Approach

Review, then design and build, then run and evolve

It starts with a fixed-fee review. After it, and after each piece of work, you choose the next move. Nothing auto-renews.

How it starts

A 30-minute call about the AI you run. If we fit, we send a quote after the call. If not, we say so.

Step by step

Pick a move: price and what you keep.

Review

AI is becoming central, and nobody has the full picture.

We map the AI you run or plan to run: how it is built, what it can reach, what it costs. You get findings and a plan.

Who it's for

  • AI runs in several teams, each built its own way
  • Nobody knows what runs where, or what it costs
  • Customers, auditors or regulators ask how your AI is controlled
  • You face a build-or-buy or platform decision

How it runs

  • Map: each system with your engineers, including where the data goes.
  • Assess: each system against the three outcomes.
  • Plan: the gaps in priority order, and the design they point to.

What you get

  • A written review of every system
  • The gaps in priority order, with reasons
  • A target design, with each decision written down
  • The rules that apply to you (such as the EU AI Act, ISO/IEC 42001, NIST AI RMF) mapped to your systems

How it ends

  • Fixed fee, quoted after the first call.
  • The plan is yours, whether your team or we carry it out.
  • Can continue into Design & build, under a separate quote.

Technical detail

STRIDE threat model for AI, risk classification (for example EU AI Act Article 6), data flow maps, cost per system, and decision records with C4 diagrams.

Design & build

You know what to build, and it has to hold up.

We design and build with your engineers. Each agent reaches only what you decided, and a person approves the actions that matter.

Who it's for

  • Several teams need one foundation to build AI on
  • Your agents are moving from answers to actions
  • They need customer data, money or core systems
  • Several customers share one platform, and their data must stay apart

How it runs

  • Design: a shared foundation, which actions wait for a person, and what each agent can reach.
  • Build: code in your repository, in commits your team reviews.
  • Test: misuse testing before launch, plus evaluation baselines and cost measures.
  • Hand over: documentation and runbooks, so your team can run it.

What you get

  • The design and a record of each decision
  • Working, reviewed code in your repository
  • Permissions, approvals and data separation built in
  • Security test results, with findings closed
  • Evaluation baselines and cost measures, kept with the code
  • Compliance evidence for the rules that apply to you

How it ends

  • Fixed fee where the scope is defined, milestones where it is not. Quoted after the review.
  • Ends with handover to your team, or continues into Run & evolve.

Technical detail

Retrieval, tool use, multi-agent orchestration, cloud and edge serving. Per-tool permissions, approval policies, per-tenant data isolation. MCP and A2A. Agent security testing with Ziran and NVIDIA Garak. Evaluation suites. MLOps and LLMOps.

Run & evolve

Your AI is live, and models, tools and needs keep changing.

We stay with your teams as long as you choose: we review changes, keep systems and controls current, re-test, and step in when something goes wrong.

Who it's for

  • Your AI is in production and changes every month
  • New teams, models and agents keep arriving
  • You want a senior expert on call when things go wrong
  • Your regulatory obligations continue after launch

How it runs

  • A monthly retainer, scoped to the systems and risks we agree
  • Design changes reviewed before they ship
  • Re-testing and evaluation when agents, tools or models change
  • Incident response when a system fails or misbehaves
  • Compliance evidence kept up to date

What you get

  • Design and decision records kept current
  • A written review of each major change
  • Incident reports
  • A quarterly governance review

How it ends

  • 3-month minimum. Nothing auto-renews. Quoted after the review.
  • Runs while it helps you, and ends with your team in charge.

Technical detail

Threat model updates, post-market monitoring and technical documentation where regulation requires it (such as EU AI Act Annex IV), and AI incident response.

When we are not the right fit

If any of these apply, another firm will serve you better.

  • You want an exploratory pilot. We work on AI that is central to your business.
  • You need engineers by the hour. We take on defined work and do not place staff.
  • You want a strategy document with no build. We stay until the work runs.
  • You want one vendor's stack resold. We are independent of vendors, clouds and platforms.
  • Your engineers cannot work with us. Without them, the work does not stick.
  • You need pentesting of systems that are not AI. A pentest firm fits better.
  • You need a certificate. We prepare the evidence. An accredited auditor issues it.
  • You need a legal reading of AI rules. We turn rules into design. A law firm interprets them.

Start with a 30-minute call