Your business is unique.
Your AI should be too.
I design and build AI workflows around how your company actually operates — grounded in your own documents, measured for accuracy, with a human in the loop from day one. Built end to end by an engineer, not a slide deck.
95% of AI projects never pay off.
MIT research found that the vast majority of enterprise AI pilots deliver zero return.* Not because AI doesn't work — because of how it gets bought and bolted on.
The fix isn't more AI. It's AI built around how you actually work, measured for accuracy, and owned by you.
* The GenAI Divide: State of AI in Business 2025, MIT.
- Built for everyone, fits no oneOff-the-shelf tools can't handle the detail of how your business actually works — so people stop using them.
- No way to know if it's rightMost AI is deployed with zero measurement. The first hallucinated answer to a customer destroys trust in all of it.
- A black box you don't ownVendor tools lock your workflows into someone else's roadmap. When your needs change, you're stuck waiting.
What this actually means for your business.
Hours back, every week
The repetitive text work — support replies, paperwork, hunting for answers — gets drafted by AI in seconds and checked by your team, instead of written from scratch.
Faster answers for your customers
Enquiries answered in minutes instead of days — without hiring more people or working longer evenings.
Your knowledge, always available
Everything your business knows — policies, prices, procedures, past jobs — answered instantly, instead of living in one person's head or a buried folder.
Fewer costly mistakes
The AI only answers from your own documents, shows where every answer came from, and hands anything uncertain to a human. You stay in control.
Customer support co-pilot
AI drafts replies from your own help docs — cited, checked, escalated to a human when unsure.
An online shop answering “where's my order?” emails in seconds, not hours.
Ask your documents
Manuals, policies, contracts and regulations — answered instantly, with the source shown.
A letting agency getting instant answers on tenancy rules, with the clause cited.
Document processing
Extract, classify and summarise the paperwork that eats your team's week.
An accountancy firm turning a client's shoebox of PDFs into a tidy summary.
Internal knowledge assistant
Everything your business knows, at your team's fingertips instead of buried in folders.
New staff asking “how do we handle refunds?” and getting the right answer.
Report & content generation
Quality drafts grounded in your business data — not generic AI filler.
A consultancy turning site-visit notes into a formatted client report by morning.
Workflow automation agents
Hand repetitive multi-step tasks to AI agents, with guardrails that keep them honest.
Overdue invoices chased, bookings confirmed and records updated overnight.
Not sure which fits? That's what the audit is for.
AI Readiness Audit
I review your workflows and deliver a report of the three highest-ROI AI automations for your business — what to build, what it costs to run, and what it saves. A fixed price, a fixed timeline, and a plan you own either way. It starts with the free intro call — the audit is the paid deep-dive that follows if there's something worth investigating.
AI Feature & Automation Build
Support co-pilots, internal assistants, document processing. Grounded answers with citations, automated evals in CI, human escalation built in — from a scoped pilot to a production system.
Web Development
Full-stack product work with the same craft that ships the AI — Next.js, TypeScript, Python. Often where an AI engagement starts or ends.
Capstan — an AI support co-pilot you can trust.
A live multi-tenant SaaS that drafts customer-support replies grounded in each company's own help docs — every answer cited back to its exact source.
Visit the live product
- Grounded & citedEvery reply traces to a source document — no guessing, no hallucinated policy.
- Accuracy is measuredAn automated eval suite runs in CI, so answer quality is tested on every change.
- Human in the loopConfidence guardrails escalate weak matches to a person instead of bluffing.
- A real product, not a demoMulti-tenant, authenticated, billed — the full commercial build, shipped solo.
Marcus took a workflow we'd been drowning in and just handled it. Every decision explained in plain English, every answer traceable to a source, delivered on the timeline he quoted.
We expected a chatbot. What we got was a proper system — with checks, sources and a human fallback. It's the first AI tool our team actually uses.
Agencies sell teams. You need one person who can do all of it.
The person you talk to builds it
No account managers, no handoffs, no junior team learning on your budget. One senior engineer, end to end.
Senior work without agency overhead
You pay for engineering, not office space and layers of management. Decisions take a conversation, not a meeting cycle.
Honest scope
If AI is the wrong answer for your problem, I'll say so in week one — not after six months of billing.
You own everything
Code, prompts, data, infrastructure — all in your accounts, all documented, no lock-in of any kind.
Discovery
A free call. We find where AI genuinely pays off in your business — and where it doesn't.
Pilot
A small, fixed-scope build that proves the value on your real data before you commit to more.
Build
The production system — evals, guardrails and escalation included, not bolted on.
Handover
Documentation, training, and no lock-in. You own everything I ship.
The things everyone asks before booking a call.
What does it cost?+
The AI Readiness Audit is £495, fixed. Builds are scoped per project after the audit or a discovery call, so you always know the full cost before anything starts — no open-ended billing. Web development is charged at a day rate.
How long does a project take?+
The audit takes one to two weeks. A pilot build typically ships in two to four weeks, and a full production system in one to three months depending on scope. You see working software from the first weeks, not at the end.
Do I own what you build?+
Yes — everything. Code, prompts, evaluation suites, data and infrastructure all live in your accounts and belong to you. When we're done, you can maintain it yourself, hand it to another developer, or keep working with me. No lock-in of any kind.
Is my data safe?+
Your data stays in your own accounts and environment. I use commercial AI APIs with training on your data disabled, and the systems I build are designed so sensitive data never leaves your control. Happy to sign an NDA before we talk specifics.
What if AI isn't the right answer for my business?+
Then I'll tell you — that's the point of starting with an audit. Some workflows aren't worth automating and some problems have cheaper non-AI fixes. The audit report is yours either way, and you'll know exactly why, not just whether.
Who do you work with?+
Small and mid-sized businesses, mostly in the UK, fully remote. The common thread is a workflow drowning in repetitive text work — support inboxes, document processing, internal questions — rather than any particular industry.
What technology do you use?+
Leading commercial AI models (primarily Anthropic's Claude), open-source models where they fit, and a modern web stack — TypeScript, Next.js, Python, PostgreSQL. Everything is model-agnostic by design, so you can swap providers as the market moves.
How do we start?+
Book a free 30-minute intro call. We'll talk through where your time goes and whether AI can win some of it back. No pitch, no pressure — if there's nothing worth building, I'll say so on the call.
I'm Marcus — a full-stack engineer with five years shipping production web applications, now designing and delivering AI systems end to end.
Most AI consulting stops at the slide deck. I take the other route: find the opportunity, scope it honestly, build it properly, and prove it works with measurable evals. Based in Bristol, UK — working remotely.