Process Audit
Technical Due Diligence
Process Build
Agent Operations
Fractional CTO
Board Advisory
Most of the companies I talk to have tried AI. Very few have it running a process. The pilots answer questions; the work still arrives in the same inbox, gets read by the same people, and gets typed into the same systems.
That is the gap I close. I build always-on AI agents that run a process end to end: they watch for work arriving, read it, match it to your records, act within rules you have written down, chase what is missing, and hand every decision that matters to a named person with the full picture. Your team keeps the judgement. The agents take the reading, matching and chasing. You absorb more volume, or grow, without hiring ahead of it.
It works in any industry. What matters is the shape of the work: steady volume, documents or email coming in, rules that cover most cases. Insurance claims intake, credit control, supplier invoice exceptions, order entry from emailed purchase orders, contract renewals, project status reporting. See what agents can run, with the evidence for each. I work with mid-market companies of 50 to 500 people: big enough for the volume to matter, without an in-house AI team or the appetite for a two-year transformation programme.
How every agent is built
- Watch. The agent monitors the inbox, portal, queue or system, around the clock.
- Read. It extracts the facts from emails, PDFs, forms and attachments.
- Match. It ties them to the right customer, policy, contract, order or project in your records. This is where most AI projects quietly fail, and why the data comes before any model.
- Decide. It applies your written rules to the routine cases, and scores its own confidence.
- Act. It updates your systems, replies, chases what is missing, and routes the work.
- Escalate. Exceptions and judgement calls go to a named person, with everything the agent found.
- Prove. Every step is logged, and every month you see the work handled, the hours returned, the exceptions and the cost per item.
Here's how the engagements fit together, and where to start:
- Want agents running one of your processes? Start with a Process Audit.
- Already know the process and have the baseline? Go straight to a Process Build.
- Have agents running and want them kept sharp? Agent Operations.
- Need senior technology leadership week to week? Fractional CTO.
- Raising, buying or reporting to a board? Technical Due Diligence.
- Have the team, need an independent voice at board level? Board Advisory.
Process Audit
Most AI projects that disappoint were decided before the first line of code: they started on the wrong process, or on records nobody had looked at. The Process Audit looks first. In two weeks I take one to three of your processes and find out exactly what an agent could run, what it would return, and what has to be fixed before it starts.
What's included
- Baseline. How the process runs today, measured: volume, minutes per item, cost, delays, error and rework rates. This is the number the agent will have to beat.
- Work map. Where the work arrives (inboxes, portals, forms, phone notes), who touches it, and where people re-key data between systems by hand.
- Records check. Whether the records the agent would rely on agree with each other and are current: customers, policies, contracts, orders. The inboxes nobody lists included.
- Decision map. Which decisions are routine and rule-based, which need a person, and what a wrong answer would cost at each step, so automation starts where it is safe and valuable.
- The build spec. The recommended first process, its scope, the systems it touches, the controls it needs, the metric that will prove it worked, and a fixed price for the build.
Who it's for
- Mid-market companies that want a process off their team's plate and need to know which one pays first
- Teams whose AI pilot works in a demo but has never handled live volume
- Finance, operations and claims leaders with a backlog that keeps growing faster than the team
- PE operating partners looking for the first agent-ready process across a portfolio
What you'll have at the end
A measured baseline, a ranked shortlist of processes with the return on each, and a build spec with a fixed price. It is yours either way: use it with me, with your own team, or with another supplier. If the honest answer is that an off-the-shelf product will do the job, the audit says so. Half the fee is credited against a Process Build started within 90 days.
Why it works
It is small, fixed and independent. I have no software licence or agency bench waiting for the build, so the recommendation follows the evidence. And the method comes from production work, including agentic AI that resolved 67% of customer service cases without a human at a European insurance brokerage, where the data was sorted out before any model was built.
Get in touch about a Process Audit →
Technical Due Diligence
The first step when the question is your technology as a whole. A deep, independent review of your technology situation, delivered as an investor-ready assessment you can use for board meetings, fundraise preparation, or internal alignment.
What's included
- Stack and architecture review. Is your current architecture fit for purpose? What are the scaling risks? Where are the decision points coming up?
- AI approach assessment. Are you building with the right AI stack for your use case? Where are the gaps between what you're building and what's production-ready?
- Team and process diagnostic. Is the engineering team structured correctly for this stage? Where are the capability gaps?
- Risk identification. Technical debt, vendor lock-in risks, security considerations, and EU AI Act compliance posture
- Investor-ready summary document. Clear, honest, jargon-free. The kind of document that gives investors confidence
For investors and operating partners
If you're a PE firm, VC, or operating partner evaluating AI readiness across a portfolio company, this is the same engagement, tuned for your lens. You'll get: portfolio company AI maturity assessment, technical debt quantification, transformation potential evaluation, and a clear-eyed view of what's real versus what's demo.
Who it's for
- Founders who have just closed a round and want to validate technical direction before scaling
- Founders preparing for a Series A or Series B who need to answer hard technology questions
- PE firms and operating partners evaluating technology and AI risk in a portfolio company (for a specific process you want on agents, see the Process Audit)
- Boards or investors who want an independent technical assessment
What you'll have at the end
A board-ready document that answers every investor question about your technology, and a clear view of what to prioritise next. No ambiguity, no filler.
Why it works
Many engagements start here or with a Process Audit. A Technical Due Diligence gives you a clear picture of where you are, what to prioritise, and whether a deeper engagement makes sense. There's no better way to de-risk the next 12 months of engineering investment.
"Both entry points are deliberately low-commitment. One of us should know whether this is a fit before we go further."
Get in touch about Technical Due Diligence →
Process Build
This is the core offering: always-on agents running one of your processes in production, on the systems you already have. Not a pilot, not a demo. A process your business runs on, with the controls a regulator or an auditor would expect.
There are two sizes, and the Process Audit tells you which one you need:
- First Process, £25,000–40,000, 6–8 weeks. One process, one or two intake channels, the core decision rules. The fastest way to prove agents in your business.
- Full Process, £50,000–90,000, 10–14 weeks. Several intake channels, deeper integration with your systems, more decision branches and several linked agents.
What's included
- The records first. The data the agent will rely on, cleaned up and connected so it matches work to the right customer, policy, contract or order. This is the step most AI projects skip, and the reason most of them stall.
- The agents. Built on your inboxes, portals and core systems, not a platform migration. Decision rules written down with your team, confidence scoring on every routine call.
- The controls. An audit trail for every item, plain-English reasons for every routing decision, human sign-off on anything consequential, checks for vulnerable customers where they apply, UK or EU data residency, no training on your data.
- The rollout. Shadow mode first, where the agent works alongside your team without acting. Then 10% of live volume. Then scale, only as the numbers justify it.
- The dashboard. Work handled, hours returned, exceptions, accuracy and cost per item, from the first week of live running.
What the agent never does
It never makes the decisions that carry real consequences: approving or declining a claim, settling a payment, writing off a debt, changing contract terms. A named person does, with everything the agent found in front of them. That line is what lets a cautious board say yes.
Examples
- Insurance claims intake: read the notification and attachments, match the policy, check vulnerability, request and chase missing documents, set up the claim, route it to the right handler, acknowledge the claimant.
- Credit control: watch the ledger, chase overdue invoices in your tone of voice, log promises to pay, spot disputes and route them to a person, report the cash position daily.
- Order entry: read purchase orders that arrive as PDFs and emails, match customer, item and price, enter them in your ERP, query the mismatches.
- Contract renewals: extract obligations and dates, warn before renewals and deadlines, flag non-standard clauses for review.
- Project status: turn meetings and email into actions, chase owners, draft the weekly report and the client update, flag slippage early.
Who it's for
- Mid-market companies (50–500 employees) with a process that grows faster than the team that runs it
- Organisations with AI pilots that work in a demo but have never handled live volume
- Insurers, MGAs, TPAs and brokers who want claims intake faster without adding handlers
- Finance teams that want cash in sooner without another hire in credit control
- PE portfolio companies where the operating plan depends on growing without matching headcount
Kill criteria
If the agent cannot reach the accuracy and the return agreed in the build spec, we stop early and you keep everything built so far. The goal is a process that runs, not sunk-cost persistence on one that does not.
What you'll have at the end
One process running on agents in production, measured against the baseline, with the controls and the audit trail in place, and a team that knows how to work alongside it. Most clients move straight into Agent Operations, and many start a second process within six months. Read more about why most AI pilots never reach production and how I approach production deployment.
Get in touch about a Process Build →
Agent Operations
An agent that runs a process is a system, not a project. Volumes change, suppliers change their invoice layouts, models improve, and your rules evolve. Agent Operations keeps it sharp, and makes sure you can see exactly what it is doing for you.
What's included
- Monitoring and alerting. Accuracy, exception rates, response times and cost per item tracked continuously, with alerts before a drift becomes a problem.
- Exception review. A regular look at what the agent escalated and why, to turn recurring exceptions into new rules where it is safe to.
- Tuning and upgrades. Prompt and rule changes, model upgrades when better or cheaper ones arrive, and small changes to what the agent covers.
- The monthly value report. Work handled, hours returned, exceptions, accuracy and cost per item, against the baseline from the audit.
- A response-time commitment. Async by default, a monthly call, business-hours cover as standard, with out-of-hours cover available if the process needs it.
Model and hosting costs are passed through at cost, with a cap agreed up front. The price depends on volume and how much change you expect; it is agreed per process, and each additional process costs less than the first.
What you'll have
A process that keeps getting better instead of slowly drifting, and a single page every month that tells your board what the agents are worth.
Get in touch about Agent Operations →
Fractional CTO
Right now, nobody has a map. The waters every business is crossing with AI are uncertain, shifting, and choppy, and visibility is poor. Everyone is sailing blind, including the people who sound the most certain.
You cannot control the wind and the waves. What decides whether you come through is who is at the helm. An experienced hand does not have a map either. What they have is the intuition to read the water and the judgement to make the right call the moment new information surfaces. That does not improve your odds a little. It transforms them.
That is what a Fractional CTO does. Most technology advisors give you strategy. What you actually need is someone who has been in the founder seat, built the team, made the architecture calls under pressure, and shipped AI that works in production. More about my background.
CTO-level capability without the full-time cost, equity, or six-month hiring process. It is also not the same as hiring an AI consultancy: here is the difference, and why the order you hire in matters.
The range exists because scope varies: a 5-person team building a first AI product needs different support than a 50-person organisation deploying AI across operations. The right number is agreed upfront, with scope defined clearly.
Maximum two concurrent Fractional CTO clients. Ask about current availability.
One rule I hold myself to: when I am your Fractional CTO, I do not also sell you a build unless your board has signed it off, with the alternatives, including off-the-shelf products and other suppliers, set out alongside it. Why that matters.
What this looks like week-to-week
- Architecture decisions. The calls your engineers escalate because they need a senior voice. You get a clear recommendation, not a committee process.
- AI roadmap. What to build, in what order, with what approach. Built around your business model and runway, not the latest benchmark.
- Hiring and team. Job specs, interview process, offer framing, team structure decisions. The engineering hiring mistakes at this stage are expensive.
- Investor and board readiness. I attend board calls, investor updates, and due diligence sessions as your technical leadership voice. Your investors get a credible CTO presence without you hiring one.
- Vendor selection. Cloud, tooling, AI infrastructure. Honest evaluation, no vendor relationships to protect, no agency behind me, and a board sign-off rule on anything I could sell you.
- Team mentoring. Your senior engineers get access to someone who can push their thinking and accelerate their development.
Who it's for
- Post-Seed or Series A founders with 3–30 engineers and no technical co-founder
- Founders whose technical co-founder has left or is stretched
- Growth-stage companies (50–500 employees) building AI into their operations for the first time
- PE-backed businesses that need senior technical leadership during transformation
What you'll have at 90 days
Your first AI system in production with measured KPIs, an architecture that scales, a team that knows where it's going, and investor updates that land. Not a strategy document. A working system. See what this looks like: 67% autonomous case resolution at an insurance brokerage and £28M in recurring annual savings identified at NHS Wales.
The commitment model
A minimum 3-month engagement ensures there's enough time to understand the business and deliver meaningful direction, not just observations. After the initial term, engagements run on a rolling monthly basis.
Maximum 2 concurrent clients at this tier. This is not a scalable model by design.
Get in touch about a Fractional CTO engagement →
Board Advisory
A lighter-touch engagement for founders and boards who have an engineering team and technical direction, but need a senior external voice for credibility at board level, or a trusted advisor for the big technology decisions.
What's included
- Quarterly technology strategy review (half-day session)
- Investor call preparation: framing technical progress in commercial language
- Technical risk assessment for board packs
- On-call availability for urgent decisions (1–2 hours/month)
- Ad hoc review of major technology decisions: architecture changes, AI stack, major vendor contracts
Who it's for
- Founders who have a CTO or senior engineer but want an independent external perspective
- Boards that want a credible technical voice on quarterly calls
- Founders at a later stage (Series B+) who need a strategic sounding board, not hands-on direction
- PE operating partners who want ongoing technical oversight of portfolio companies
What you'll have
Confidence at the board table, with a credible technical voice and quarterly progress backed by metrics, not slides.
Application process
This tier is by referral or application. If you found your way here through a recommendation, reach out directly. If you're applying cold, include a brief note on your company, stage, and what you're looking for.