AI systems integration
I build AI systems that actually run your business.
Right now they are drafting £700k to £1m of monthly buying, reconciling stock for the UK's leading leisure centres and finding the next customers worth calling. Built on top of the software you already run: audited, human-approved, working in weeks. Not strategy decks, not a chatbot, not an agency.
live systems in manufacturing, distribution, retail and compliance · based in the UK, working anywhere
Simulated example of how a system I build behaves. You ask in plain English, for example: “Who owes us what? Draft the chase-ups.” The work arrives drafted: seven overdue invoices worth £23,400, chase-up emails ready, oldest at 62 days. The system checks it, then waits: nothing is sent until a person approves it, and every action is logged with who, what, and before and after state.
Synthetic data. The workflow shape is real.
01the problem
Sound familiar?
Four things owners tell me in the first ten minutes.
Month-end takes a week.
Your best person spends the day retyping numbers between systems.
Invoices get chased when someone finally has an hour.
You tried ChatGPT. It wrote a poem.
None of this needs new software. It needs a working layer on top of what you have: AI carries the admin, your people keep the relationships.
02how it works
From one conversation to a working system
A conversation
Watch the real work
Pick one workflow
Connect the tools
Draft, not do
Your team learns it
Keep it, grow it
01 It starts with a chat about the work that eats your week. No audit, no deck, no jargon.
02 I sit with the people who actually do the job and watch how it really happens.
03 We pick one workflow, not a grand plan. Something that gives hours back within days.
04 I connect AI to the software you already use. It can read your data; it cannot change it.
05 The AI drafts the work: the order, the email, the numbers. Nothing happens until a person says yes.
06 Your team builds it with me, so the skills stay in your business and your team runs it day to day.
07 Everything is logged and checkable. Keep it, extend it, or stop whenever you like: no lock-in.
03proof, not promises
I do not just advise. I build.
Every system I ship is audited, human-approved and running in production, at a scale where a wrong answer costs real money. That standard is not a promise: it is live right now, in manufacturing, distribution, retail and compliance.
- monthly
- a whole sale-or-return month-end run end to end: counts in, stock reconciled, invoices drafted
- £700k to £1m
- of monthly buying drafted from live sales trends, every line approved by a person
- Monday
- a briefing of leads found and qualified, first messages drafted; the team spends the week selling
- 95%+
- certificate reading accuracy in production, anything uncertain routed to a person
Compliance taught me the bar: if a certificate is read wrong, a crew member does not board. That is the standard every one of these systems is built to.
Each number belongs to a named case study, told honestly with a status line.
Read the builds04what I do
Connect, build, hand over
Advising or building, the outcome is the same: a system your team runs day to day.
01
I connect AI to your systems
Whether you run a 20-year-old back-office system or a modern cloud stack, I wire AI in so it can read what your business already knows: stock, invoices, customers, jobs. Reading only; your systems stay untouched.
the systems you already run · read-only · audited
02
I build the custom pieces
When the job is bigger than a workflow, I design and build complete production AI systems: web applications, dashboards, document checking, verification, compliance. Full software, always in service of an AI system that earns its keep.
production-grade · audited · human review built in
03
Your team runs it
Every build ends the same way: your team shown how to run it, on their own real work, as part of delivery. Not a training course, not an extra line on the invoice. The system stays useful because your people own it day to day.
included in every build · your real workflows
the operating rules
The rules every system I ship obeys
01
AI never invents a number.
Plain maths sizes the decision. AI explains it in plain English.
02
Nothing is sent unchecked.
AI drafts, the rules check every line, and a person approves anything unusual before it moves.
03
Every action is audited.
Who did it, what changed, the before and the after. The log is always on.
04
Your data stays in your systems.
New layers are additive. I do not rip out what works.
Kadio · Conrad Empson
non-negotiable
Screenshot this and send it to the person who owns the spreadsheet.
06recent work
Built, shipped, running
running monthly
Blu Concepts
The system checks in with every site each month, collects the counts, reconciles the lot and drafts every invoice. A person approves anything odd.
every sitechecked in with, counted, reconciled and invoiced, monthly
Read the build
running weekly
Taylor Made Designs
Every Monday a briefing lands: who to contact, why now, first message drafted. The team sells; the system searches.
12stages between an instruction and a verified name worth calling
Read the build
in production
Nevis Marketing
Reads live stock, sales and suppliers, then drafts the weekly buying with every line explained. A person approves.
£700k to £1mmonthly purchasing supported
Read the build
agents in production
CrewPass
Thousands of certificates a month, read, checked and verified by a team of agents. Anything uncertain goes to a person.
95%+certificate reading accuracy in production
Read the build
07the timeline
Working in days, trusted in weeks
I build with AI as well as building AI, and it has changed the maths: the first working version lands in days, not quarters. The weeks after are what earn trust.
Day 1
The map: where AI helps, where it should not go.
Week 1
One broken workflow, working by Friday.
Weeks 2 to 6
Harden it with your team: edge cases, accuracy checks, audit trail.
Ongoing
Support if you want it, never a lock-in. Stop on 30 days' notice.
The speed is not a promise either: two of the systems above existed as working software within days of starting. Development used to mean a three-to-six-month roadmap; because I build with AI, a working system lands in the first week and the weeks after go into edge cases, accuracy checks and the audit trail: the parts that make it production-grade.
08fit
Worth a call, sorted or not
who it is for
- You run a business, or the operations inside one, and can get a decision made.
- There is real operational pain: quoting, purchasing, reconciliation, month-end.
- You want a straight answer on where AI fits before you spend real money.
- You want your team to end up running it themselves.
come anyway even if
- You think you want a chatbot. Working out what you actually need is the job.
- You are not sure AI even applies. That is what the call is for.
- You have been sold the hype before and it burned you.
- You only half-know what is broken. I will help you name it.
Based in the UK, working anywhere. Remote-first, on site when it matters.
09the person
Direct with the builder
Conrad. I build production AI systems for businesses in manufacturing, distribution, retail and compliance. I founded CrewPass, a compliance platform for the superyacht industry, and built the AI engine inside it, so I look at your business the way an owner does, not just an engineer. Based in the UK, working anywhere.
Book a 20-minute call
Bring one workflow that eats your week, or just the problem. Twenty minutes, direct with me, and I will tell you straight whether it is worth fixing.
Opens my calendar in a new tab. Pick any slot that suits you.
Prefer email?
A few lines about what eats your week is plenty. I reply within 24 hours, and if it is worth fixing, the next step is the call above.
Start by email