Team

The people on the call are the people on the work

Five people, each of whom leads a discipline rather than supervising someone junior doing it. There is no delivery tier behind us, which is the point: you get the person who made the decision, not a summary of it.

  • Tom

    Founder · AI infrastructure & security lead

    Owns the platform build: hardening, the air-gap design, and the assurance evidence.

    Tom’s background is software engineering and quality engineering rather than management consulting: test automation, device management and security tooling before AI, which turns out to be unusually good preparation for it. The disciplines that make an AI system defensible in a regulated setting are the ones quality engineering has always cared about: measure it, define what correct means before you build, assume it will fail and design for that, and never ship something you cannot test.

    On an engagement he owns the infrastructure and the security position: the DISA STIG and CIS Level 2 baselines, the air-gapped artefact pipeline, the NCSC CAF mapping, and the deviation list that honestly accompanies both. He founded Flowing Mind and still does the work.

    • Air-gapped platforms
    • DISA STIG & CIS L2
    • NCSC CAF
    • Quality engineering
  • Nigel

    Lead architect

    Decides where the boundary sits, and what the system looks like either side of it.

    Nigel owns the system design, which in this work means the decisions that are expensive to reverse: where the data boundary is drawn, what runs inside it, how models are served and sized against real concurrency rather than a benchmark, and how applications reach them.

    That last point is the architectural commitment the whole consultancy rests on. Applications talk to an interface the client owns rather than to a model directly, so the model underneath can be replaced, or different data classifications routed to different backends, without rewriting everything above it. Architecture you can leave is the deliverable.

    • Platform architecture
    • Data boundary design
    • Model serving
    • Capacity sizing
  • Dave

    Delivery lead

    Runs the engagement: scope, sequence, and the assurance timetable.

    Dave leads complex technology delivery in high-stakes enterprise and government environments, with a programme and project management background across large multi-supplier implementations, from discovery through to post-launch support.

    AI programmes rarely fail for want of a good model. They fail on sequencing, on a stakeholder who was not in the room when a decision was made, and on an assurance gate nobody put in the plan. Dave owns that half: what gets built in what order, who signs what and when, and keeping the evidence work moving alongside the build rather than trailing behind it.

    • Programme delivery
    • Government & enterprise
    • Assurance scheduling
    • Multi-supplier
  • Andy

    AI development lead

    Leads the build: retrieval, agents, evaluation and the serving stack.

    Andy leads the engineering. Retrieval over client material with citation that holds up, agents where the task genuinely varies rather than where an agent sounds impressive, the inference stack that serves them, and the tests and CI that make the result something a client team can safely change after we have gone.

    The evaluation harness is his as much as anyone’s, and it is the part that decides whether a build is defensible. A system with a measured baseline on the client’s own material can be improved, argued for, and re-checked after a model swap. One without it is a demo with a deployment pipeline attached.

    • Retrieval & RAG
    • Agents
    • Evaluation harnesses
    • Inference stacks
  • Tammy

    Lead business analyst · AI assistive technology for SEN

    Turns the stated requirement into the one that was meant, and specialises in assistive AI for special educational needs.

    Tammy runs the analysis: what the process actually is rather than what the process document claims, where the time genuinely goes, and which of the things people ask for would survive contact with the work. A meaningful share of proposed AI projects turn out to be a reporting problem or a form design problem, and discovering that during analysis costs a fraction of discovering it during a build.

    She also specialises in AI as assistive technology for special educational needs. It is one of the few applications where these tools change what a person can do rather than only how quickly they can do it, and one where the distinction between supporting a learner and making decisions about them matters enormously. In that setting accessibility and equality impact are the assessment, not an annex to it.

    • Requirements analysis
    • Process discovery
    • Assistive technology
    • SEN
    • Accessibility

How we staff

Small and senior is the product, not a stage we are passing through

Every person on this page leads their own discipline and does the work themselves. We do not carry a bench, we do not mobilise a team of forty, and we will tell you plainly when a piece of work needs more capacity than we have rather than stretching to take it.

Where an engagement needs capability we genuinely do not hold, specialist legal advice, clinical safety, sector-specific regulatory expertise, we say so and bring in somebody who does, rather than improvising at your expense.

On clearance: tell us the level and any facility requirement in the first conversation. We will say straight away whether we meet it as we stand and where we would need to work alongside your own cleared staff or a cleared partner.

Talk to whoever you actually need

If your question is architectural, ask Nigel. If it is about hardening or an accreditor, ask Tom. If it is whether this can be delivered by March, ask Dave. Book a call and we will put the right person on it.