AI’s Hidden Frontier: Unlocking Tribal Knowledge

This is a continuation of my series spotlighting the companies that are truly AI native: organizations rethinking work at its core, not just bolting algorithms onto legacy processes. Each profile explores how these pioneers treat AI as an operating system for execution rather than a surface-level add-on.

Notch represents one of the boldest examples. Instead of starting with process documents and standard operating procedures, they dig into the unwritten rules of how work really happens. Their focus is on tribal knowledge: the subtle judgment calls, shortcuts, and exceptions that define real execution but rarely make it into a manual.

The Notch Difference

Most AI projects start with documents: SOPs, workflows, training binders. Notch starts with what really happens. Four practices define their approach:

1. Find the hidden workflows – They sit with teams, listen to calls, and observe the “real” work: shortcuts, tacit rules, unwritten exceptions.
2. Break into decisions – They map the key if/then forks where agents apply judgment: when to escalate, how to recognize a VIP, which workaround resolves fastest.
3. Test and refine – Every mapped flow is validated against live cases. Gaps get filled. Loops get tightened. Exceptions get codified.
4. Make it usable – The output becomes a playbook: clear guides, rules, and mini-agent flows that can be shared, trained, or automated.

These steps turn tribal knowledge into scalable resolution logic. Notch deploys it as modular workflows: deterministic rules for high-stakes actions, AI models for messy interpretation, and specialized mini-agents to stitch it together.

Beyond Today’s Metrics

Founder Rafael Broshi is clear, the cost of resolution is collapsing: “Three years from now, it won’t be 60 cents or a dollar per case. It’ll be closer to one cent. At that point, the only thing that matters is whether you can resolve faster, cheaper, and more completely than anyone else.”

Deflection metrics won’t survive commoditization. Only resolution, AI that absorbs tribal knowledge well enough to take entire workflows off a rep’s plate, delivers durable savings.

How to Apply Notch’s Practices (Starting Now)

1. Shadow reality. Sit with your best agents. Capture what they actually do, not what’s in the binder.
2. Map decision forks. Chart the “if/then” pivots and exception paths. Don’t sanitize them.
3. Validate in live fire. Test flows against recent cases. Revise until agents nod and say, “Yes, that’s how I do it.”
4. Codify into tools. Turn the flow into checklists, playbooks, or automatable rules. Teach it, share it, or hand it to AI.
5. Track resolution, not deflection. Measure end-to-end resolution rates, cost per resolution, and CSAT; not how many tickets disappeared.

The Result: When leaders make it a practice to capture and scale tribal knowledge, the payoff is tangible. Costs fall as entire workflows move off human plates. Teams shrink without losing effectiveness. Handling times drop, customer satisfaction rises, and execution becomes more consistent. The gains don’t come from chasing surface metrics. They come from unlocking the hidden expertise that actually drives work forward.

This article was originally published at Forbes