AI automation consultant
for UK SMEs
I help growing SMEs turn messy admin, inboxes and spreadsheets into reliable AI-assisted workflows.
Independent, UK-based, and hands-on. I design, build, test and monitor GenAI, agentic workflows and custom automations around the systems your business already uses, with delivery standards shaped by work across national critical infrastructure, government and local authority digital systems.
Built in the real world, not the demo room
Alongside DataTune I work in advisory at Amey, one of the UK's largest infrastructure services firms. The work is building digital and data systems for government departments internationally and UK local authorities. Multi-stakeholder programmes, regulated environments, very low tolerance for things going wrong.
That's the bar I hold AI work to here. If a system can't be evaluated, integrated and operated by your team after I'm gone, I won't build it.
Six AI capabilities, wired into the systems you already use
GenAI integrations
Agentic workflows
Document understanding
Retrieval (RAG)
AI in sales & support
Custom AI solutions
Where this actually pays back
Supplier invoice intake
~40 min per invoice → under 2 min. Extraction, validation against the PO, exceptions routed to a human, full audit trail.
Inbound lead triage & qualification
Hours-long first response → minutes. Inbound parsed, enriched, qualified and dropped into the CRM with a drafted reply for approval.
Internal Q&A over SOPs and policies
Grounded answer system over your own documents. Cuts ticket volume to the ops team and stops tribal-knowledge bottlenecks.
Quote drafting from RFQs
Inbound RFQ → drafted quote in minutes, priced against your rules and live supplier data. Sales rep reviews and sends.
Contract clause extraction
Back-catalogue of contracts indexed and queryable: renewal dates, liability caps, jurisdiction, change-of-control.
Same-day onboarding agent
New starter or vendor: accounts provisioned, access granted, docs issued, kickoff scheduled, compliance sorted. All from one trigger. You approve what matters.
The tools I build with
Models
Retrieval
Orchestration
Evaluation & monitoring
Integration
Hosting & data residency
No vendor allegiance. The right model and the right orchestration layer depend on the workload, the data sensitivity and what your team can operate after launch.
Three phases, in this order
Diagnose
Two weeks. Map your workflows and data, rank candidate AI capabilities by value, feasibility and risk, and pick the one or two worth building.
Build
Four to eight weeks per capability. Design, integrate, evaluate against a representative test set, ship to a production pilot.
Operate
Monitor real-world quality, latency and cost. Tune, expand scope, and add the next capability once the first one is earning its keep.
What I won't build
- Public-facing chatbots with no grounding or guardrails
- Autonomous agents on irreversible actions (payments, sends, deletes) without human approval
- AI strategy decks with no build attached
- A replacement for the junior hire you actually need
- Anything that can't be evaluated against a test set
AI is genuinely useful in a lot of places. It's also a confident-sounding way to ship risk. The job is telling the difference.
Three ways to work together
AI diagnostic
A clear-eyed map of the highest-value AI opportunities in your operation, with a recommended build and an honest view of what's not worth doing.
Pilot build
One production AI capability, integrated with the systems you already use, evaluated and monitored. The thing actually goes live.
Ongoing build partner
Rolling work across multiple AI capabilities once the first pilot is in production. Small number of clients at a time.
Put AI to work inside your business
A short call is the fastest way to find out whether AI is the right next step, and which problem to start with.