AI decisions compound. The early ones hardest of all.
We work with leadership teams on the decisions that are difficult to reverse: what to build, what to buy, where agents genuinely create advantage, and where a person has to stay in the loop.
Committing budget to something that changes every quarter.
Most leadership teams are not short of AI ideas. They are short of a way to choose between them that will still look sensible in two years. Vendors are moving faster than procurement cycles, internal enthusiasm is uneven, and the cost of picking a direction and reversing it is high.
The result is usually one of two failure modes. Some organisations commit early to a single platform and discover the lock-in later. Others run a long tail of pilots that never converge on anything, because no one has decided what the operating model should be.
The decisions that set everything downstream.
Engagements are shaped around whichever of these is actually blocking you. Most start with two and widen from there.
AI operating model design
Who owns AI decisions, where capability sits, how work moves between central teams and the business, and what has to be standard rather than left to each team. This is usually the question underneath the others.
Build, buy or partner
An honest read on which capabilities are worth owning and which are becoming commodity. We include the cost of the integration you will own forever, which is the part these decisions usually miss.
Where agents create advantage
A prioritised view of where automation changes the economics of a process rather than shaving minutes off it, and where the effort is better spent elsewhere.
Where humans stay in control
Deciding in advance which actions require a person, and why. Doing this early is much cheaper than retrofitting oversight after a security or audit review asks for it.
Avoiding lock-in
Where committing to one cloud, one framework or one agent platform is a reasonable trade, and where it quietly removes options you will want back.
Organisational impact
What automation does to team shape, headcount planning and the skills you need to hire for. Leadership teams tend to want this discussed openly rather than implied.
Short, structured, and aimed at a decision.
Advisory work is easy to let drift. These engagements are scoped to produce something a leadership team can act on rather than a document that gets filed.
Understand the position
Interviews across leadership, engineering and risk, plus a look at what is already running. We are listening for where the real disagreements are, because those are usually the decisions worth spending time on.
Frame the options
We set out the realistic choices with their trade-offs, costs and reversibility, including the option of doing nothing yet. The aim is a decision your team can defend rather than a recommendation you have to take on trust.
Agree the direction
A working session to land the decisions, followed by a written position covering what was agreed, what was rejected, and what would have to change for that to be revisited.
Positions you can take into a board meeting.
- A written operating model covering ownership, standards and how work moves
- A prioritised map of where agents change process economics
- Build, buy and partner recommendations with the reasoning attached
- An explicit position on which actions keep a human in the loop
- The lock-in risks you are accepting, stated rather than assumed
- A sequence for the next two to three quarters
Questions we get asked about this work.
How long does a typical engagement run?
Most are measured in weeks rather than months. Advisory work loses value when it stretches, because the market moves underneath it. If a question needs longer than that, it is usually a delivery question wearing an advisory hat.
Do you recommend your own product?
Only where it fits, and we will say plainly when it does not. Plenty of organisations are not at the point of needing a control plane, and telling you that early is more useful than selling you one. The advisory practice and the platform are deliberately separate conversations.
Who needs to be involved from our side?
Whoever can actually make the decision, which usually means someone from technology leadership, someone from risk or security, and someone who owns a business outcome. Engagements that involve only one of those three tend to produce conclusions the other two will not accept.
We already have a strategy. What would you add?
Often the gap is not strategy but the operating model underneath it: who decides, what is standard, and where the exceptions get handled. That is where good strategies usually come apart, and it is where we spend most of our time.
Related: AI security and governance, agentic frameworks, the full consultancy practice.
Bring us the decision you keep postponing.
A free 30 minute conversation about where you are and what is actually blocking it.