ai-capability-pattern-matching

Uses a ready-made, research-based library of 13 AI capability patterns (see ../../references/ai-capability-pattern-library.md) to pose diagnostic questions to a new client/industry, to assemble a raw

SKILL.md

GitHub

AI Capability Pattern Matching

Purpose

Produces a raw list of AI opportunity candidates top-down, hypothesis-driven, using a ready-made capability pattern library — as a counterweight to the bottom-up, data-driven approach of ../task-level-decomposition-and-automation-fit/SKILL.md (process/task mining one task at a time). The two are complementary:

  • Bottom-up (task-level-decomposition): works through an existing process/role one task at a time and classifies each one. Strong when the process is already precisely described, but won't surface opportunities that would only emerge from redesigning the process.
  • Top-down (this skill): poses the client the diagnostic questions of 13 patterns before the process has been precisely described. Faster first-pass mapping, surfaces opportunities the client wouldn't have named themselves ("we hadn't thought about it that way"), but requires validation (point 4) before a candidate goes to scoring.

Use this skill especially at the start of a discovery workshop/ interview, before detailed process description has been done — and use ../task-level-decomposition-and-automation-fit/SKILL.md afterward, once a specific process has been selected for closer examination.

Anchored in research

  • ../../references/ai-capability-pattern-library.md — 13 patterns, abstracted from 81 verified use cases in a broad industry report (2026) and cross-checked against a second, independent AI use-case digest (63 cases, 16 functions).
  • ../../../opportunity-recognition/skills/pattern-and-analogy-connector/SKILL.md — the general Capability Pattern Mapping abstraction method, of which the pattern library is the concrete AI application.

Method

  1. Before the meeting: select the 4–6 most relevant patterns from the pattern library based on the client's industry/situation (not all 13 at once — too many questions at once drowns the conversation). Use the source material's industry-specific function weighting as a rough guide (e.g. manufacturing → patterns 4, 5, 6, 11; professional services → patterns 2, 3, 9).
  2. Pose each selected pattern's diagnostic question to the client as written, don't turn it into something more technical or AI-jargon heavy. The questions are deliberately phrased in business language, not technology language (e.g. "where does a highly paid expert have to search for anomalies..." not "could an LLM read documents...").
  3. Log every "yes, we have a situation like that" answer as a structured candidate: pattern name, the client's own description of the situation, who does the work today, estimated volume/ frequency (if known). Don't score it yet at this stage — that happens in point 5.
  4. Validate every candidate before further processing, with three checks:
    • Is the situation genuinely recurring/high enough volume to be an opportunity, or a one-off exception case?
    • Does the pattern's assumed AI type (Agentic/Physical/other) match the organization's current maturity level, or is there a gap risk (e.g. a Physical AI pattern in an organization with no sensor data at all)?
    • Is there an obvious reason why this would NOT work in this particular context (regulation, union agreement, safety criticality)? If so, flag it visibly, don't hide it.
  5. Move validated candidates into ../ai-opportunity-portfolio/SKILL.md for 5D scoring and 2x2 prioritization — this skill only produces the raw list, it doesn't prioritize.
  6. If the client doesn't recognize any pattern as their own, that's information in itself: either the organization is already highly automated in these areas, or the conversation hasn't reached the right level in the organization (try a different role/team) — don't force a fit.

What this skill does NOT do

  • Doesn't replace ../task-level-decomposition-and-automation-fit/SKILL.md — it produces a fast, hypothesis-driven raw list, not a precise task-level classification. It's worth using both in the same engagement, at different stages.
  • Doesn't score or prioritize candidates — that's ../ai-opportunity-portfolio/SKILL.md's job.
  • Doesn't claim that all 13 patterns fit every client — some patterns are clearly more industry-specific (e.g. the Physical AI patterns) than others.
  • Doesn't expand the pattern library's examples with details that aren't in ../../references/ai-capability-pattern-library.md — if you need a deeper example, refer to the original sources rather than filling in from memory.
  • Isn't an exhaustive listing of every possible AI opportunity — the 13 patterns are a curated sample, not a comprehensive taxonomy. New patterns will be found over time; add them to the library using the same method (see the final section of ../../references/ai-capability-pattern-library.md).

Refinement notes

The pattern library is built on a research basis (from two independent industry AI use-case reports). Areas to keep deepening with real practice:

  • your own observations on which patterns resonate most often with which types of clients
  • new patterns you've identified yourself but that aren't in the sources — add them to ../../references/ai-capability-pattern-library.md in the same format
  • a concrete workshop template/question sheet (into ../../references/)

Once this section is filled in and validated in practice, update the maturity field in skills_index.json to draft, validated, or canonical (see ../../../meta/maturity_levels.md). Don't add new fields to the frontmatter — name and description are the only ones allowed (see ../../../meta/frontmatter_schema.md).

Continue from here

  • Before this (the general method): ../../../opportunity-recognition/skills/pattern-and-analogy-connector/SKILL.md
  • A parallel, bottom-up approach: ../task-level-decomposition-and-automation-fit/SKILL.md
  • Next in this pack: ../ai-opportunity-portfolio/SKILL.md — scores and prioritizes the validated candidates this skill produces.
  • If the whole process is run as a paid engagement: ../ai-discovery-engagement-design/SKILL.md
  • A ready-made skill chain for this situation: see ../../../playbooks/
  • This pack's shared guardrails: ../../CLAUDE.md

References

  • ../../references/ai-capability-pattern-library.md — the library of 13 patterns with examples
  • ../../references/ — the pack's shared background material
  • ../../CLAUDE.md — the pack's shared guardrails