Skip to main content

Direction

Technology strategy

Build, buy or stop. Costed, with the trade-offs written down and the reversible decisions separated from the ones you only get to make once.

Technology strategy work is mostly the practice of making fewer decisions, later, with better information. The expensive mistakes are rarely the wrong database. They are the irreversible commitments made early, on thin evidence, because a decision felt overdue.

We separate the choices that can be changed cheaply from the ones that cannot, get the second group right, and deliberately defer as much of the first group as possible. Then we cost it, including the cost of running it two years from now.

What you get

Named artefacts, not a slide pack. Each one is a thing your team can open, run, or hand to an auditor.

ArtefactWhat it contains
Current state assessmentWhat you run, what it costs, what depends on what, and where the single points of failure are, including the human ones.
Options with costsTwo or three genuine options, each costed for build and for three years of running, with the assumptions listed separately so they can be argued with.
Reversibility analysisWhich decisions are cheap to undo and which are not, so effort concentrates on the second group.
RoadmapSequenced by dependency and risk rather than by quarter, with the decision points marked.
Vendor assessmentWhere buying is the answer: a comparison against your requirements, including exit cost and data portability.
AI feasibility noteWhere AI is on the table: what it would need to be true, what it would cost, and whether the data supports it. Frequently the answer is not yet.

How it runs

  1. 01 · 1–2 weeks

    Frame

    The decision that actually needs making, which is often not the one we were asked about.

  2. 02 · 2–4 weeks

    Prove

    Where a key assumption is load-bearing, test it. A small spike beats a long argument about whether something will work.

  3. 03 · 3–6 weeks

    Build

    The costed options, the roadmap and the write-up, produced with your team rather than presented to it.

  4. 04 · 1–2 weeks

    Hand over

    The document belongs to you, in your systems, in a format you can edit. We walk the board or leadership team through it if that helps.

What we use

Chosen per engagement against your constraints. We have no reseller relationships and no incentive to recommend one of these over another.

Assessment

Architecture reviewCost modellingDependency mappingThreat modelling

Cloud and hosting

AWSSelf-hosted LinuxManaged PostgresCDN and edge

AI adoption

Retrieval-augmented generationOn-premises inferenceEvaluation harnessesData readiness

Governance

Architecture decision recordsRisk registersData protection impact assessments

What we do not do

Knowing where our usefulness stops saves everyone a procurement cycle.

  • We do not hold reseller or referral relationships with any vendor we might recommend. If we suggest you buy something, we do not get paid when you do.
  • We do not recommend a rewrite as an opening move. It is occasionally correct and almost always the most expensive answer available, so it needs a much higher standard of evidence than it usually gets.
  • We do not produce a roadmap detached from what it costs to run. A plan without the three-year operating cost in it is a wish list.
  • We do not recommend adding AI to a process nobody has measured. Without a baseline there is no way to tell afterwards whether it helped.

Questions we are asked

Everyone is telling us we need an AI strategy. Do we?

You need a decision about AI, which is not the same thing. In most small and mid-sized companies the useful version is narrow: one or two processes where the work is high-volume, text-shaped and currently done by someone expensive, with a measurable baseline to compare against. That is a page, not a strategy document. If the honest answer is that none of your processes qualify yet, we would rather tell you that than sell you a programme.

How do you cost something that has not been designed?

In ranges, with the assumptions stated separately from the numbers so you can see which assumption is doing the work. A cost estimate whose reasoning is hidden cannot be challenged, which makes it worse than useless in a board discussion. Where a range is too wide to decide on, that usually points at the assumption worth testing during Prove.

Will you recommend your own delivery service?

Sometimes, and where we do we will say plainly that we have an interest in the answer. Strategy engagements are priced and contracted separately from build engagements for exactly this reason, and you are free to take the roadmap to anyone. Several of the options we present will typically not involve us at all.

What if we already know what we want to do?

Then the useful engagement is much shorter: a challenge exercise on the decision you have already made, looking specifically for the assumption that would be most expensive to have wrong. That is a couple of weeks, and it either strengthens the plan or catches something. Both are good outcomes.

Related

Start with what is in the way.

Most of this work is shaped by the obstacle rather than the ambition. A deal that keeps stalling at the same stage, a process nobody owns, a product that will not sell, a decision that has been open for months. Tell us yours and we will say honestly whether we are the right people for it.