Most consulting teams do not have a memory.  They have decks, models, interview notes, meeting transcripts, Teams chats, partner instincts, and a few people carrying too much context in their heads.

That works until the initial enthusiasm of starting a new job wears off: the assumptions become tacit, contradictions are papered over and the team forgets why a decision was made. A finding that looked solid in week two becomes fragile in week six because the evidence around it has shifted.

This creates effort at the worst time. The week before a deliverable lands, when the manager asks where a number came from. During the executive review, when a senior client challenges a finding and the team realises the underlying assumption was never validated. The handover, when a consultant rolls off and takes their context with them.

This creates rework and can create late hours or late deliverables (or both); everyone accepts this as being part of the job.

I call the artifact team memory: a living record of what the work currently believes, why it believes it, how confident it is, and what has changed.

In practice, every assumption is captured. Every assumption is sourced: who said this, when, in what context. Every assumption carries a confidence rating: validated, working assumption, flagged unknown. When two inputs contradict each other, the contradiction is flagged rather than silently resolved.

The artifact isn't a deliverable, clients don't need to see it. It sits underneath the deliverables, holding them upright. Deliverables become consistent with each other by construction, not by heroic effort at the end.

The important point here is that team memory is not something the AI “has”. It is something the team builds and maintains. The artifact holds the memory; AI just makes it practical to keep current as the work changes.

Maintaining a record like this by hand was never realistic on a complex engagement. The administrative load was too high. AI doesn't change the judgement required: a senior consultant still has to decide what counts as a source, what confidence to assign, what contradiction matters. But the maintenance load that used to make the discipline unaffordable is now possible.

When a new piece of evidence contradicts an earlier finding, you find out the day it happens, as soon as the meeting ends, not the week before the deliverable lands.

Every consultant, lawyer, analyst, and researcher knows the week before a deliverable lands. The artifact removes most of what makes that week painful.

Team memory reveals misalignments that sit underneath the formal strategy:

Two senior stakeholders can be interviewed weeks apart and appear to agree. They use the same words. They support the same outcome. But one is assuming the executive team will make the trade-offs, while another is assuming the business will absorb them. One is talking about approval. The other is talking about ownership.

On paper, there is alignment, however in practice the organisation is carrying different assumptions about how the strategy will actually be realised.  Without team memory, those differences usually remain invisible until execution starts to drag. The work comes back for clarification and then people disagree about what was already agreed.

With team memory, the misalignment appears earlier, while it is still specific enough to resolve. The client can see that the issue is not commitment to the strategy, but the structure beneath it: who owns what, which trade-offs have actually been made, and where agreement is only verbal.

That is the real value, the artifact does not just help the consulting team manage information. It helps the client see why strategy that looked agreed is not yet executable.

It is worth being precise about this, because it is easy to describe team memory as if the AI is doing the remembering.

AI isn't deciding what counts as a source, it isn't rating confidence, it isn't deciding that two answers are subtly contradictory rather than superficially aligned. Those are human judgements and they remain human judgements. 

What AI changes is the cost of maintaining the record. 

That’s an infrastructure for judgement, not a replacement for it.

The WEF's Future of Jobs Report 2025 names analytical thinking and critical reasoning as the highest-value human skills of the next five years. Team memory matters because it gives the team space to be able to use those skills. In a shared artifact the team can collectively inspect, challenge, update, and carry forward.

Most discussion of AI in professional services has lived in two places. 

Between those two sits a layer that gets less attention, the team layer. Shared artifacts that hold complex multi-person work together and it’s worth taking the effort to build.

It is portable across consultants, a second person can pick up the engagement and be productive in a day rather than a fortnight, because the artifact is the context. 

It is replicable across engagements, the method travels even when the content doesn't. And it scales the judgement of good consultants, not by replacing them, but by giving them infrastructure.

If your consulting, advisory, strategy or transformation team is carrying too much context in people’s heads, get in touch. That is the problem I am working on.