Starting with what each team is for
Companies are complex. How functions work together, what each team measures, what it optimizes, how decisions actually get made: grounding all of this is the hardest part of enterprise AI, and nobody inside the company is trained to articulate it. We found a simpler way in. Every team exists for a goal. Every goal has measurable milestones. Every milestone has workflows. Break the enterprise into that mesh, and AI transformation becomes fast, grounded, and massive.
the whole model
one team
and workflows
layer of the warehouse
The digital brain, a context graph of the enterprise, is the right way to ground AI in how a company actually works. But building one the obvious way is slow. You have to capture structure, cross-functional behavior, metrics, and decision logic from people who have never had to explain any of it.
We reimagined the starting point. Each team exists for a goal. Each goal is proven by measurable milestones: hit them all, and the goal is achieved. Each milestone is moved by workflows the team already runs. Ask the right questions and any team can be captured this way in days, not months.
Do it across the org and the enterprise becomes a mesh of goals, milestones, and workflows. That mesh becomes the fourth layer of the Intelligence Warehouse, where every goal gets its own agent and OTP agents keep the local wins honest to the global ones. This paper walks through it, with examples.
01The grounding problem
Enterprises are complex structures. Not complicated like a machine, complex like an organism.
Sales depends on supply. Supply depends on finance. Finance constrains marketing. Every function measures different things, optimizes different things, and decides things in different ways.
For AI to be useful inside a company, all of that has to be grounded: written down in a form a machine can use. The structure. How functions work with each other. What each team measures and optimizes for. How decisions actually get made, and by whom.
Here is the hard part. Nobody inside the company can hand you this.
Not because they don't know it. They run on it every day. But people and teams are not trained to articulate how they operate. Ask a team how they work and you get a job description. Ask what they optimize and you get this quarter's number. Ask how they decide and you get "it depends."
The knowledge is real. The articulation does not exist.
02The right destination, and the brutal starting point
AI itself is not the problem. The models are ready. And we know what the destination looks like: a digital brain, a context graph of the enterprise, the Intelligence Warehouse. One connected graph of the company's entities, metrics, decisions, and judgment that every agent reasons against. This is the correct architecture, and it works.
The problem is the starting point.
Building that graph the obvious way means grounding everything at once: every system, every definition, every process, every decision rule. Months of discovery before anything runs. That cost is why most context-graph efforts stall in the mapping phase. The destination is right. The road into it was too long.
03The reimagining
So we changed the unit. Instead of grounding the whole company at once, we ground one thing: what each team is for.
Three statements. That's the entire model.
The second statement is the engine. A goal on its own is a wish. A goal with milestones is a checklist you can verify. And if the milestones don't add up to the goal, you have found a gap in how the company understands itself, before a single system is built.
Teams cannot articulate their goal directly. But they answer questions about it easily. What breaks if your team vanishes for a month? How do you know a good week from a bad week? Walk me through the last time it went wrong. Forty-five minutes of the right questions, and the goal, the milestones, and the workflows are all on the table, in the team's own words.
One team. A few conversations. Days, not months. That is the new starting point.
04What goals and milestones actually look like
This gets concrete fast, so let's make it concrete.
A compliance team. Everyone assumes compliance has no goal, they "just file things." Wrong. The goal: the company is due-diligence ready, any day, any hour. The milestones: every board meeting's minutes compliant. Every statutory filing done on time. Every license current. Hit all three, the goal holds. The workflows: the minutes review after each board meeting, the filing calendar run, the license renewal tracker.
An e-commerce sales team. The goal: grow and defend the brand's share online. The milestones: hero products in stock in the priority pincodes. Price within the agreed band of the competitor. The brand visible in top search results. Share above target in each city. The workflows: the daily availability check, the price response, the search-rank review, the weekly business review.
Now run the same lens across a company:
| Team | Goal | Milestones (hit all = goal achieved) | Workflows that move them |
|---|---|---|---|
| Supply chain | Serve demand at cost | Fill rate above target at every node · zero stockouts on hero SKUs · inventory days inside the band | Replenishment runs, reorder consolidation, expedite calls |
| Finance | Protect the margin pool | Gross margin above floor in every category · promo spend inside budget · working capital inside the band | Monthly margin review, funding approvals, variance analysis |
| Marketing | Win discovery and demand | Share of search above target · campaign ROI above hurdle · brand scores holding | Campaign reviews, spend reallocation, creative testing |
| Customer service | Keep the customers who want to stay | First response inside SLA · resolution rate above target · repeat complaints below threshold | Ticket triage, escalations, root-cause reviews |
| Talent / HR | Right people in the right seats | Critical roles filled within 45 days · regretted attrition below target · a named successor for every key role | Pipeline reviews, exit interviews, succession planning |
| Manufacturing | Make the plan, safely | OEE above target · zero lost-time incidents · schedule adherence above target | Shift handovers, maintenance windows, quality checks |
Read any row and notice three things. The goal never expires; it is why the team exists. The milestones are measurable; each one can be checked from data. And the workflows are not new work; they are what the team already does, finally connected to why they do it.
Notice one more thing. The targets inside milestones, the 45 days, the margin floor, change every planning cycle. The structure does not. Numbers are visitors. The mesh is permanent.
05Do it for every team, and the mesh appears
Capture one team this way and you have a well-understood team.
Capture every team and the structures start connecting. Supply's fill-rate milestone serves the sales team's availability milestone. Finance's margin floor bounds the sales team's price response. Marketing's share-of-search milestone feeds the same goal as sales' availability. One workflow moves several milestones. One milestone serves several goals.
It flows down, and it flows sideways. The sideways connections are not noise. They are how the company actually works, and for the first time they are written down.
The rollout even sequences itself. Every conversation names neighbors: whose number gets worse when you push yours, where do you escalate. The team named most often is the next conversation. And each team captured makes the next one cheaper, because half its connections are already known from the other side.
Every small action traces up to the big goal it serves. Every big goal traces down to the small actions that move it. That is what AI has been missing. With the mesh, a machine can win the micro goals while winning the macro ones, because the nesting is explicit.
06What this sets up: the transformation
The Intelligence Warehouse has three layers: entities, metrics, decisions. The mesh becomes the fourth.
Each goal's slice of the graph is a bounded territory with a measurable objective. That is exactly what you can hand to an agent.
One agent, one goal. It watches the milestones, runs the workflows, escalates where its team said to. The compliance agent keeps readiness at one hundred percent and thinks about nothing else. Small world, total accountability.
Above them, OTP agents optimize the larger goals the small ones roll into. When two local wins collide, growth pushing against margin, they apply the trade-off rule the teams agreed to, or hand the call to the human who owns it.
This is why the mesh makes transformation efficient. Micro agents deliver local wins in weeks, because each one has a real goal and measurable milestones from day one. And the wins roll up, by construction, into the enterprise goals they serve. Local speed, global direction, at the same time.
Governance comes built in. Every decision carries the escalation line its owner drew: "if it would breach the margin floor, it goes up. I don't take that one myself." The agent inherits that line as a wall.
And the whole thing compounds. Every run adds a case. Every planning cycle appends new targets. Every new team enriches the mesh. The system gets sharper with use, not staler.
07One more thing
The craft of drawing out what people know but have never said has a name. It is called consulting. In this model, consulting stops producing documents and starts producing the mesh itself: goals articulated, milestones pinned, workflows attached, typed straight into the graph the agents run on.
The deliverable no longer describes the company. The deliverable runs the company.
And it starts smaller than you would think. One team. Eight questions. Forty-five minutes. The mesh takes it from there.