Services

AI consultancy services, prices published

Five services: decide what to build, decide where it runs, build it, contain the agents, prove it is controlled. From a UK cloud region to fully air-gapped on your own accelerators.

  • NCSC CAF
  • NCSC secure AI
  • DISA STIG
  • CIS Level 2
  • NIST
01

AI readiness assessment

A review of your data, systems, controls and people. You get ranked AI opportunities with costs, risks and a build order, and a written note on what to leave alone.

From
£4,500
Duration
2-3 weeks
You end up with
A ranked roadmap, a risk register and a costed first build
Full details 02

Private, self-hosted & air-gapped AI

Open-weight models on your own hardware or in a UK region you control, up to fully air-gapped on NVIDIA DGX or HGX B300. Hosts hardened to DISA STIG and CIS Level 2.

From
£18,000
Duration
4-8 weeks
You end up with
A running private model estate, with monitoring and an exit plan
Full details 03

AI implementation

Retrieval over your documents, agents for variable work, automation inside the systems you already run. Built in your repository and handed over.

From
£25,000
Duration
6-12 weeks
You end up with
Working systems in production, with tests and documentation
Full details 04

AI agent setup & security

Coding, document and operations agents set up to do real work, each in a sandbox with least privilege, full logging and a kill switch someone has practised.

From
£18,000
Duration
4-8 weeks
You end up with
Agents in production, contained, logged and with a tested stop
Full details 05

AI governance & compliance

The policy, risk register and evidence a regulator, auditor or procurement panel will ask for. So AI proposals get approved, not parked.

From
£7,500
Duration
3-6 weeks
You end up with
An approved AI policy, a live risk register and audit-ready evidence
Full details

The usual sequence

How they fit together

You do not need all of them. Buying them as a bundle is usually a mistake. The common path:

  1. Assess

    Three weeks to establish what is worth building, what data it touches and where it can run. The list usually gets shorter.

  2. Decide where it runs

    Data classification decides the architecture. Sensitive workloads go private. Low-classification work can often use a hosted model under proper terms.

  3. Build the first thing

    One well-chosen system in production, measured against an evaluation set, handed to your team. If it acts, it goes in a sandbox with a kill switch.

  4. Make it defensible

    Policy, register, DPIA and evidence pack, so the next systems do not re-fight the same argument.

Governance often runs alongside the first build. Evidence is far cheaper to produce during construction than to reconstruct later.

Next step

Start with a 30-minute call

No deck. Tell us the problem. We will say whether AI is the right tool, what it would take and roughly what it would cost.

Book a call Pricing