Skip to content
Jake Bowles

The AI decisions on your board agenda are not waiting for you to hire.

I am a practising CTO. Nine years on Azure for UK organisations, from Defence to regulated enterprise. I spend two days a week with a small number of UK leadership teams: putting agents into production, getting engineers shipping with them, and building estates that regulators accept.

This site, and the agent running on it, were built the same way I help teams build.

Book the first conversationSee the agent at work

My fractional work is agreed with Synapx. Synapx clients and live prospects are carved out, and any potential conflict is declared and stepped back from before it becomes one.

  1. Mon

    Synapx · Chief Technology Officer

  2. Tue

    Synapx · Chief Technology Officer

  3. Wed

    Synapx · Chief Technology Officer

  4. Thu

    • 09:00Board session: AI as an operating model
    • 11:00Agent architecture and governance review
    • 14:00Engineering: agents in the delivery workflow
    • 16:30Sovereignty assessment: what the vendor really guarantees

  5. Fri

    • 09:00Leadership: technology direction and the quarter ahead
    • 11:00Team clinic: plan-first, review-gated agentic development
    • 14:30Sovereign and hybrid estate: architecture options
  6. An illustrative week. The split is yours to set.

What the highlighted days are for

Three kinds of work I do in the room rather than advise on from outside it. Each has a longer explanation in the console; the engagement itself is described on the fractional CTO page.

  1. Agentic AI in production

    Agents that take delegated, governed action, built by engineers who ship with agents themselves.

    Ask the console →
    What it is.
    Systems where a model plans, calls tools, observes results and iterates towards an outcome, inside boundaries you set. The unit of value is a completed task, not a generated paragraph. The teams that build them work the same way: coding agents plan, implement, test and review under human direction, and this site is a worked example.
    Why it matters.
    Tool use, long context and cheap inference crossed the threshold together. The organisations that win treat agents as an operating model, delegated authority with audit, rather than a feature. Inside engineering, specification, review and verification become the scarce skills, and most teams under-invest in guardrails and over-invest in prompts.
    How I help.
    Board-level framing and the operating model. Agent architecture and the governance scaffolding, permissions, evaluations and observability, that makes production agents defensible. Team operating models, tooling choices, plan-first and review-gated workflows, and honest measurement of where agents pay off.

  2. Sovereign cloud and AI

    Keeping UK data, models and operations under UK control without giving up capability.

    Ask the console →
    What it is.
    "Sovereign" is a set of concrete constraints: where data rests and moves, who can compel access, which models you can run and inspect, and who operates the stack. Slogans do not satisfy any of them.
    Why it matters.
    Regulated sectors and the public sector need AI they can defend to regulators and citizens. The practical options are widening fast, and the trade-offs are rarely explained plainly.
    How I help.
    Sovereignty maturity assessments, architecture for sovereign and hybrid estates, and vendor scrutiny that separates real guarantees from marketing.

  3. AI due diligence and assurance

    For investors and boards who need to know whether an AI claim is real before money moves.

    Ask the console →
    What it is.
    An independent read of a target or a portfolio company: the system as built, the numbers behind the demo, the contracts behind the numbers, and what a buyer would need to know before completion.
    Why it matters.
    Four questions decide most AI-led deals and few data rooms answer them. Is the AI claim real? What does it cost per unit of output at ten times the volume? What happens when the model provider changes price or policy? And what regulatory exposure transfers with the deal?
    How I help.
    Technical and AI due diligence for private equity and boards: architecture and evaluation review, unit economics at scale, provider dependency and exit options, and the regulatory exposure that transfers, written for the investment committee rather than the engineering team.

The record

I am Chief Technology Officer at Synapx. Before that I built and led the Data and AI practice at Expert Thinking, was a BCS IT Awards Rising Star finalist in 2023 and 2024, and spent nine years delivering on Azure for UK organisations, from Defence to regulated enterprise.

The full record, with every employer and credential, is on the CV.

Work I can point to

  1. Go-Ahead London

    An idea that would overhaul how the business worked.

    • From idea to production: I delivered the solution that overhauled their business.

  2. Royal London Asset Management

    A build that had to reach production.

    • Go-live support: the trusted guiding hand from a finished build to a system running in production.

  3. Expert Thinking

    No Data and AI practice, just as ChatGPT arrived.

    • I grew the practice from inception into a core business function, through the years in which the agentic era began.

  4. Synapx

    A consultancy delivering in low code.

    • I took engineering from low code to pro code through AI tooling, with agents in the delivery workflow.

Where the return comes from

A fractional CTO is not a cheaper CTO. It is a different shape of one: senior judgement applied at the moments that move the business, without carrying a full package for the hours in between. The return shows up in four places, and we agree in the first month how each will be measured.

  1. Decisions stop circling

    AI, cloud and vendor questions that have sat on the board agenda for a quarter get an accountable answer in the week they are raised, from someone who has made them before and will stand behind them.

    In practice: decisions that had taken weeks take days.

  2. One salary line becomes a share

    You pay for one or two days of a CTO, not a full-time package, a recruiter’s fee and the months a search takes. The role is filled the week we agree it.

    Set two days a week against a full-time package, a recruiter’s fee of around a quarter of first-year salary, and a search that typically runs three to six months.

  3. The organisation moves faster than the market around it

    Agents in production, engineers building with agents, and a sovereign estate that regulators accept are the three shifts every UK organisation now has to make. You make them with someone who has already done each one, so you skip the expensive first attempt.

  4. Capability stays when the engagement ends

    The operating model, the guardrails and the engineering practices are built into your teams, not rented from a bench. What you keep is a leadership team that can govern AI and an engineering function that ships with it.

How the return is measured →

Notes for decision-makers

Essays I write for people who have to decide about agentic and sovereign AI, not build it.

All essays →

See an agent do the work

The Agent Console is a small agentic system over this site. Ask it something and it plans, calls tools and renders the answer. I built it with coding agents, the same way I help teams build.

jake@grid ~ agent v5.12/6 online

> show me his career

Plan · 2 steps

1. fetch cv.timeline

2. render <Timeline/>

▸ fetch cv.timelineok 183ms

Sep 2025 to now

Chief Technology Officer · Synapx

Nov 2022 to Aug 2025

Expert Thinking

Run this for real →

Try

> whoami

> career

> sovereign ai

> read sovereign

> hire

Bring all six nodes online and it grants root. How it was built.

Three ways to engage

Three shapes the work takes. Every engagement starts with a first conversation about the problem, and I will tell you in that conversation what it costs.

  1. Fixed-scope diagnostic

    A short, fixed scope, for a fixed fee.

    • Where AI, cloud and delivery decisions are stuck, and why.
    • A written view the board can act on, with the first ninety days sequenced.
    • The guardrails, evaluations and observability a production agent would need in your estate.

  2. Advisory retainer

    One or two days a week, agreed term, the seat held.

    • Technology direction owned, board and leadership meetings attended.
    • Engineering led in the room: architecture, review gates, agents in the delivery workflow.
    • A one-page monthly report against the return we agreed in the first month.

  3. AI due diligence

    Fixed fee per target, scoped to the data room and management access.

    • Is the AI claim real, and what does it cost per unit of output at ten times the volume.
    • Provider dependency: what happens when the model vendor changes price or policy.
    • The regulatory exposure that transfers with the deal, written for the investment committee.
Jake Bowles, fractional CTO

Let's talk about your week.

Open to fractional CTO engagements, advisory and speaking.

Book the first conversationLinkedIn

Or write to j@kebowl.es.