An AI Roadmap
Your Business Can Use
A £10m+ business usually has several AI ideas, a few unapproved tools and no agreed order for dealing with them. We turn that into a costed plan that operations, finance and IT can support, then help prove the first use case.
95%
NO MEASURABLE P&L IMPACT
36%
OF LARGE UK BUSINESSES USING AI
Costed
WITH OWNERS AND MEASURES
An AI roadmap needs to lead to a first build
A consultant runs a half-day workshop. The team votes on twenty possible use cases. A document arrives. Nothing happens. Six months later the same conversation starts up again, with a slightly different consultant.
MIT's Project NANDA reported 95% of organisations in its GenAI study saw no measurable P&L return. RAND cited estimates of 80%+ of AI projects failing. Gartner expects at least 30% of GenAI projects to be abandoned after proof of concept by the end of 2025.
The roadmap includes a first use case, owner, baseline and approval path so the plan can be tested through delivery.
A TYPICAL "AI STRATEGY"
- A workshop with twenty post-it ideas
- A maturity score nobody contests
- Use cases nobody owns or sizes
- "Build a data lake first" homework
- No mention of the ICO or the AI Act
- No one building it on Monday
WHAT YOU NEED INSTEAD
- Two or three named use cases
- Each one ranked by payback in £
- An owner, a cost, a date
- A first phase with an owner, baseline and approval route
- ICO, NCSC and AI Act requirements accounted for
- A written plan for the next twelve months
What the roadmap needs to cover
Leadership needs named projects, accountable owners, costs, dependencies and dates. The plan also has to give finance and IT enough evidence to approve the first one.
Where you are
Who uses what, on which accounts and with which data. Larger teams often find unapproved AI use before leadership has agreed a policy.
Where it pays
Use cases sized in hours saved or revenue freed up, not "transformation". Two or three winners, the rest binned or parked.
What you build first
One use case picked, scoped and live. Real users, real data, a real number against the spend at the end of it.
How you keep it safe
ICO data protection, NCSC secure AI guidelines, EU AI Act dates if you sell into the EU, ISO/IEC 42001 if a buyer asks. Written into the plan, not bolted on after.
How it compounds
Each use case feeds the next: shared data, shared evals, shared playbook. You stop running every AI project from scratch.
What the leadership team gets
An audit, a ranked plan, the first build. Then you decide whether you carry on with us, with your own team, or with no one.
Each phase has a written scope, decision owner, price and output. You can use the roadmap with us or hand it to your own team, and any build phase ends with the code, records and operating notes agreed in the scope.
BOOK A ROADMAP CALLAudit
We talk to your leadership, sit with two or three teams, read the tools they use, and look at the data behind them. You get a written report: what's running today (sanctioned and not), where the risk is, and where the hours are hiding.
Ranked plan in plain English
Two or three named use cases. For each: who it helps, how much it saves, what it costs to build, who owns it, and how it changes the job for the team doing it now. The ones that don't pay back get killed on the page, not in a steering committee a year later.
Build the first one
The top use case, live. Real users, real data, real measurement on the way in and out. Built on the AI stack we use across our agency work, with proper logging, evals and a safety gate. Your IP, your hosting if you want it.
Hand over a roadmap that keeps going
A twelve-month plan with dated milestones, a written AI usage policy your team can follow, and a short list of what's safe to do in ChatGPT and what isn't. Optional retainer if you'd like us building the next use case alongside you.
What UK adoption data says
Larger businesses are adopting AI faster, but most have yet to prove a revenue result. That makes prioritisation and measurement more useful than another list of tools.
36% of large UK businesses were using AI.
DSIT's 3,500-business study found adoption increased with company size. It also found most adopters reported higher productivity while most had yet to see a revenue change.
95% of GenAI pilots had no measurable P&L impact.
MIT's Project NANDA studied generative AI rollouts inside organisations and found 95% delivered no measurable P&L impact. The report says the gap is driven less by model quality than by whether the tool is embedded in the workflow.
AI adoption is now an operating priority.
The AI Opportunities Action Plan set out fifty recommendations covering infrastructure, adoption and public-sector use. The commercial question for an established business is still local: which process has an owner, usable data and an outcome worth measuring?
Sources: Department for Science, Innovation and Technology, AI Adoption Research, fieldwork February to May 2025. MIT Project NANDA, The GenAI Divide, 2025. HM Government, AI Opportunities Action Plan (CP 1241), 13 Jan 2025, and government response (CP 1242).
Costs, tax and governance
The ICO covers data protection, the NCSC publishes security guidance, and the EU AI Act can apply to UK businesses operating in or serving the EU. ISO/IEC 42001 also gives buyers a recognised AI management-system standard.
Data protection on AI
The Information Commissioner's Office has guidance on AI and data protection, plus the AI and Biometrics Strategy published on 5 June 2025. The Data (Use and Access) Act 2025 updated the rules around automated decisions. If your AI touches customer data, this is what applies.
Secure AI development
The National Cyber Security Centre's Guidelines for Secure AI System Development, written with the US CISA and twenty-one other agencies, covers secure design, development, deployment, operation and maintenance. A practical checklist for "is this thing safe to put in front of customers".
If you sell into Europe
Prohibited-practice rules and AI literacy duties have applied since 2 February 2025, with general-purpose model duties following on 2 August 2025. Article 50 transparency duties apply from 2 August 2026. The Commission's current timeline puts Annex III high-risk areas on 2 December 2027 and product-embedded high-risk systems on 2 August 2028. UK companies can still be in scope when an AI system is placed on the EU market or its output is used in the EU.
The new AI standard
Published in December 2023, ISO calls it the world's first AI management system standard. Think ISO 27001, but for AI governance and lifecycle. We make sure your plan doesn't get in the way of getting there later.
We use this approach inside our own company
Raq.com is the operating system behind our own AI work. It gives people and agents shared tools, knowledge, permissions and records, which means our advice comes from running these systems rather than presenting an AI maturity model.
When this is worth discussing
We work best when there is a real operating problem, enough volume to measure and people from the affected teams who can make decisions.
Usually a good fit
- An established UK business, usually with annual revenue above £10m
- A repeated process with a known cost, delay, error rate or capacity problem
- A senior sponsor and a day-to-day owner who understand the work
- Access to the relevant staff, systems, sample records and security requirements
We may point you elsewhere
- A standard product already covers the process well
- The requirement is a one-off small build with no wider operating case
- There is no owner or access to the people and data needed to test the result
- The plan relies on AI making high-impact decisions with nobody responsible for review
Questions before committing
We already have several AI pilots. Why do we need a roadmap?
Separate pilots compete for data, budget and technology time. A roadmap gives the business one set of criteria for value, risk and readiness, and makes ownership clear before more pilots are approved.
Should we wait until it's settled down?
No. Models and rules will keep changing. We favour use cases with a short payback period and keep the model behind an interface, so replacing it doesn't discard the whole investment.
Isn't ChatGPT enough?
For drafting emails and tidying up a CV, yes. For anything that touches customer records, integrates with your systems, or needs to give the same answer twice, no. ChatGPT is a tool. A roadmap tells you which jobs it's safe for, which ones need something purpose-built, and where the line is.
Don't we need to "get our data right" first?
That's the line every consultancy uses to sell a three-year data programme. You don't need a data lake. You need the right slice of your data in the right place for the first use case. We build that, deliver the use case, and the cleaned data is a free by-product. The rest gets done when the next use case needs it.
What about the EU AI Act and the ICO?
Written into the roadmap from day one, not bolted on at the end. The ICO's AI and Data Protection Risk Toolkit gives the playbook for the data side. The NCSC's secure AI guidelines cover the engineering side. If you sell into the EU, we map your use cases against the AI Act and tell you which ones are prohibited, which are high-risk, and which sit outside the parts that matter to you.
How much does it cost?
The audit and ranked plan is fixed-fee. The first build is priced against the chosen use case, scoped per phase. We tell you the number before we touch a line of code, and you can stop after the audit without doing the build.
What do you need from us?
A sponsor at the leadership table, three or four hours of time from two or three team leads, read access to whatever systems your use cases would touch. That's it. We don't need a data team or an AI champion. You can hand it to us non-technical and we translate.
Is there UK government funding for this?
Some manufacturers may qualify for Made Smarter support through a regional programme. Availability, eligibility and grant terms vary, so we check the current programme rather than including funding in the business case until it is confirmed.
Build an AI roadmap for your business
Tell us what the business has tried, which teams are asking for AI and where leadership expects a return. We will explain what a useful roadmap would need to settle before anyone approves the first project.