two questions max. “How was your experience?” (1-5) and one open text field. Simple enough that people actually fill it out.

Sentiment routing. The AI reads the free-text response and classifies the mood. Negative responses trigger an immediate alert to the owner and a draft recovery message. Positive responses get a different flow: thank-you plus return offer. The routing happens automatically, no one has to read every response.

The return offer. A discount code sent automatically to guests who rated their experience well. No manual list, no forgetting to send it. Goes out the same night of their visit.

Look, you don’t need a CRM or a loyalty program or a marketing team to do this. You need a form, a Zapier workflow, and about a weekend of setup.

The stack: Tally for the feedback form, Zapier to route responses, Claude to classify sentiment and draft replies, Twilio for SMS delivery. Under $50/month to run for most restaurants. No developer needed.

The results for the restaurant we worked with: 200-plus direct feedback responses in the first 60 days, 11 complaints resolved before they hit Google, and a 22% increase in repeat visits from first-timers who received the return offer.

None of that happened because they hired someone. It happened because they stopped letting the conversation end at the door.

If you want this built for your restaurant, let’s talk. And subscribe — I find the workflows, you focus on the food.

Subscribe now

And every time a grant opportunity comes up, someone spends weeks writing an application from scratch. Researching the funder. Matching language. Making sure the narrative fits the requirements.

Sometimes the grant is $5,000. And the application takes 40 hours to write.

Sound familiar?

This is the part of nonprofit work nobody talks about: the writing tax. Every grant needs the same information about your organization, told in slightly different ways, formatted to meet slightly different requirements. And because there’s no dedicated grant writer, it falls on whoever has the most writing experience. Which usually means the executive director. Which means everything else waits.

Subscribe now

We helped a small nonprofit cut their grant writing time from 3 weeks to 3 days. Not by cutting corners. By building one document that does most of the work upfront.

Here’s what the process looked like before:

Grant opportunity found —> staff member reads requirements —> starts writing from scratch —> pulls past applications for reference —> spends days matching funder language —> multiple revision rounds —> submit, exhausted

Here’s what it looks like now:

Grant found —> requirements pasted into AI alongside org profile doc —> AI drafts a first application matching the funder’s language and priorities —> staff edits for 2-3 hours —> review —> submit

What’s the org profile doc? It’s a master document with everything the organization knows about itself:

• Mission, history, and founding story

• Programs with specific outcomes and numbers

• Demographics served

• Budget information

• Past grant successes and what they funded

• Staff bios

• Testimonials and quotes from people served

You build it once. You update it once a year. Every application drafts from it.

Does the AI produce a submission-ready grant?

No. Someone still reviews it. Someone still adds the specific metrics the funder asked for. But starting a grant application used to take a full day just to get a rough first draft. Now it takes about 45 minutes.

The part I’ve seen trip people up: the org profile doc needs to be solid. If it’s thin or outdated, the AI produces thin content. Garbage in, garbage out. The real investment is building that document properly the first time. After that, it pays off on every application.

Three things this setup handles:

First draft generation. Paste the grant requirements and the org profile into Claude with a prompt like: “Draft a grant application for this opportunity based on our organization profile. Match the funder’s priorities and language.” A solid first draft comes back in a few minutes.

Language matching. Funders have specific priorities in their RFPs: “economic mobility,” “underserved communities,” “capacity building.” The AI picks up on those terms and weaves them into the draft naturally. That kind of careful alignment used to take hours of deliberate, manual work.

Boilerplate sections. Organization background, executive summary, statement of need — these sections are nearly identical across every application. The AI writes them fast and consistently from the org profile. Staff time focuses on the sections that actually require customization: the project plan, evaluation approach, and budget narrative.

The stack: a well-maintained Google Doc as the org profile, Claude or GPT-4 for drafting, and a simple prompt template saved for each grant type. No special software. No subscription beyond what most small nonprofits already have access to.

The result: the nonprofit we worked with went from submitting 4-5 grant applications per year to 12-15. Same team. Same hours. They just stopped spending those hours staring at a blank page.

That’s a lot more shots at funding. Without burning out the one person who knew how to write.

If you’re running a nonprofit and still writing every application from scratch, this is one of the most practical things AI can help with right now. The technology is there. The main thing you need to build is the org profile doc.

If you want help setting that up and building the prompting workflow around it, let’s talk. And subscribe — I find the workflows, you focus on the mission.

Subscribe now