Week 1 - 2: We observe the work 

 

Over the first two weeks, we embed ourselves just enough to see how work really happens. We pay attention to how analysis moves through the team, where decisions are made, and where they quietly stall. We examine how information is shared, where judgment resides, and how people collaborate under real constraints, not ideal ones.

 

This phase isn’t about tools or AI yet. It’s about understanding the rhythms, pressures, and informal practices that shape everyday work.

 

Week 3: We surface the team’s agentic capability

 

Every team already works with a form of agency; some decisions are automated, some are delegated, and others are tightly held by individuals. In week three, we map this agentic landscape:

 

• What the team decides independently

• What requires consensus or escalation

• Where judgment is exercised vs. avoided

 

This reveals the team’s unique strengths, blind spots, and decision patterns. It also shows where AI could support the team and where it would create friction or risk.

 

Week 4: We explore how AI fits individually and collectively

 

With a clear picture of how the team thinks and acts, we explore how AI could realistically be used:

 

• By individuals, in focused moments of work

• By the team, as a shared cognitive resource

• Across the organization, without breaking trust or policy

 

We don’t assume AI should be everywhere. We test where it adds clarity, where it reduces cognitive load, and where it should stay out of the way. 

 

Week 5: We design how AI works for this team

​

Using what we’ve learned, we shape clear roles between humans and AI. Where AI takes on volume, synthesis, or recall, and the team retains judgment, accountability, and consequence. It’s a set of interaction patterns designed specifically for how this team works, decides, and collaborates.​

 

Week 6: We stress-test the system through real scenarios

 

In the final week, we challenge the team and work through realistic scenarios involving:

 

• AI drift over time

• Ambiguous recommendations

• Over-trust and under-trust

• Team behaviour friction

• Operational and reputational risk

 

These scenarios surface where the system holds and where it needs reinforcement. The team learns how to recognize early warning signs and intervene before small issues become systemic problems.

​

By the end of six weeks, teams have:

 

• A clear understanding of their own capabilities

• A shared language for human–AI collaboration

• Designed ways of working that feel natural, not imposed

• The confidence to use AI without losing judgment or control

​​