Home Work Services
Website Design & Development Shopify Design & Development Google Ads Management SEO & Local Optimisation Social Media Management Software & App Development AI Agents & Automation Brand Design
About Process Blog Start a project Say hello tristan@clickfusion.cloud
A developer at a desk with two monitors of colourful code in a dark room, focused, headphones on

How AI has made custom software cheaper and faster (and what it still can't do)

AI coding tools have made custom software cheaper because they cut the hours a developer spends on routine work: standard code, first versions, connections between systems, admin screens and tests. That puts a booking system, quoting tool or client portal within reach of a small business. Design, security, data protection and upkeep still need an experienced developer.

At Click Fusion Cloud we build with AI coding tools every day. They are genuinely good at some jobs and quietly bad at others, and knowing which is which decides whether the software you pay for is an asset or a liability.

What does AI actually make faster in software development?

Most of the cost of custom software development is developer time. Anything that takes hours off the job takes money off the quote. These are the jobs where AI saves the most:

  • Boilerplate. Every app needs the same plumbing: log-in pages, forms, database tables, settings screens. Now a developer describes it and reviews what comes back.
  • Prototypes in days. A clickable first version of a quoting tool or a customer portal can be in front of you within days, using your real services and prices. You see it and change your mind early, when changes are cheap, instead of after weeks of work.
  • Integrations. Connecting your booking system to your calendar, or your website forms to your accounts software, means reading another company's documentation and writing the glue between the two. AI reads documentation quickly and drafts that glue well.
  • Admin dashboards. The internal screens where you search customers, change a booking or export a month's jobs are important but rarely clever. AI builds them fast.
  • Tests. Automated tests check the software still works after every change. They were often the first thing cut from a small budget. AI makes them cheap enough to write properly.
  • Moving off old systems. Plenty of small businesses run on an ageing Access database, a tangle of spreadsheets or a bespoke system whose developer has long gone. AI is good at reading old code and old data formats, which makes it far quicker to work out what the old system does before replacing it.

What does the research say about AI and developer speed?

Less than the headlines suggest, and the studies point in different directions.

The study behind most "55% faster" claims is a 2023 controlled experiment by researchers from Microsoft, GitHub and MIT. They recruited 95 freelance developers through Upwork and asked them to build a simple web server in JavaScript. The group with GitHub Copilot finished 55.8% faster than the group without it. The authors are clear about the limits: it was one small, standardised task, not work on a large project, and the study did not measure code quality, including security.

A 2025 randomised trial by METR found the opposite in a different setting. Sixteen experienced open-source developers worked on 246 real tasks in large projects they knew well, using early-2025 tools. With AI allowed, tasks took 19% longer. The developers had expected AI to cut their time by 24%, and afterwards believed it had cut it by 20%.

METR ran a follow-up and reported in February 2026 that developers are probably faster with AI now than in early 2025. It also said its new data is only very weak evidence of how much faster, partly because many developers declined to take part because they no longer wanted to work without AI.

Read together, the gains are largest on new, well-defined work and smaller on big, mature systems, and feeling faster is not the same as being faster.

Sources: Peng et al., The Impact of AI on Developer Productivity: Evidence from GitHub Copilot (2023); Becker et al. (METR), Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity (2025); METR, uplift update (February 2026).

What can a small business now afford to have built?

Tools that used to make sense only for firms with an IT department. A few examples of the kind of thing that is now realistic on a small business budget:

ToolExampleWhat it replaces
Quoting toolA kitchen fitter enters measurements and chosen units, and gets a priced quote with your terms, ready to sendAn evening of spreadsheets after each survey
Booking systemA dog groomer's booking page that knows a large double-coated breed needs a longer slot and the van can only do two a dayBooking apps that can't follow your rules, plus a lot of messages back and forth
Client portalAn event planner's couples log in to see the timeline, approve suppliers and upload the seating planLong email threads and attachments nobody can find
Small CRMA heating engineer tracks each property, its boiler, service dates and certificates, with reminders when a service is dueA generic sales CRM forced to fit, or a whiteboard

Off-the-shelf software is still the right answer more often than not. If a standard product fits how you work, use it. The AI already built into everyday tools, covered in what AI can actually do for a small business in 2026, may be enough on its own, and so may the simpler automations in ten jobs you can hand to AI this month. Custom earns its place when your process is unusual or the workarounds are costing you time. We go through that decision in custom CRM vs off-the-shelf.

What still needs an experienced developer?

AI writes code. It doesn't take responsibility for it. These are the parts where experience matters more than speed.

  • Architecture. How the data is structured, where it lives and how the pieces fit decides whether the system copes with year three or has to be rebuilt. AI tends to solve the problem in front of it, not the one coming next.
  • Security. A study by Stanford researchers, published at the ACM CCS conference in 2023, found that participants with an AI assistant wrote less secure code on four of five security-related tasks, and were more likely to believe their code was secure. Most participants were students, which the authors note as a limit, but the pattern is the one to watch: confident code that hasn't been checked.
  • Data protection. UK GDPR requires data protection by design and by default: collect only what you need, keep it only as long as you need it, and limit who can see it. If personal data is breached and it puts people at risk, you must tell the ICO within 72 hours where feasible. Those duties sit with your business, whoever or whatever wrote the code.
  • Edge cases. Two customers booking the last slot at the same moment. A bank holiday Monday. The clocks going forward. A refund on a part-paid deposit. AI handles the normal path well. The unusual paths are where real businesses lose money.
  • Reviewing AI output. Every line AI writes needs reading by someone who could have written it. AI-generated code can look tidy and still be wrong in ways only experience spots.
  • Hosting and upkeep. Software needs a server, backups, security updates and someone to call when it stops working. None of that goes away because the first version was quick to build.

Sources: Perry et al., Do Users Write More Insecure Code with AI Assistants? (CCS 2023); ICO, data protection by design and by default; ICO, report a personal data breach.

Is it safe to build your own app with AI?

For some things, yes. "Vibe coding", where you describe what you want and let an AI tool build it without reading the code, is fine for a personal tool: a calculator for your own pricing, a page that tidies a spreadsheet, a prototype to show a developer what you mean.

It gets risky the moment the app holds customer data, takes payments or lets people log in. The usual problems with DIY apps are rarely visible on screen:

  • Secret keys for payment or email services sitting in code that anyone can read in their browser.
  • A database that anyone with the address can read, because access rules were never set.
  • Admin pages with no log-in, found by guessing the web address.
  • No backups, so one bad change or one failed server loses everything.
  • Nobody who understands how it works, so the first serious bug stops the business.

Before a DIY app goes live with real customers, ask a developer to check five things: where the secret keys are kept, who can read and change each part of the database, that every admin page needs a log-in, that backups run and can be restored, and what personal data it stores and why. It's a short job, and it's far cheaper than a breach.

What should you ask a developer who builds with AI?

  1. Who reviews the AI's code? The answer should be a named, experienced person, not "the tools are very good now".
  2. Who owns the code and the data? You should, with access to both in a form another developer could pick up.
  3. What does the first version include? Start with the smallest version that does one job properly, then add to it.
  4. How is personal data handled? Where it's stored, who can see it, how long it's kept.
  5. What happens after launch? Hosting, backups, updates and support, and what each costs.

AI has changed what a small business can sensibly have built, and that is good news. The faster building is real. The judgement around it is still the part you're paying for. If you have a tool in mind, Click Fusion Cloud offers software development in Hertfordshire and across the UK, starting with a conversation about whether it's worth building and a fixed quote before anything is built.

Quick answers

Has AI made custom software cheaper?

Yes, for most small business projects. Most of the cost of custom software is developer time, and AI coding tools cut the hours spent on routine work like standard code, prototypes, integrations, admin screens and tests, which brings tools like booking systems and client portals within reach.

Does AI make developers faster?

Often, but not always. A 2023 experiment found developers using GitHub Copilot finished one small task 55.8% faster, while a 2025 METR trial found experienced developers on large, familiar projects took 19% longer with early-2025 AI tools, so the gains depend heavily on the kind of work.

Is it safe to build a business app myself with AI?

It's fine for a personal tool or a prototype, but risky once the app stores customer data, takes payments or has log-ins. Common problems include exposed secret keys, open databases, unprotected admin pages and no backups, so have a developer check it before real customers use it.

What still needs an experienced developer when using AI?

Architecture, security, data protection under UK GDPR, unusual edge cases, reviewing everything the AI writes, and hosting and upkeep after launch. AI writes code quickly, but it doesn't take responsibility for it.

Want this done for you?

Software & App Development from Click Fusion Cloud

Built by our studio in Hertford, Hertfordshire, for businesses across the UK.

See our software and app development service

Start a project

Got an idea?Let's make it happen.

Tell us what you're planning and we'll come back with ideas and a clear quote. You'll hear from a real person, usually within a day.

  • Free, no-obligation quote
  • Reply within 24 hours
  • Payment plans available
Tristan, founder of Click Fusion Cloud

Tristan
Founder, Click Fusion Cloud


By sending this you agree to our privacy policy.