Start with knowledge management, not tools
Audit and structure what the organisation knows before selecting AI tools; the limits of AI output are set by the limits of its input context.
Before selecting or deploying AI tools, an organisation should audit and structure its own knowledge. This is counterintuitive — tools are visible and concrete and have a procurement budget, while knowledge management is unglamorous and has no obvious owner. But AI’s performance depends almost entirely on the context it is working with, and context comes from knowledge. A well-chosen tool on top of a chaotic knowledge base produces confident wrong answers faster. The same tool on top of well-structured knowledge produces genuinely useful output.
The practical corollary is simple. When a client asks which AI tool to buy, the honest first answer is usually: before that, tell me how your knowledge is organised. If the answer is “it’s all in SharePoint” — or in Notion, or in Google Drive, or in the heads of senior staff — the next question is whether the documents are current, accurate, and structured in a way that AI can consume efficiently. For most organisations, the answer is no.
The reason is not complicated. The ceiling on what AI can do for an organisation is set by the inputs it is given. Tool-first sequencing treats that ceiling as adjustable through procurement, which it is not. Knowledge-first sequencing addresses the ceiling directly and makes every subsequent tool choice more valuable.
“First” does not mean “only”. AI strategy, data governance, leadership and technology infrastructure all have to be addressed in parallel — none of them can be deferred until knowledge management is finished, because knowledge management itself is never finished. The sequencing argument is narrower: of the various preconditions for useful AI, the state of the underlying knowledge is the hardest to bypass and the one most often assumed rather than examined. “First” means first among the actionable steps, not the exclusion of the rest.