AI you understand, trust, and can defend.

Yes. User readiness is one of the six dimensions in every assessment. We look at what tools your teams have, what they’re actually doing with them, where they’re improvising unsafely, and what training would close the gap. The training programme is included in the deployment engagement — covering your tools, your use cases, and your compliance boundaries. Not generic AI training.

That’s one of the core architecture decisions the assessment produces. The answer depends on your data classification, regulatory context, latency requirements, and the controls your security team already has in place. We document the decision, name the reasoning, and sign it — so it’s a written, evidenced conclusion, not an informal judgment call made during a vendor sales process.

We don’t sell tools and we have no preferred vendors. We assess what’s right for your specific stack, regulatory environment, and team — and we say so in writing before any deployment begins. If a tool you already have is the right answer, we’ll tell you. If it isn’t, we’ll tell you that too. Our fee is fixed and does not change based on which tools you adopt.

A governed workflow is one where data boundary, identity, prompt, model, output, reviewer, and retention are all written down before code is shipped, and where every run produces a linked, hash-stamped record that an auditor can open later. If any of those is missing, the work is not yet governed.

We score across six dimensions — data boundary, audit trail, identity, retrieval, rollback, and vendor posture — each on a 0–100 scale with documented evidence per score. The overall score is a weighted average; we tell you the weights, and we tell you which dimensions a regulator will check first.

The evidence packet — input, prompt, model version, output, reviewer, timestamp, and hash — is what auditors ask for, in the form they ask for it. We have walked the artifact through audits in finance, healthcare, energy, and public sector. You keep the packets, not us.

Every deployed workflow is pinned to a model version and ships with a rollback contract — written, signed, tested before production. Rollback is a normal operational lever, not an incident response.

Yours. Your cloud accounts, your identity provider, your secret manager, your observability. We are tool-agnostic and stack-agnostic by design — we deploy where you already operate, against the controls your security team already trusts.

Handover is the deliverable. Runbooks, on-call shape, retention contracts, dashboards, and the documented architecture are part of the engagement. We do not retain the workflow; you operate it from day one of production.

From a clean intake: 2–3 weeks for the readiness brief, 3–5 weeks for architecture design, 6–10 weeks for deployment. Most teams want a smaller first slice in production before they widen scope — we plan for that.

Only where you have written, signed permission for it to. Most engagements run with in-region inference and a closed data boundary. Where a third-party model is used, the contract is named, the residency is named, and the retention is named in the architecture document.

Your identity provider, your roles, your groups. No shadow identities, no service accounts standing in for real reviewers. Every action against the workflow resolves to a named person via your IdP.

A small, named team. The architect who wrote your design is the same person who deploys it. There is no offshore handoff and no rotating delivery team. You will have phone numbers.

Fixed-fee engagements priced by scope, not by hours. We send the price in writing before we begin, and we do not bill against scope changes we did not warn you about.