AI Reshuffle Opportunity Framing

AI Reshuffle Opportunity Framing

Purpose

Runs BEFORE ai-opportunity-portfolio, not instead of it — a premise check on how an AI opportunity is being framed, before it's scored. Most AI initiatives fail the same way: the team asks "how can AI make this existing process faster or cheaper" instead of "how does AI change what customers fundamentally need, and where does value now sit in this industry." The first question produces AI-enhanced ideas that get outcompeted by whoever asks the second question and rebuilds the value chain around it. Use this at the very start of AI opportunity work, before any idea reaches scoring.

Anchored in research

Both parts of this skill come from Sangeet Paul Choudary's Reshuffle: Who Wins When AI Restacks the Knowledge Economy (2025), winner of the 2025 Thinkers50 Strategy Award. Choudary's central thesis: strategy frameworks built from the 1970s through the early 2000s assumed stable industry structure; AI creates "structural uncertainty" instead — a model trained in one industry can be ported to another, so the old frameworks' fixed boundaries no longer hold. His core prescription, independently confirmed across multiple sources: companies don't need an AI strategy, they need a strategy for the world AI creates. The three-orders-of-effect model below is his own illustrative device from the same book, built on the historical shipping container.

Method

1. Ask the reframing question directly, before anything else. Not "how can we use AI to do this task faster or cheaper" but "how does AI change what our customers fundamentally need, and where does value now sit in our industry as a result?" If a proposed AI opportunity can only answer the first question, flag it explicitly as automation-framed before moving on — don't let it pass as strategic by default.
2. Apply the three-orders-of-effect test, using the shipping container as the reference model (a physical box didn't just make loading ships faster — it restructured global trade in three escalating stages):
   • 1st order — automation. The most visible, most obviously "AI" move: doing the same task faster or cheaper with AI in the loop. (Container case: port loading got faster.)
   • 2nd order — standardization. A less visible effect: the AI capability creates a common interface or format that lets previously incompatible systems, teams, or partners interoperate. (Container case: standardized dimensions let ships, trains, and trucks handle the same unit without repacking.)
   • 3rd order — unbundling and rebundling. The effect almost no one plans for, and the one that actually restructures the industry: previously vertically-integrated activities split apart and recombine around entirely new constraints, in a configuration that didn't exist before. (Container case: standardization enabled just-in-time logistics and global, disaggregated manufacturing — an industry structure the container itself never "automated," it made possible.)
3. Locate the proposed opportunity on this scale explicitly. Most ideas that reach this skill will honestly be 1st-order — that's not a failure, but it should be named, not disguised as transformative. The test this skill exists to run: is ANYONE in the room discussing the 3rd-order possibility, even if the immediate project stays 1st-order? An organization that never asks the 3rd-order question is the one that gets reshuffled by a competitor who does.
4. Name what would unbundle and what would re-link, for the 3rd-order case specifically. Which parts of the current value chain are only bundled together today because of a constraint AI is now removing (cost, coordination difficulty, scarce expertise)? Where would those parts re-link if a competitor — or a new entrant with no legacy structure to protect — designed the value chain fresh today?
5. Use Perplexity as the reference case for a genuine 3rd-order move, independently confirmed: it didn't automate search results (faster link lists), it occupies a different position in the value chain entirely — reading sources, synthesizing an answer directly, and citing for verification, eliminating the "click through and read it yourself" step search always assumed. By 2026 this reshuffled position supports a subscription-only, ~$500M ARR business with enterprise and API tiers — a business model automation alone wouldn't have created.
6. Check whether the opportunity assumes a strategy that isn't actually clear yet. AI doesn't fix organizational weaknesses, it amplifies whatever is already there — a well-documented 2026 concern, not a one-off warning (see References). If the team's strategic direction on this part of the business is genuinely unclear, an AI initiative built on top of it won't produce clarity, it will produce louder, faster output at whatever quality the underlying strategy already had — "strategic noise" at scale rather than strategic advantage. If this opportunity depends on a strategic premise the team hasn't actually agreed on, name that gap before scoring the opportunity, not after.
7. Hand off framed opportunities to scoring. Once an opportunity is explicitly located on the 1st/2nd/3rd-order scale and step 6's premise check is clear, it's ready for ai-opportunity-portfolio's 5-dimension scoring — this skill doesn't replace that scoring, it makes sure what enters it is honestly framed first.

What this skill does NOT do

• Doesn't score or prioritize opportunities — that's ai-opportunity-portfolio's job; this skill only tests how an opportunity is framed before it gets there.
• Doesn't claim every AI opportunity must be 3rd-order to be worth pursuing — plenty of legitimate, valuable AI work is 1st- or 2nd-order; the point is naming which one honestly, not forcing every idea toward the most ambitious framing.
• Doesn't design the actual rebundled value chain for a 3rd-order opportunity — it identifies that the question needs asking; designing the answer is deeper strategy and business-model work (see ../../../specialisation-packs/business-model-canvas/skills/bmc-innovation-pattern-matching/SKILL.md and ../ai-native-business-model-canvas/SKILL.md).

Refinement notes

• What's the clearest real case where a client's "obviously 1st-order" idea turned out to have a 3rd-order version worth naming, once this test was applied?
• How do you get a client team to take the 3rd-order question seriously instead of retreating to the comfortable 1st-order framing?
• Is there a fourth order worth naming in your own practice, beyond Choudary's three?

Continue from here

• Use first: before any AI opportunity reaches ../ai-opportunity-portfolio/SKILL.md.
• Related: ../capability-commoditization-tracking/SKILL.md — the same reshuffle logic applied to which of the company's OWN capabilities to keep investing in.
• Related: ../conways-law-ai-architecture-check/SKILL.md — checks whether the organization's own structure can actually support a 3rd-order move.
• Next, once an opportunity is framed: ../ai-opportunity-portfolio/SKILL.md, then ../ai-native-business-model-canvas/SKILL.md if the opportunity is transformative.
• This pack's shared guardrails: ../../CLAUDE.md

References

• ../../references/ai-native-reshuffle-heuristics-research.md — selection and grounding notes for this skill and its siblings
• Forbes Technology Council, "AI Won't Fix Organizational Weaknesses — It Will Amplify Them" (Aug 2026), and independent 2026 research on AI as a "strategic amplifier" — grounding for step 6's weakness-amplification caution (attributed to these independent sources rather than to a specific named individual whose exact quote on this point could not be verified — see ../../../human-ai-collaboration-design/references/hitl-partnership-heuristics-research.md for the verification detail)
• ../../references/ — the pack's shared background material
• ../../CLAUDE.md — the pack's shared guardrails