Our best people's know-how isn't written down — can AI even use it?
95% of generative AI pilots are failing because uncodified expert decision-making causes customer-facing AI agents to fail, triggering correction cascades that degrade performance and consume tokens.
The question
Our most valuable operational knowledge is tacit, held by a handful of experienced people, and what is written down is scattered across documents nobody maintains. Can we operationalize that knowledge so AI systems can actually use it, what does it take, and is it worth doing before we invest further in AI tooling?
Counsel's position
Yes, immediately launch a targeted program to operationalize high-value tacit knowledge into structured, agent-friendly formats, then integrate with AI tooling.
Verdict
The verdict: Yes, immediately launch a targeted program to operationalize high-value tacit knowledge into structured, agent-friendly formats, then integrate with AI tooling.
Uncodified expert decision-making causes customer-facing AI agents to fail
Given your scattered documentation, you must build "judgment infrastructure" that translates tacit principles into explicit rules before deploying agents.
Converting tacit judgment into AI rules requires a seven-step method
Given your reliance on experienced staff, operationalizing their knowledge requires systematically testing and refining AI assumptions rather than just writing prompts.
Unstructured institutional knowledge causes correction cascades that degrade agent performance
Given your scattered documentation, you must convert knowledge into structured, governance-aware units to prevent agents from constantly violating organization-specific conventions.
Agent-friendly knowledge fabrics reduce token consumption and accelerate time-to-answer
Given your unmaintained documents, you must rebuild your knowledge base using concise, event-driven formats designed specifically for AI consumption.
95% of generative AI pilots are failing without durable open standards
Given your hesitation to invest further in AI tooling, focusing on vendor-agnostic AI artifact catalogs protects your institutional knowledge from proprietary lock-in.
Read another verdict
- Buy a tool for this process, or build around our own knowledge?
- Our documents are a mess. Clean them up before AI, or after?
- How do we measure the return on an AI workflow — and what baseline is honest?
- Automate this workflow, or redesign it before we automate?
- Which process should we point AI at first?
- Our AI pilot works but nobody uses it — fix the workflow or kill it?
- Rent AI from a vendor, or run your own?
- Let non-developers ship AI-generated code?