Can AI Coding Tools Really Build You a CRM? (Cursor, v0, Bolt)
It’s Sunday evening and you should be switching off. Instead, you’ve just watched someone build a CRM in under ten minutes by typing a few sentences into an AI coding tool. Contacts, pipeline stages, a tidy revenue dashboard, the lot. Then you remember the hundreds of pounds you pay for software each month, and the messy spreadsheet your team still prefers. Could you just build something better yourself? It’s a fair question, and the demo makes it look easier than cancelling your broadband.
Can an AI CRM tool actually build you a working system?
Yes, it can. Cursor, v0 and Bolt will all get you a CRM with contacts, pipelines, a database and secure logins. Whether your business can lean on it is another matter, and that comes down to planning, testing and support.

What “working” really means
Much depends on what you mean by working. If you can type a test customer’s name into a box and watch it save, congratulations, that’s working of a sort. A team handling fifty new enquiries a day is asking for rather more: deleted records they can get back and follow-up messages that still go out while you’re on holiday in Portugal.
AI coding tools can get you to either place. The second one just takes far more time and technical effort, and the work doesn’t stop once it’s built. The video you watched, funnily enough, tends to end before any of that starts.
Cursor, v0 and Bolt do different jobs
People tend to lump these three together, which is a bit unfair on all of them. Cursor’s really aimed at developers. It lives in the files on your own computer, where it’ll read your code, add to it and run commands while you get on with something else. If you know your way around software, you’re very much in the driving seat, and when something breaks you can roll it back. But a person in your business still has to look at the code and judge whether what changed is safe to use.
v0 is known for producing visual web pages quickly, although calling it a design tool undersells it. You can build interactive screens, link them to a live database and set up the logic for storing customer details securely. Bolt, meanwhile, lets you build whole applications and databases in your web browser, with ready-made features for user accounts, passwords and security settings. As with v0, those features won’t set up and test themselves, so you’ll need someone who knows what they’re doing.
Your choice really comes down to who’ll build the thing and who’ll look after it afterwards. Raw code might suit a technical colleague, and a browser-based builder is easier for a quick experiment. What none of them can tell you is whether building your main customer database is a sensible use of your time.
Simple, everyday requests that make software complicated
Picture a small recruitment agency. Version one of their custom software stores employers, candidates and live vacancies, and for about a day everyone’s pleased with it.
On day two, a recruiter wants to link one candidate to three vacancies without creating three duplicate files. A manager then asks for certain interview notes to be hidden from junior staff. By the end of the week somebody has imported an old spreadsheet with the same client company spelt three different ways.
None of this is unusual. Every one of those requests needs clear rules, though, about how data links together, who can see which notes and what happens when two bits of information conflict. Email makes it messier still. When a client writes in, should the message sit on the candidate’s file, the employer’s account or the vacancy? And what does a colleague see when they’re covering for someone off sick?
AI can write the code for all of that. Deciding the rules, testing each scenario and checking the app sticks to them every time is still down to you.
The bigger question: what problem are you actually trying to fix?
All of which raises an awkward point about why you wanted to build in the first place. Most people who fancy their own system are, if we’re honest, fed up: with missed follow-ups, with mandatory fields nobody wants to fill in, with a team that won’t use the software properly. That frustration is real, but it’s worth working out how much of it comes from the software and how much from the way your business runs.
Is it the software or the setup?
Your current system might just have the wrong sales stages built into it. Perhaps updating a deal takes five clicks instead of one, so people quietly stop doing it, or the screen is full of boxes nobody can explain. A custom build promises to sweep all that away, and occasionally it’s exactly what a company needs. More often, the platform you’ve got needs a better setup, less clutter and clearer responsibilities for your staff.
A shiny new interface won’t decide who picks up the phone
Take a small plumbing and heating firm. The owner emails a quote to a homeowner on Thursday afternoon. The office administrator assumes the owner will call to follow up, while the owner assumes the administrator already has. By the following Wednesday nobody’s spoken to the customer, who has by then gone with a competitor.
An AI-built system could show that quote on a beautifully designed screen, and the gap would still be there. What would close it is some plain process design: a named owner, a follow-up date and a recorded next action on every open quote, plus agreement on when to stop calling, when a quote counts as lost and what gets written down after a call.
That’s usually where we start at Waggle Dance. We look at what your team does during a hectic week, and design steps people will actually follow. Sometimes this ends in new software. Other times it ends with fifty unnecessary boxes deleted and a decision about who makes the next phone call.
Owning software means owning the maintenance work
Building your own system is a bit like fitting out your own office. You get the desks exactly where you want them, and you’re also the one on the phone when the plumbing leaks. With custom software, that means paying a developer on a support contract or leaning on someone in your team who has the time and the skill to fix bugs.
Who looks into it when a web enquiry never reaches your inbox? Who checks that a browser update hasn’t broken your booking form, or removes someone’s access the day they leave? AI can help write the fix, but a person still has to notice the problem, weigh up the suggestion and confirm everything’s working again. In-house technical support makes this manageable. If the answer to all of those questions is “the founder”, add that workload to your sums.
How to decide without wasting a month building the wrong thing
If you’d still like to explore a build, and plenty of people will, start with one operational headache and test a basic fix against your current tools before your team has to rely on it. These five steps should keep the experiment small.

1. Choose one piece of work
Just the one job, mind. It might be chasing open quotes, or the way website enquiries get picked up, or those overdue tasks everyone keeps walking past. Jot down what happens today and what you’d like to be different. “Every open quote must have a named owner and a dated next action” gives you something to test. “We need a modern system” doesn’t. Agree what success looks like before you start, too. Could a colleague sit at your computer and find every quote due for a follow-up today without asking you a single question?
2. Build a small prototype with dummy data
Build only what that one task needs, and keep real customer files out of it while you’re experimenting. Then give it to a colleague and don’t explain the screen. Feed it awkward data: two customers with the same name, an enquiry with no email address, a quote reopened after being marked lost. Watch where they get confused, because if they can’t tell who owns an enquiry, another colourful chart on the dashboard isn’t going to help.
3. Compare it with your existing options
Run that same task through your current setup, a properly configured commercial platform and your prototype. How quickly can someone finish the work, see the customer’s history and hand it to a teammate? You may find your existing platform could do this all along and nobody configured or explained it properly. Or you might uncover a requirement that commercial platforms handle badly, which is a genuine reason to consider going custom. There’s a sensible middle option as well: a small specialist tool that does one job and connects to your main database.
4. Name the person responsible after launch
Before anything goes live, decide who fixes bugs, updates settings, manages user accounts and answers staff questions. Work out the full running costs while you’re at it, including hosting, database storage, SMS and email fees, outside technical support and the value of your own time. Setting the monthly price of an AI tool against a software subscription leaves out most of the real work.
And can you get your data back out? Try it. One day you might want to move to different software or a new hosting provider, and a pile of customer records you can’t open or use won’t be much comfort. As for maintenance, if nobody’s got the time or the know-how, treat the build as an experiment for now.
5. Test permissions, data recovery and automated messages
A password screen doesn’t make your customer data safe. Arrange a security check that suits the information you hold, and make sure regular staff can’t wander into private management notes or export the full customer list to a personal device.
Backups need testing too. Knowing a file exists somewhere in the cloud is nice, but can you actually restore your records within twenty minutes of a crash? Then look at your email rules. What happens if the internet drops mid-send? Could a customer get the same reminder three times in a row? Does an unsubscribe carry across all your sales lists?
Then there’s the legal side, which nobody ever puts in the demo video. If you’re in the UK and you hold people’s personal details, UK GDPR and the marketing rules apply to you. Off-the-shelf software will cover a good chunk of those checks. Build your own and they’re all yours to sort out.
Today's Deep Dive
What to do next
A custom AI-built system can be a brilliant investment if it solves a truly unique need and you can support it properly. For most service businesses, though, setting up a proven platform gets the team working productively much faster. Where you go from here depends on which of these sounds most like you.
You need a CRM configured around your service business
If your needs centre on capturing enquiries, managing pipelines, automating communication and keeping up with follow-ups, a configured platform is your best route. Waggle Dance CRM is built on GoHighLevel and configured for UK service businesses, at £149 per month plus VAT. A quick live demo will show you whether it fits how your team works, with no obligation.
You want a better setup for the software you already pay for
Already paying for licences? Then let’s look at why your staff are working around the system. It could be confusing pipeline stages, dozens of unnecessary required fields or automated reminders nobody trusts. Our setup and optimisation service fixes those headaches without making you switch software.
The real issue is your wider business processes
Sometimes software has little to do with it. Follow-ups get missed because roles overlap, everyone’s busy and nobody has time to review what’s slipping. Our £499 per month plus VAT coaching bundle includes Waggle Dance CRM and fortnightly business coaching, which gives you dedicated time and guidance to refine daily processes, assign clear roles and build habits that last.
FAQs
Can I use Cursor to build a system without being a developer?
You can, in the sense that plain prompts will generate and change application code. The tricky part is knowing whether that code is secure, reliable and easy to maintain. If you can’t judge its quality yourself, line up technical support before any real customer data goes in.
Can v0 or Bolt build more than just a software mock-up?
Yes. Both support live applications connected to real databases. You’ll still have to specify, configure and test every feature your business relies on, user access levels and automated messaging rules included.
Is building a custom tool with AI cheaper than buying software?
At first it can be, particularly for a tiny internal utility. Over two or three years the picture changes, once you add hosting, messaging services, technical fixes, updates and your team’s time. How fast you built it won’t tell you what it costs to run.
What is the difference between Waggle Dance CRM and GoHighLevel?
GoHighLevel is the underlying engine. Waggle Dance CRM delivers it pre-configured and streamlined for UK service businesses, with expert support behind it. So when you’re comparing systems, look at the setup help and ongoing support as well as the raw features.
Can Waggle Dance help us if we want to keep our existing software?
Yes. We offer platform setup, optimisation and business coaching independently of our software packages. Everything starts with how your team manages its daily work, and switching to our software is never a requirement.
Still weighing up whether to build a custom tool, buy a new platform or fix the software you already own? Book a free Clarity Call with our team and bring along one real task that keeps getting missed in your business. We’ll help you work out the best way to fix it.



