Hiring an AI developer? Start with how you’ll win customers

Most businesses looking for an AI developer don’t need a developer. They need to be more competitive. That means customers finishing a quote, a booking or a sign-up without waiting, getting a real answer at 10pm, and choosing you over a rival. Decide what should be better for your customers, and how you’ll measure it, before you decide who to hire.

For most businesses starting with AI now, we’d get there with a partner first, then hire someone to own it. LLMs and agents are new, the tools change every month, and one person can’t carry all of that alone. If AI is already core, continuous work at the heart of what you sell, hire sooner.

Your customers have already moved

With LLMs and agents this useful, we’re seeing customer automation turn up inside the software people already use. The app answers properly. The form fills itself in. The quote arrives while the customer is still interested.

Once people get that elsewhere, a three-day email chain looks slow. A better, more competitive experience lands far better than another internal efficiency drive. The useful question isn’t “should we use AI?” It’s “how far is our product from that, and what would closing the gap be worth?”

Start with what should be better

“We need AI” isn’t a brief. An outcome is. A good one names something customers do, a number and a date you’d be happy to be held to.

Compare two briefs. “Add AI to the website” gets you a chatbot nobody uses. “Let customers get a firm quote in two minutes instead of two days, and convert more of them” gets you a project you can judge.

One example from our own work. For Everyone’s Invited, we built AI anonymisation of submissions into the service itself. It saved about 12,500 hours of review, and 52,794 testimonies have been published.

Where it lands with customers

The first real win is usually a customer journey that already exists but stalls.

• Quoting. An agent asks the questions your sales team always asks, then produces a quote within limits you set.
• Booking and changes. The customer says what they need, and the agent checks availability, books it and confirms.
• Onboarding. It collects what’s needed, checks it’s complete and chases the gaps, so new customers get started sooner.
• Answers and support. Real answers from your own information, inside your app or site, with a person behind anything tricky.
• Finding what you sell. Customers describe what they want in their own words, rather than guessing your menu.

Our guide to AI for UK SMEs goes further on agents for customers. For a worked support example, see AI customer service for small business.

How far away are you?

Most businesses are closer than they think, or further than they hope. Four questions place you.

1. Do customers already do the job with you online? If quoting or booking happens by phone and email, that’s the first step, and AI can help build it.
2. Can your systems be reached? An agent needs to look things up and make changes. That means your booking, stock, pricing or customer records can be connected to safely.
3. Is your information in a fit state? AI answers from what you give it. If prices, policies and product details are scattered or out of date, that comes first. Our guide to getting your data ready for AI explains what “ready” looks like.
4. Do you know where customers drop off? If you can’t see where they give up today, you can’t prove the new version is better.

Four yeses, and you’re close. Two or three noes are normal, and they’re exactly what a partner should help you close first. Our AI readiness scorecard is a quick way to check.

What it takes to get there

The model is the easy part. You can rent a very capable one. The outcome depends on everything around it.

1. One customer journey, measured. Pick the one where customers wait or give up most. Record today’s numbers, so there’s something to beat.
2. The foundations. Connect the systems, tidy the information it answers from and decide what it’s allowed to do.
3. A way to check its work. Test answers before customers rely on them, and keep a person behind anything that costs money if it’s wrong.
4. A working version with real customers. Start with a slice of them, then compare with today’s numbers.
5. Rollout and measurement. Open it up, then keep measuring. An outcome you stop measuring quietly stops being one.
6. Someone to own it. It needs looking after as your products, prices and the tools change.

Steps one to three are the foundations. They’re also where a single new hire is most likely to struggle.

What usually goes wrong

Most customer-facing AI that disappoints doesn’t fail on the technology. It fails on the order of things.

• Starting with the tool. Someone adds a chatbot, then hopes customers find a use for it.
• No number to beat. Without today’s conversion or wait time, nobody can say whether it worked.
• An agent that can’t do anything. It talks well but can’t book, quote or change a thing, so customers still end up on the phone.
• Nobody checking its work. It runs unwatched until a customer finds the mistake for you.
• Nobody owning it. Prices change, the agent doesn’t, and it quietly starts getting things wrong.

Partner, contractor or hire?

Judge each option on the outcome, not the job title.

Why we’d start with a partner, right now

• One person carries everything. They become the only one who understands the system. If they leave, the knowledge leaves with them.
• You can’t easily judge them. If nobody in the business has done AI work, a good hire and a confident one look the same in an interview.
• The first job is foundations, not code. A team that has connected systems and set up checks many times does it faster.
• A partner isn’t tied to one tool. They see many projects and can pick what fits yours.

None of that rules out hiring. It changes the order. Start with a partner to get a working outcome, then hire someone to own it. They’ll join a system that works, with a handover, rather than a blank page.

For the wider comparison, see agency, freelancer or fractional CTO.

How to tell a partner will deliver

Ask these before you sign anything:

• What will be better for our customers, and by how much? “It’ll improve engagement” isn’t an answer.
• What’s today’s number, and who measures it? No baseline means no proof.
• What’s the smallest useful first version? A good partner shrinks the scope before growing it.
• What can the agent actually do, not just say? Quoting, booking or changing a record is where the value is.
• Who checks its work, and what happens when it’s wrong? A named person beats “the model is very accurate”.
• Where does our customers’ data go, and who owns it all afterwards? That covers the code, the accounts and the prompts, the written instructions the AI follows. Our AI and GDPR guide and who owns your code cover the detail.

A partner who answers these plainly has been asked before. One who calls them “a bit much” has told you what you need to know.

What it costs, and what it’s worth

Set the cost against what the outcome is worth, not against a salary.

In our experience, an AI agent can cost as little as £500 to set up. Realistically, £5,000 is the lower bound for a well-managed, organised project. Where yours lands depends mostly on the state of your data and what you want it to do. Our guide to what an AI agent costs to build and run covers the running costs too.

Then do the sum the other way. If one more customer in ten finished a quote, what would that be worth over a year? If customers stayed a month longer on average?

For comparison, the ONS puts median full-time pay for programmers and software development professionals at about £59,900 in 2026, on provisional figures. That’s before employer costs, recruiting and the time it takes them to learn your business. If you use a contractor instead, check whether the off-payroll rules, known as IR35, apply.

Where Contic fits

If you want AI to make your product more competitive and win more customers, start with a free chat. You’ll get guidance and advice on what we’d do, from people who do this all the time. We’ll tell you plainly how far away you are, and if you don’t need a developer at all. An architectural review is free too.

A paid discovery doesn’t have to follow. When it does, it can end in a strategy you take away and run with yourself, or in a scoped plan and a price. Either costs far less than getting it wrong.

Our founder was the founding engineer at AI Music, which Apple acquired. We work AI-augmented, and senior engineers review and own everything that ships. Our own team builds it, with a named technical lead and a named product lead. We hold Cyber Essentials Plus and are registered with the ICO.

You can read more about our AI work, or talk to us.