EU AI Act transparency duties apply now; high-risk duties from December 2027. Check your exposure
Insights About us Careers
Contact us
Solutions

AI Services for Executive and Strategy Teams

The most valuable thing we do at executive level is usually recommending that something is cancelled.

Leadership
Function group
Stop list
As useful as the roadmap
One policy
Twelve delivery plans

Most organisations of any size now have more AI activity than anybody can describe, and almost none of them have a defensible view of what it is producing.

In one paragraph

AI services for executive and strategy teams cover portfolio assessment across current and proposed activity, board level reporting on exposure and progress, governance framework design, vendor and build decisions, capability and readiness assessment, and independent review of business cases.

The portfolio view almost nobody has

Activity accumulates function by function, each with its own vendor, its own business case and its own definition of success. The aggregate picture is rarely assembled, which means the same questions recur at every board meeting without an answer.

  • Count what is actually running. Pilots, vendor features switched on inside existing products, and departmental tools bought on a card all belong in the inventory.
  • Separate produced value from projected value. Most portfolios contain a small number of things delivering and a larger number still describing what they will deliver.
  • Identify duplication. Three functions buying similar capability separately is common and expensive, and nobody sees it from inside a function.
  • Find the concentration. Dependency on one vendor, one model provider or one individual is a risk the functional view cannot show.
  • Be honest about the stop list. A portfolio where nothing is ever cancelled is a portfolio nobody is managing.
Worth knowing

The stop list is the deliverable executives value most

Every portfolio assessment we have run produced a list of activity that should be discontinued: pilots that proved the point and were never closed, tools with three users, duplicated capability, and projects whose sponsor has left. Recommending cancellation is uncomfortable for a supplier because it reduces the work available, and it is the finding executives act on fastest, because reallocating committed spend is easier than finding new budget.

Reporting to a board that has to ask harder questions

Boards are now asking about exposure, not just opportunity

The questions have shifted from what AI could do for us to what we are running, what it decides, who is accountable and what happens when it is wrong. A reporting pack built for the first question does not answer the second.

Report the uncomfortable numbers

Number of systems making or informing decisions about people, number with an appeal route, number tested for outcomes, number with a named owner. These are answerable and they are what a board should be asking.

Distinguish material from interesting

A board needs to know about the systems that could cause regulatory, financial or reputational harm. A list of every departmental tool is noise that hides them.

Progress reporting should use business measures

Adoption rates and model metrics do not tell a board whether the investment worked. The function's own performance measure does.

Governance that scales, and the build or buy question

DecisionOur usual readNote
One governance framework or severalOneThe obligations are largely the same; delivery plans differ by function
Central platform or federated toolingDependsOn whether your functions have comparable data and skills
Build or buy a general capabilityBuyDocument processing, transcription and general drafting are competitive categories
Build or buy your differentiating layerBuildYour own data, your own process, your own boundaries
Central AI team or embeddedBothStandards and review centrally; delivery close to the function
Vendor concentrationManage itPortability and exit terms matter more than current pricing
Pilot or production from the startProduction pathPilots that cannot become production are expensive demonstrations
Public commitments about AICarefulDisclosure obligations and greenwashing style scrutiny both apply
Worth knowing

Buy the general capability and build the part that is yours

Document processing, transcription, translation and general drafting are competitive product categories where building rarely repays the investment. What no product supplies is retrieval over your own material within your own confidentiality boundaries, models fitted to your own process, and integration with how your organisation actually works. That is a much smaller programme than building the general capability, and it is where any durable advantage sits. Organisations that invert this spend heavily and end up with a worse version of something they could have licensed.

Process

How an engagement runs

Inventory and honest assessment first, because the stop list funds the rest.

Weeks 1 to 3

Portfolio inventory

What is actually running, including vendor features and departmental tools.

Weeks 4 to 6

Assessment

Produced value against projected, duplication, concentration and ownership.

Weeks 7 to 10

Governance design

One framework, with delivery standards functions can actually apply.

Weeks 11 to 14

Board reporting

A pack answering exposure and progress in business measures.

Ongoing

Operation

Reassessed each cycle, with the stop list revisited rather than quietly dropped.

Deliverables

What you receive

A defensible view of what you are running, and what should stop.

01

Portfolio inventory

Everything actually running, including vendor features and tools bought locally.

02

Value assessment

Produced value separated from projected, by activity.

03

Stop list

Activity that should be discontinued, with the reasoning.

04

Concentration and dependency analysis

Vendor, model provider and key person exposure.

05

Governance framework

One policy and standard set, with delivery plans by function.

06

Board reporting pack

Exposure and progress in measures the business already uses.

Fit check

Is this the right starting point?

Worth being direct. There are situations in executive and strategy where custom AI work is the wrong spend, and those are listed rather than buried.

Worth doing if

  • Nobody can say how many AI systems the organisation is running.
  • The board is asking questions the current reporting does not answer.
  • Several functions have bought similar capability separately.
  • Pilots accumulate and none are ever formally closed.
  • Governance is being written twelve times in twelve functions.

Do something else if

  • An inventory is wanted without honest delivery status from the functions.
  • The assessment must not recommend cancelling anything.
  • Board reporting is expected in model metrics rather than business measures.
  • A single central platform is mandated regardless of functional readiness.
Questions

Frequently asked questions

Marked up with FAQPage schema so these answers can surface directly in search results and inside AI assistant responses.

What does a portfolio assessment actually produce?

An inventory of what is running, a separation of produced value from projected value, a map of duplication and concentration, and a stop list. The last one is the deliverable executives act on fastest, because reallocating committed spend is easier than finding new budget. Every assessment we have run has found pilots that proved their point and were never closed, tools with a handful of users, duplicated capability across functions and projects whose sponsor has since left.

Do we need one AI strategy or one per function?

One governance position and separate delivery plans. The obligations that matter are largely the same across functions, so a single policy, a single risk assessment approach and one set of documentation and testing standards is both cheaper and more defensible than twelve. What genuinely differs is the data, the systems, the measure of success and the appetite, and those belong in each function's own plan. Trying to run one delivery programme across every function usually produces a roadmap that satisfies nobody.

What should we build and what should we buy?

Buy the general capability and build the part that is specific to you. Document processing, transcription, translation and general drafting are competitive product categories where building rarely repays the effort. What no product gives you is retrieval over your own material within your own confidentiality boundaries, models fitted to your own process, and integration with how your organisation actually works. That is a far smaller programme and it is where any durable advantage sits.

What should we report to the board?

Exposure and progress, in business measures. On exposure: how many systems make or inform decisions about people, how many have an appeal route, how many have been tested for outcomes, how many have a named accountable owner. Those are answerable and they are what a board should be asking. On progress, use the function's own performance measure rather than adoption rates or model metrics, which do not tell a board whether the investment worked.

Will you tell us not to do something?

Regularly, and it is the part of the work clients say they value most. We have recommended cancelling projects, buying instead of building, fixing data before modelling anything, and on several occasions concluding that the honest answer was that AI was not the right instrument for the problem described. A supplier who never recommends against their own further involvement is not giving you an independent assessment, and an executive team can usually tell the difference.

Tell us what the problem looks like.

Thirty minutes, no charge, no deck. We will tell you whether this is an AI problem, a data problem, or a process problem, and we will say when the honest answer is to buy something rather than build it.