Most small businesses run on two things: a spreadsheet and a WhatsApp group. The spreadsheet holds the jobs, the customers and the hours. The group chat is where the work actually gets handed out. "Can someone cover Mrs Patel's leaking tap at 2?" "Who's got the van keys?" "Send me your hours by Friday."
It works, until it doesn't. Messages get buried. Two people edit the sheet at once and one version wins. Someone finishes a job and forgets to mark it off, so the invoice goes out a week late. At the end of the month, somebody spends a Saturday scrolling back through the chat to work out who did what.
Until recently the fix was either off-the-shelf software or a developer, and many owners couldn't justify either. That has changed. AI app builders can now build a working internal tool from a plain description, and small teams are using them to replace the spreadsheet and the group chat with an app that fits how they already work.
What an AI app builder actually does
You describe the app in ordinary words, the AI writes it, and you check the result on your screen as it goes. "Each technician signs in and sees only their own jobs for today. They can add photos, mark a job done and log their hours. I see everything on one page." A few minutes later there is something you can click through and change.
That conversation can also happen in the AI assistant the business already uses. With a tool like Floot, you type the description into ChatGPT or Claude, and Floot builds and runs the app behind it.
The difference from a spreadsheet is that it's a real app: each person has their own login, the data sits in a proper database rather than a shared file, and it works on a phone without pinching and zooming.
A worked example: a seven-person maintenance business
Picture a property maintenance business with an owner, an office manager and five technicians. Today, jobs arrive by phone and email, go into a Google Sheet, and get handed out in a WhatsApp group. Hours come in by text on Friday afternoon.
Here is the app they would build to replace that:
Jobs. The office manager adds each job with the address, the problem and the customer's number. It shows up on the assigned technician's phone, not in a group of seven.
On site. The technician taps the job, gets directions, adds before and after photos, notes the parts used and marks it done.
Hours. Each job has a start and finish button, so timesheets build themselves. No more Friday texts.
The office view. One screen shows what's open, what's done and what's ready to invoice, with an export the bookkeeper can use.
Notifications. A new or changed job sends the technician a notification on their phone, so nobody has to scroll a chat to find it.
None of this is unusual. It's what job management software does. The point is that it does exactly what this business needs and nothing else, and the owner can change it next month without a support ticket.
Built this way, the technicians install it from the App Store or Google Play like any other app, sign in with their own account, and get a notification when a job lands. The owner never touches a server, a database or a line of code.
How to build one in an afternoon
1. Write down what you do today. List the steps of a typical job from first call to invoice, and who touches it at each step. This list becomes your description to the AI. Be specific about who should see what: "technicians see only their own jobs" matters more than any colour scheme.
2. Describe it and build the first version. Start with the smallest useful version, such as jobs and sign-in, and add hours and photos once that works. AI tools do best with one change at a time. If you're using ChatGPT or Claude, you can connect it to Floot in a couple of minutes and build in the same chat window you already use for emails and quotes.
3. Move your data in. Export your spreadsheet as a CSV, attach it in the chat and ask the AI to import it. Check a handful of records by hand.
4. Test it with two people for a week. Pick the most patient technician and the office manager. Keep WhatsApp running alongside so nothing is lost while you find the gaps.
5. Switch over. Once the week goes smoothly, move everyone across and make the app the single place jobs live.
Choosing a tool
There are three broad routes, and each suits a different kind of business.
A staff app has a few needs a simple form doesn't: a login for each person, notifications that reach their phone, somewhere to keep photos, and an app the team can install rather than a link they have to bookmark. There are three broad routes, and they differ most on exactly those points.
Spreadsheet-based app builders such as Glide, Softr and Google's AppSheet turn your existing sheet into a simple app. They are quick to start and a good fit if what you mostly need is a nicer way to view and edit the same data. They get harder to stretch once you want per-person views, notifications or anything the spreadsheet wasn't designed for.
AI app builders such as Lovable, Bolt and Replit build the app from a conversation in their own editor. They give you more freedom than a spreadsheet-based tool, but you learn a new editor, and you usually pay through the builder's own AI credits, so every change costs something.
Building from the AI assistant you already use. This is the route Floot takes. You describe the app in ChatGPT or Claude, and Floot provides everything it needs to run: the database, staff sign-in, file storage for photos, email, phone notifications and hosting. It can also publish the app to the App Store and Google Play, so staff install it like any other app. Because the building runs on the ChatGPT or Claude subscription you already have, changing the app later doesn't cost extra credits. There is a free plan to try it on one real problem first.
Whichever route you choose, check four things before you commit: whether each staff member gets their own login, where the data is stored, what it costs per month once you're past the free tier, and whether you can export your data if you leave.
The honest trade-offs
Someone has to own it. An app you build is yours, which means somebody in the business needs to look after it. It doesn't take a developer, but it does take an hour here and there when you want something changed. The upside is that a change is usually one message in the chat, not a feature request to a software company.
Off-the-shelf might already fit. If your business is a standard trade, tools like ServiceM8 and Tradify were built for exactly this and include things like quoting and accounting integrations out of the box. Build your own when the off-the-shelf options force you into a workflow that doesn't match yours, or when you're paying for ten features to use two.
Staff data deserves care. Timesheets, phone numbers and customer addresses are personal information. Give everyone their own login, make sure people only see what they need, and don't leave the old spreadsheet shared with former staff.
Start smaller than you think. The most common mistake is trying to rebuild the whole business on day one. Replace one painful thing, usually job handout or timesheets, get the team using it, then add the next piece.
The bottom line
Spreadsheets and WhatsApp aren't the problem. They're how most good small businesses got started. They just weren't made for running a team of five or ten people every day. What's changed is that building something better no longer needs a developer or a big software contract. If you can describe how your business works, you can now build an app around it. Pick the one job your team complains about most, try it on a free plan this week, and see whether the next Saturday is spent somewhere other than the group chat.
Keep up to date with our stories on LinkedIn, Twitter, Facebook and Instagram.