Design it once, properly.
Scope, feasibility, options, cost and a target architecture — written down before anyone starts building, so the thing that gets built is the thing you decided on.
You have something to build, and one shot at building it right.
- Discipline
- Architecture
- Engagements
- 04
- Typically bought by
- CTO · Head of Architecture · Engineering leadership
- Anchors
- £4,000 – £15,000+
Set every figure below against the build it protects — and against paying for that build a second time. That comparison is the whole argument for hiring an architect, and it is why the fees are published rather than quoted on request.
AI Discovery & Scoping
Define the problem, requirements, feasibility, options, costs and delivery approach.
- An executive mandate to "do something with AI" without a defined problem.
- Competing internal proposals with no consistent basis for comparison.
- A vendor proposal you cannot independently evaluate.
One defined problem branches into three costed options — buy, build and a hybrid — each assessed against feasibility, cost and delivery risk, converging on a single recommendation with the rejected options recorded.
- 01Discovery report with the problem statement and success criteria
- 02Feasibility assessment and evidence base
- 03Costed options analysis with a recommendation
- 04Delivery roadmap and resourcing shape
- 05Risk and governance flag register
- 06Executive read-out session
- 01Problem definition and the decision the system is meant to support
- 02Success criteria, and what would count as evidence of them
- 03Data availability, quality, rights and access constraints
- 04Technical feasibility assessment against the stated criteria
- 05Build, buy and hybrid options with their trade-offs made explicit
- 06Indicative cost envelope and run-rate drivers
- 07Delivery approach, sequencing and the capability required to execute it
- 08Early risk, governance and regulatory flags
AI Solution Design
Implementation-ready enterprise AI architecture and technical design.
- A funded initiative with no architecture behind it.
- An engineering team building without an agreed target design.
- A design that has not been tested against non-functional requirements.
A layered reference architecture: sources and data flow into retrieval and an orchestration layer, which calls model providers through an abstraction; guardrails and evaluation sit across the path to production, with human oversight and monitoring attached.
- 01Solution design document, implementation-ready
- 02Architecture and data-flow diagrams
- 03Architecture decision records with the rejected options recorded
- 04Evaluation and acceptance criteria
- 05Non-functional requirement set
- 06Control points mapped for governance sign-off
- 07Design walkthrough with the delivery team
- 01Target solution architecture and component design
- 02Data architecture, flows, lineage and retention
- 03Model, retrieval and orchestration strategy with the reasoning recorded
- 04Interface and integration contracts
- 05Evaluation strategy — what is measured, when, and against what threshold
- 06Non-functional requirements: performance, cost, resilience, scalability
- 07Security architecture and control points
- 08Human oversight and intervention design
- 09Observability, monitoring and incident paths
- 10Build sequencing and architectural decision records
AI Architecture Review
Independent assessment of an existing AI architecture.
- A system approaching production with no independent assessment.
- Costs or latency behaving very differently from the original assumptions.
- A vendor-delivered architecture you need verified before signing.
| Ref | Severity | Finding | Remediation | When |
|---|---|---|---|---|
| F-01 | High | No evaluation gate before release | Add threshold gate to the deploy path | Now |
| F-02 | High | Single model provider, no fallback | Provider abstraction + tested fallback | Now |
| F-03 | Medium | Retrieval index has no access control | Enforce entitlement at query time | Next |
| F-04 | Medium | Cost scales linearly with usage | Cache + routing by task complexity | Next |
| F-05 | Low | No architecture decision records | Record decisions from this point on | Later |
- 01Architecture review report with severity-rated findings
- 02Risk-weighted remediation backlog
- 03Target-state recommendation and transition options
- 04Evidence pack for architecture or assurance boards
- 05Findings walkthrough with the owning team
- 01Architecture assessment against the stated intent and requirements
- 02Data flow, dependency and coupling analysis
- 03Model and retrieval design review, including evaluation adequacy
- 04Scalability, performance and cost-behaviour assessment
- 05Security, access and data-exposure review
- 06Resilience, failure modes and fallback behaviour
- 07Operability: monitoring, drift detection, incident response
- 08Governance and control-point gaps
- 09Prioritised findings with remediation options
Enterprise AI Platform Architecture
Organisation-wide target and reference architecture with a transition roadmap.
- Multiple teams independently standing up their own AI stacks.
- No consistent way to evaluate, approve or monitor AI across the estate.
- Model and platform spend growing faster than the value being delivered.
A platform layer providing model access, evaluation, guardrails, identity and observability as shared capabilities, with three delivery teams building on paved paths above it instead of each standing up their own stack.
- 01Enterprise AI reference architecture
- 02Capability model and current-state gap assessment
- 03Reference patterns and paved-path definitions
- 04Vendor and build/buy position per capability
- 05Transition roadmap with phased decision gates
- 06Executive and architecture-board briefing pack
- 01Current-state assessment across teams, platforms and vendors
- 02Target reference architecture and the capabilities it must provide
- 03Model access, routing and vendor-abstraction strategy
- 04Shared evaluation, guardrail and safety tooling
- 05Data and knowledge-access architecture
- 06Identity, entitlement and tenancy model
- 07Observability, cost attribution and FinOps design
- 08Reference patterns and paved paths for delivery teams
- 09Governance and stage-gate integration
- 10Transition roadmap with sequencing, dependencies and decision points
Tell us what you are building, or what you have to defend.
A scoping conversation ends with a written view of the engagement, its deliverables, its duration and its fee. No obligation on either side.