The questions that change the decision.
Operational answers about what we build, how work starts, when it stops and how claims stay inside the evidence.
Are you consultants or a software house?
Neither category describes the work alone. We combine process understanding, research when the answer is unknown, building and responsibility through adoption.
What is an External AI Department?
An external operating model close to the process. It is not staff augmentation, a partner division or unlimited availability.
What do you mean by capability?
Something the organisation could not previously do reliably and can then use, verify and govern. It includes people, process, data, system and controls.
Do we need to know what to build?
No. We start from the problem, constraint and what should become possible. Choosing the solution is part of the work.
How do you decide what is worth doing?
With explicit criteria, progressive proofs and the partner Capability Roadmap. Options include proceed, do not proceed, test, wait, scale, retire and reassess.
How are you different from SaaS or an internal AI team?
SaaS offers a product whose shape is already defined; we begin with the constraint and build only when a specific capability is missing. An internal team has its own organisational continuity and context; the External AI Department works within a bounded external scope alongside an internal owner, without replacing one.
When do you stop?
When access is insufficient, value does not justify cost and risk, the proof does not converge, adoption is unsustainable or an existing product solves the problem better.
How do you measure?
Criteria agreed before development, representative cases, errors, failure modes, real use, cost and value where measurable. There is no single metric for every capability.
What happens after deployment?
The capability can be reassessed against performance, use, technology, cost, risk and process change. That is Continuous Capability Evolution when agreed—not automatic perpetual maintenance.
Who owns the system?
Code, infrastructure, data, intellectual property and transfer are defined in the contract. The principle is to avoid opaque dependency and keep domain decisions and knowledge with the partner.
What do you reuse across clients?
Techniques, infrastructure and generic components may compound. Partner data, documents and operating know-how do not travel between organisations.
Do you only work in industry?
No as a method. Current public evidence does come mainly from the lab’s starting contexts; we do not turn cross-sector ambition into claimed experience.
Are the Capability Roadmap and the lab Roadmap the same?
No. One orders partner capabilities. The other describes the lab’s Now, Next and Horizon.
How do you handle security and data?
Depending on risk, the assessment considers data boundaries, access, criteria, human oversight, logging and responsibility. Concrete controls depend on the project.
Cloud, hybrid or on-premise?
It depends on data, integrations, risk, cost and desired control. Sovereign by choice makes that decision explicit without imposing one architecture.
Does your question depend on context?
Send it with the constraints that cannot be ignored.