Skip to module content
🧠AI Academy
0/91
Sign in
Module 11 · ~7 min

The Engagement Arc & Opportunity Radar

One page mapping every finding onto your delivery pipeline — plus the templated lead-magnet.

Reading progress
0/7 · 0%

The big idea

💡Key idea
Most AI initiatives fail for documented, avoidable reasons: generic tools, shallow integration, no organizational learning, no measurement. The winners pick narrow high-friction workflows, integrate deeply, keep expert humans in structurally designed loops, and let system and organization compound together. Every stage of that playbook is operationalized by an artifact you already build.
Quick check
1 question · instant feedback
0/1
  1. The engagement arc's discovery stage evidence-backed upgrade is:

Deep dive

2/2 open

Most AI initiatives fail for documented, avoidable reasons: generic tools, shallow workflow integration, no organizational learning, and no honest measurement.

The initiatives that succeed share a consistent pattern: they pick narrow, high-friction workflows; integrate deeply into how those workflows actually run; keep domain experts in structurally designed human oversight roles; and let both the system and the organization compound their learning over time.

Every stage of that playbook maps directly onto an artifact you already build — the Opportunity Radar, the ROI model, the risk register, the eval harness. None of it requires inventing new deliverables.

The consultant who can cite the research behind each decision AND ship the technical system is selling the only version of 'AI strategy' that the evidence actually supports.

The Opportunity Radar is designed to reach a defensible decision in 45 minutes — either solo or in one working session with a client.

You finish with three outputs: a ranked shortlist of opportunities, a named list of readiness gaps that become your engagement scope, and a 90-day proof plan with pre-agreed verdict criteria.

Every step is grounded in published research — MIT, HBS, OpenAI, Anthropic, NIST — and every step directly neutralizes one of the failure modes the MIT NANDA study documented. That research grounding is what separates this from the dozens of other 'AI readiness checklists' sitting in your prospect's inbox. The citations aren't decoration; they're the differentiator.

Quick check
1 question · instant feedback
0/1
  1. The Opportunity Radar's Step 4 (Readiness) failure to tick items becomes:

How to run it

  1. Discovery
    Inventory workflows (not brainstorm use cases); run the learning-loop audit; capture baselines during discovery; note shadow-AI usage as demand signal. Source: MIT NANDA · HBS §2 · §8.
  2. Opportunity scoring
    Gate candidates on domain specificity × workflow integration before rubric scoring; bias the inventory toward back-office friction; flag retrofit vs reimagine per candidate.
  3. ROI
    Three metric tiers; leading indicators; sensitivity ranges; the learning asset counted; 90-day verdict pre-agreed.
  4. Risk
    Register rebuilt on Govern/Map/Measure/Manage; cross-functional decision rights; traceability as an architectural requirement.
  5. Solution design
    Simplest shape first (the ladder); buy-vs-build priced with the 67/33 evidence; frontier map with structural human oversight at the boundary.
  6. Build
    Volume 1, end to end — the delivery layer that makes all of the above real.
  7. QA / evals
    Evals as the adoption unlock (lesson #1), the Measure function, the frontier locator, and the metric source — one harness, four strategic jobs.
Quick check
1 question · instant feedback
0/1
  1. Why do the vendor case studies (OpenAI, Anthropic) count as evidence here?

In the field

🔬Worked example
The AI Opportunity Radar: 5 steps, one 45-minute working session. Step 1 · Inventory workflows (not brainstorm use cases). Step 2 · Test each for a learning loop. Step 3 · Score survivors and pick a shape (via the domain × integration gate + the shape ladder). Step 4 · Check readiness (owner named, baseline captured, feedback capture designed, accountability agreed). Step 5 · A 90-day proof plan with pre-agreed verdict criteria. Offer as a free gated PDF or a facilitated session; rebrand the cover, keep the evidence citations.
🚫When not to reach for it
The Radar is a discovery tool, not a promise of automatic wins. A prospect who completes Steps 1–3 has produced your discovery inputs; Step 4's unticked lines are your engagement scope; Step 5 is the proposal skeleton — but the discipline of running it, per opportunity, is what separates the 5% from the 95%.
Quick check
1 question · instant feedback
0/1
  1. The four-sentence synthesis of the volume:

Pitfalls & takeaways

Failure modes

  • Selling any single element (evals, RAG, agents) without the pipeline discipline that gives it strategic meaning.
  • Skipping the Radar for a big client 'because they know what they want' — that's exactly the workshop-brainstorm failure the evidence names.
  • Treating the Radar as a lead magnet only, and not running it inside every engagement.

Durable takeaways

  • One synthesis: narrow + integrated + human-in-loop + measured = the 5%.
  • The Radar is diagnostic before proposal — mechanically qualifies leads.
  • Cite the research honestly; the caveats build credibility.
  • The evals-and-delivery layer is what makes the strategy defensible.

Do the work

📦Artifact to produce
AI Opportunity Radar (branded PDF + facilitated session variant).

Sources

  • · Synthesis: MIT NANDA · MIT SMR × BCG · Harvard Business School · OpenAI · Anthropic · NIST · Duke Fuqua