AI systems for UK service businesses

Less admin.
More profit.

We build the systems that quote in minutes, chase every follow-up and join up your CRM, diary and accounts — then prove, in pounds and hours, what they saved.

A 60-second tour · always on, never off sick · then run your own numbers

or start with the 2-week diagnostic →

Fixed fee. Two weeks. You keep the roadmap either way.

Taking on new work now — built a few days a week, inside a real business we already run.

The interactive bit — this is what we do, in miniature

Drop AI into a business.
Watch the numbers move.

Meet Harborough Drains — a fictional firm doing about £700k a year. Add the systems I actually build — drag on a desktop, tap on a phone. Watch who does each step, and what the swap costs against what it saves.

EnquiriesQuotesBookingJobsInvoicesAccountswaiting…The brain

Tap a system to install it — or drag it onto the model.

Who does each step?16 of 16 steps done by a person

Enquiries Log & re-key each call Remember who to ring back
Quotes Work out each price by hand Type up & send the quote Chase it before it goes cold
Booking Phone tag to find a slot Diary it & confirm by text
Jobs Print & hand over job sheets Ring the engineer with changes
Invoices Re-type the job into accounts Match payments to invoices
Accounts Month-end reconciliation Chase late payers
Questions & numbers Dig out policies & prices Pull the board-pack numbers Sign off the big calls

Signing off the big calls stays with a person — by design. That's the deal.

An illustrative model. The uplifts are banded from one real UK services business — details under NDA. Worked out live on your device; view source.

Why this is different

Everyone gets a chatbot.
Or the business gets a system.

Handing out licences doesn’t fail loudly — it decays. Built systems quietly pay for themselves.

The usual way

Give everyone a chatbot

key leaked
Monthly spend63 seats · ~£1,900
Outputslop

Seats multiply, prompts drift, two bots answer the same question differently — and nobody tested any of it. The spend climbs either way.

What we build

Build the system

Instant quotes Auto follow-up Board reporting
Monthly spend~£150 · flat
Outputmeasured

A small number of targeted systems — scoped, tested before they ship, keys locked down. The system does the work; the team gets the results.

The mess is optional. Start with the one system worth building →

An illustration, not a simulation — a composite, not any one company. But this is how it goes.

I build the AI a business runs on.
And I run the business it's built for.

One person, both halves of the job — the engineering, and the P&L it feeds. More about me →

Systems in production.

One real business, daily use. Figures banded for confidentiality.

Instant quote & bookingA customer gets a price and a slot in seconds, wired into the CRM. 2×+online conversion
A central pricing engine — the demo above is a working miniature of it — behind an AI-assisted quote flow and self-service booking. A large share of bookings now completes with no human touch. The full case →
One brain for the businessAsk it anything; it answers from the company's own records, showing its sources. 17×more likely to surface the right record first time
We rebuilt how it finds answers — and, more importantly, how it is tested. Every change is scored against real cases and must beat the last version before it ships. The testing code is public. The engineering →
The automation backboneCRM, job management, accounts and e-sign, joined into one flow. ~200 hrshanded back, every month
Roughly half of routine office work now runs itself, built so nothing fires twice and nothing gets dropped — and month-end reconciliation is gone. How it's wired →

Every system, in depth →

Security & data handling

Built so an auditor can say yes.

Certification-ready, not certified — stated plainly. Every system is built to the control principles behind SOC 2 and ISO 27001, so the posture exists before the paperwork does.

Least-privilege access

API keys scoped to the minimum they need. Capability tokens, not shared logins.

Secrets stay out of code

Credentials never sit in source, chat or tickets. Private source control, every change reversible.

Everything leaves a trail

System actions are logged and attributable. Audit trails are designed in, not bolted on.

Fails closed, not open

Idempotent webhooks, rate limiting that blocks by default — and nothing ships untested.

UK GDPR-aware by design

Data handling built with retention and subject-access requests in mind from the start.

Certified foundations

Major cloud and Stripe underneath hold SOC 2 and ISO 27001; the application layer on top is built to the same controls — ask for the list.

These controls run a live business every day — reviewed at board level, month in, month out.

Two ways in.

Hiring for a permanent Head of AI & Operations role? The CV → — it prints properly.

Start here · low risk

The 2-week diagnostic

We audit your systems and your P&L, and hand you the one system worth building first — written up and costed. Fixed fee, agreed before we start. You keep the roadmap either way.

If it doesn't surface a system worth building, you don't pay.

Book the diagnostic

How a fractional engagement works →

Charlie built and runs the AI and systems our business depends on: instant quoting, forecasting, and the board reporting they feed. He's the rare person who can ship the engineering and own the number it moves.

Get in touch.

One email. I reply personally within a working day.

or write to charlie@thoughtcoast.com