Services
Five ways to engage.
You buy a business question answered — framed with an accountable sponsor, built on governed data, validated against figures you already trust.
IaaS
Innovation as a Service
Turning enterprise data into AI-driven advantage, one answerable question at a time.
Rather than a single large programme, value is delivered through a repeatable cycle in which each investment is tied to a defined business question and justified on its own merits. The unit of delivery is a question, not a programme phase, so value can be demonstrated early and investment can follow evidence.
The cycle
- Frame
- A defined business question, an accountable sponsor, a reference measure, and a confirmed view of data readiness.
- Build
- Sources connected in place, definitions agreed and governed, solutions validated against figures you already trust.
- Realise
- Adoption embedded; the agreed measure tracked against its reference point in your own terms.
- Scale
- Further use cases reuse the same foundation. Your teams are enabled to extend it.
Six service lines
- 01 · Frame · Value and data discovery
- A structured engagement translating strategic ambition into a prioritised portfolio of business questions, each with a value hypothesis, an accountable sponsor and an objective assessment of data readiness.
- 02 · Build · Governed use case delivery
- The core of the service: sources connected in place, definitions agreed and published, solutions built and validated against figures the client already trusts, results delivered with full lineage.
- 03 · Realise · Value realisation
- Adoption embedded and agreed measures tracked against their reference point — creating the evidence base for the next investment rather than asserting it.
- 04 · Embed · Capability transfer
- Client teams equipped to operate and extend the capability, with runbooks, definitions, a clear split of responsibilities and a defined path to self-sufficiency.
- 05 · Assure · Governed AI enablement
- AI made enterprise-ready: grounded retrieval over governed data inside the client boundary, provenance on every answer, policy enforced at the point of use, model-agnostic deployment including fully local.
- 06 · Sustain · Managed operation
- For clients who prefer a managed model: platform operation, a regular cadence of new use cases, and governance reviews anchored in business outcomes — with a clear transition path at any point.
The conventional alternative is a warehouse or transformation programme that requires migration and extended remodelling before any value is visible. Boards increasingly favour focused, evidence-led investments over large programmes with distant and uncertain returns.
AIaaS
AI as a Service
Governed AI capability, run for you, inside your boundary.
For organisations that want the capability without building the team: we operate the governed AI layer on your infrastructure, deliver a regular cadence of new use cases, and hold the governance reviews. Your data never leaves. Your models are yours to choose. You keep the option to take it in-house at any point, and we will help you do it.
What it includes
- Operated platform — Provisioning, monitoring, upgrade and capacity, on your infrastructure or a cloud you control.
- A cadence of use cases — An agreed number of new answerable questions delivered per quarter, each validated against your own figures.
- Grounded AI in the boundary — Ontology-grounded retrieval over governed data, with provenance on every answer and inference on local models.
- Agent operation — Governed agents configured, gated, monitored and reported — with their budgets and approval boundaries maintained.
- Cost control — Model routing, budgets that can decline, and spend attributed to team and use case.
- Governance reviews — Periodic review against business outcomes, with the evidence pack an auditor would ask for.
Most organisations cannot hire the people to run this, and most do not want to. What they want is the capability, the control and the audit trail. Those are separable from the headcount.
Professional services
Specialist Capability
Our engineers and specialists, with your team — on site or remote — working the problem alongside you.
A professional services capability. We place named specialists inside your team, in your context, on your problem — forward deployed engineers, AI specialists and data specialists, individually or as a small pod. Not consultants producing recommendations and not a support desk answering tickets: practitioners with the platform in their hands and the authority to build. They learn your domain, find where the real friction is, and ship something that works against your actual data within days instead of quarters.
- Forward deployed engineer
- Sits inside your team and builds. Owns a working outcome end to end, from understanding the business to something running against real data.
- AI specialist
- Grounding and retrieval design, evaluation and refusal behaviour, agent design and action-gate mapping, model selection and routing, and the evidence model an auditor would ask for.
- Data specialist
- Source reachability, identifier continuity, entity resolution, quality rules, definitions and lineage — the work that decides whether anything downstream can be trusted.
- Solution architect
- Deployment topology, entitlement and policy model, integration approach, and the security review pack.
- Delivery lead
- For a pod: scope, cadence, governance and the written handover plan.
How it works
- Embedded or remote, your call — On site in your workspace when the work benefits from it, remote when it does not — the same named people either way.
- Domain first, technology second — The first week is spent understanding the business, not configuring software.
- Ship in days — Working against your real data from the start. A prototype that survives contact with a sceptical user beats a specification that does not.
- Build the client's capability — Every engagement has an explicit handover plan. The measure of success is that we become unnecessary.
- One accountable person — You get named people, not a rotating bench.
- Individually or as a pod — One specialist, or a small team with a delivery lead. Scaled to the problem, not to a utilisation target.
When to use it
- The problem is not yet well-formed — You know something is wrong but not what to specify. An FDE finds the question worth answering.
- The estate is unusual — Legacy systems, unusual protocols, or data nobody has successfully joined before.
- Speed matters more than process — A regulatory deadline, a board commitment, a competitive window.
- A programme has stalled — Pilots that demonstrate well and never reach production usually have an architecture problem, not an enthusiasm problem.
- You want capability, not dependency — The engagement is designed to transfer, and the handover is written down at the start.
The hardest part of enterprise AI is rarely the model or even the platform. It is the translation between what a business actually needs and what a system can actually do. That translation does not happen over email. It happens when an engineer who can build is sitting in the room where the problem is described.
POC
Proof of Concept
One question, your own data, a working answer, and an honest verdict at the end.
A short, defined proof of concept against your own data, inside your own boundary. Agreed in advance: the question, what would count as success, who decides, and what happens next either way. It is designed to produce a decision rather than a demonstration — including the decision not to proceed, which is a legitimate and useful outcome.
What it includes
- A written success criterion — Agreed before anything is built, so nobody argues about the result afterwards.
- Your data, your boundary — Read-only access to the agreed objects. Nothing leaves. Nothing is written back.
- A working answer — Something a sceptical user can try, not a slide describing what it would do.
- An honest verdict — What worked, what did not, what it would take to make it production-grade, and whether we would advise doing so.
Most proofs of concept fail for the same two reasons: no agreed success criterion and no sponsor able to accept the result. We fix both before starting. If you want to find out whether AI can make a real difference to a specific problem you have, this is the cheapest honest way to know.
Discovery
Discovery and Assessment
A written answer to 'is this feasible, and what is it worth?' before anyone commits.
A structured, time-boxed assessment producing a documented view of your data readiness and a prioritised portfolio of questions worth answering. Source inventory, identifier continuity, exception granularity, volumes, retention and residency constraints, and the read pattern for each object — in writing, and useful to you whether or not you engage us further.
What it includes
- Prioritised question portfolio — The questions that are material to the business and not answerable today, ranked.
- Value hypothesis per question — Estimated from your own records, against a reference measure your finance function already reports.
- Data readiness assessment — Per system, per object: what is reachable, what is missing, and what would need to be agreed.
- Recommended first use case — With scope, approach and an indicative investment.
Opportunities that carry a defined question, an accountable owner, a reference point and accessible data move quickly from conversation to delivery. Where any of the four is missing, an assessment establishes it before anything is scoped, which is cheaper for everyone than discovering it mid-build.
Engagement models
Matched to how clearly the opportunity is defined.
- Scoped delivery
- Defined scope, price and deliverables. The standard model where the question and the data sources are clear enough to scope.
- Capability subscription
- A standing allocation drawn over time. Suited to clients with a portfolio of questions who prefer a continuous cadence to successive procurements.
- Discovery and assessment
- A structured feasibility and data-readiness assessment first. Suited to less well-understood estates.
- Outcome-linked
- A base fee plus a share of measured value. Available where a reliable baseline already exists — one option among several instead of the default.
- Embedded capacity
- Specialists within a wider transformation programme, including forward deployed engineers.
Outcome-linked pricing needs a reliable baseline. Where there is not one we will say so and propose something else, rather than create a measurement argument later.
Which of the five fits?
Tell us roughly where you are and we will say which engagement makes sense — including when the answer is none of them yet.