We redesign how intelligence moves through work.
What gets noticed. Whose knowledge makes it into the room. How judgments are formed. Where AI and other systems participate. What gets checked. Who can make the call. Whether decisions become action. What the organization learns afterward.
We don’t start by asking where AI fits. We start by asking what intelligence the work requires.
What do you mean by “intelligence”?
Not individual IQ.
We mean the knowledge, judgment, and learning an organization needs to produce a breakthrough outcome.
The problem is often not that the organization lacks intelligence. It’s that intelligence fails to survive the organization’s ways of working.
How do you redesign the work?
We begin with the finished outcome.
What would have to be known, judged, challenged, and remembered for this work to be excellent? Then we trace that intelligence backward through the workflow. Only after we understand the work do we decide what should change.
Say we’re helping Nike. We imagine a new shoe drawing lines around the block at Footlocker. We then ask about all the forms of intelligence and know-how that would make that shoe possible: trend identification, customer obsession, competitive intelligence, supply chain analysis. We don’t just name these, we identify what about each is most critical to get right at this moment. We then trace back the process of producing that shoe and look for where those sources of intelligence entered, and how well they contributed to the final product. We go looking for where information got lost or degraded, where teams may have been silenced, where groups fell victim to cognitive biases. With that assessment, we design a new process to overcome these traps. And then we test that process on real work and adjust from what we learn.
Ultimately, a smarter organization must get better at three things.
What do you actually look for?
Our current working model groups combination failures into three kinds.
How do you decide what people and AI should each do?
We don’t divide work into a permanent list of “human tasks” and “AI tasks.”
Automating one part of a system changes the work around it. We design the combination, then revisit it as the technology changes.
This is also why we care about sequence. Showing a model’s answer first can anchor human judgment. Telling everyone to verify usually means nobody truly owns verification.
What trade-offs do you help us make?
A lot of organizational design is deciding what not to maximize.
We help leadership teams make those trades deliberately rather than discover them later as side effects.
Do you diagnose first and implement later?
Usually not.
An organization on paper is not the organization in action. We believe design and implementation are actually the same process. We learn where authority really lives by asking someone to make a decision. We learn whether a handoff works by changing it. We find political resistance by changing something people care about.
Assessment, design, and implementation inform one another. That matters even more with AI, because adoption changes roles, skill, and the work around the technology.
What do the changes actually look like?
Usually more concrete than “transformation” sounds.
The work may touch structure, but it often lives in everyday operating mechanisms.
- A leadership meeting where topics now end in explicit decisions.
- A handoff that carries context and rationale, not just a deliverable.
- An AI workflow where people form an independent view before seeing the recommendation.
- A named checking role with real capacity and escalation rules.
- A learning review whose answers return to the people who raised the issue.
- An exploration team with capacity that’s actually protected from lights-on work.
- A new operating rhythm with clear roles, a short protocol, and a loop that closes.
We prefer changes that can survive ordinary work. A practice that depends on permanent enthusiasm or heroic facilitation usually won’t.
How do you know it worked?
We don’t consider adoption a result.
Using more AI isn’t a result. Running the new meeting isn’t a result. Completing the training isn’t a result. We want to know whether the work got better.
- Decisions become more accurate or easier to defend.
- Decisions close faster and reopen less often.
- Rework caused by handoffs falls.
- Important actions actually close.
- The same failures recur less often.
- People retain critical skills without the tool.
- The combined human + machine workflow outperforms the person or machine alone, where the work allows that test.
We compare the organization against itself over time. We use behavioral outcomes where possible, leading indicators where necessary, and we look for ways the metric itself could be gamed.
Do you advise or implement?
Both.
We tend to work in one of two postures, and the same relationship can move between them.
Who do you work best with?
Someone with enough influence to change the organization, or who’s prepared to build and spend that influence. We usually have both a senior sponsor whose ambition is driving the work, and a strong internal operator who helps us understand how the organization really works.
What will you ask from us?
- Make room for the new work. We may ask you to kill, pause, or deprioritize something else.
- Involve the people doing the work. Some of the intelligence needed to redesign the system exists only in their lived experience.
- Make real decisions. We’ll push for clear authority and ownership.
- Let us challenge you. The relationship doesn’t work if the answer has already been decided.
When are you the wrong partner?
- When the real brief is downsizing.
- When you want blue-sky strategy detached from live work.
- When you need deep specialist expertise in one narrow function.
- When the sponsor can’t affect change and isn’t positioned to build the influence required.
- When leadership wants the organization to change but isn’t willing to reconsider its own assumptions or behavior.
How long does the work take?
Sparring relationships can be ongoing, or punctuated around consequential decisions. Embedded work often runs six to eighteen months, because the point is to change real working conditions, not just produce a recommendation.
What does success look like at the end?
The organization is better at making itself smarter without us.
We don’t need every transformation to be finished before we leave. We need clear ownership, working mechanisms for continued change, and people who’ve stopped waiting for someone else to fix every problem.