Patterns

Every new technology gets deployed in three stages. (1) Absorb — do the old work better. (2) Innovate — do things only possible with the new tech. (3) Disrupt — redefine the question entirely. Most organizations stop at stage 1 and mistake it for "doing AI."

The work that produces the deliverable (write the code, build the spreadsheet, draft the brief) is the *task*. The hard part — implicit knowledge, taste, opinion, judgment about what the deliverable should be — is the *job*. AI absorbs tasks. Jobs survive.

When automation makes a previously-expensive task cheap or free, the existential question for an incumbent is whether the cost of that task was the very thing protecting your business. If yes, your moat is now a puddle.

The most generative AI strategy question. Don't ask "how do I automate the existing process?" Ask "what previously-impossible question can I now ask?" Existing-process automation is incremental; new-question discovery is transformative.

Org-level maturity framework for engineering AI adoption. Five stages across six capability dimensions. Use to self-assess where the team is and what has to change to move up a stage.

AI should be used at the construction layer to produce deterministic, auditable, thermodynamically efficient runtime artifacts — not in the runtime itself. The agent's job is to write the workflow, not to be the workflow.