AI Opportunity Discovery
The skill students cite most. Roughly 95% of enterprise AI initiatives fail — and the root cause sits earlier than anyone looks: in how the business decides what to build in the first place. A complete, sequenced system for finding, qualifying, sizing, and prioritizing AI opportunities, and for defending those priorities against the forces that routinely derail them.
“A lot of topics piqued my attention, especially the initial assessment and opportunity discovery. How to get buy-in from C leaders was invaluable.”
Certification studentTAUGHT BY THE AUTHOR OF FROM DATA TO PROFIT
The failure is upstream, not in execution.
Most buyers assume their AI problem is a delivery problem. It usually isn’t — it’s a selection problem, and selection happens in a room you may not be in. Every problem below is one named or worked through in the course itself.
Part one · What’s broken at the companyA 95% failure rate on AI initiatives.
Resources consumed by initiatives that were never going to create value. Once a bad one reaches a roadmap and gets communicated to a board or to investors, it is nearly impossible to dislodge. Discovery is the only point of leverage.
→ Whole course · framed in Lesson 1Money is spent on the wrong opportunities because nobody sized them first.
Without upfront estimation there’s no basis for prioritization, budget, or buy-in — so the portfolio drifts toward whatever is loudest or newest.
→ Lesson 9 · The Opportunity Estimation FrameworkBolt-on AI consumes the budget bigger opportunities needed.
Replacing an existing feature with a more expensive AI version, with no re-engineered workflow and no new monetization, is a net negative that also forecloses better uses of the same resources.
→ Lesson 1 · bolt-on vs. re-engineered monetizationExpensive technology applied where cheap technology would do.
AI is among the most expensive options available and gets used by default rather than by justification. Complexity and uncertainty are the two-word test for when it’s actually the right tool.
→ Lesson 1 · the cheapest-viable-technology ruleUse-case thinking instead of pipeline thinking.
With only two or three bets in flight, the business cannot walk away from a failing one — “if we don’t do this, I don’t know what we’ll do.”
→ Lesson 2 · the opportunity pipelinePrioritization by squeaky wheel, or by coolest job title.
Resources redirected by the loudest customer or the most senior title, at the cost of committed initiatives with booked revenue — and displaced revenue is never surfaced.
→ Lesson 13 · when discovery goes wrongInitiatives blocked mid-flight by data problems.
Access, silos, extraction cost, and quality discovered months in rather than in the first week. Meanwhile technical teams are buried in an unfiltered request queue.
→ Lesson 8 · problem, data, and solution spaceMargin pressure answered with price increases, then layoffs.
The reflex when costs rise and pricing power falls — and the course’s central worked counter-example, reframed as growth. Discovery is treated throughout as a growth function, not a cost-reduction function.
→ Lesson 7 · the retail margin caseShipping products customers aren’t prepared for.
Real value trapped inside a product an unready customer base won’t adopt, with no part of go-to-market allocated to preparing them. Innovation with no adoption journey isn’t an opportunity.
→ Lesson 6 question 3 · Lesson 12Building what everyone else can build.
Without an identified information or data advantage, competitors replicate the initiative immediately and the opportunity evaporates. A 6–18 month transient advantage is fine — if you priced it in.
→ Lesson 7 · the four-point progressionData given away or left unmonetized.
Data-generating processes treated as exhaust rather than as an asset — including arrangements where a partner captures the value and the originator gets nothing.
→ Lesson 3 · Reddit, Disney, LyftDownstream breakage nobody anticipated.
Every major strategy change breaks something further down. Unadvertised risks that surface later turn the organization against the strategy.
→ Lesson 7 · downstream breakage analysisYou’re not in the room where it’s decided.
Opportunity discovery is where the year’s work gets chosen. If you aren’t in it, you inherit the results — and going back later to reopen the decision costs you credibility rather than winning the argument.
→ The whole system, sequenced to get you in the roomYou get handed initiatives you know won’t deliver.
Someone else’s disconnected KPI lands on your roadmap. You build it. When it produces nothing, the technology team absorbs the blame — and your role becomes a candidate for cutting.
→ Lessons 1, 7 · tying work to a critical KPIYou’re told how to do your job.
“Do this with AI” is an executive specifying your architecture. The diagnosis is uncomfortable but useful: it’s a symptom that leadership doesn’t trust the technical organization to connect technology to value on its own.
→ Lesson 13 · framework certaintyYou can’t quantify the value of your own work.
Without an ROI estimate you can’t defend a priority, justify a budget, or explain what’s lost by switching to the next shiny object. You’re reduced to arguing from opinion against people arguing from opinion.
→ Lesson 9 · three-band estimationYou can’t tell hype from a real opportunity.
Executives arrive enthusiastic and non-technical, having seen a demo. You need questions that filter hype without dampening enthusiasm.
→ Lesson 6 · the four questionsNobody says anything in the session.
Silence in the first session is close to universal, and it usually isn’t sabotage — it’s confusion about the mission, or fatigue, or not knowing what’s being asked for. There’s a documented recovery for it.
→ Lesson 13 · the silent roomYou’re too good at it and end up owning everything.
If you supply all the ideas, participants leave wondering why they were there — and you’ve lost the ownership transfer that makes the process stick. Worse: anyone who leaves feeling stupid doesn’t come back.
→ Lesson 13 · don’t take overYou don’t know how to define a problem without prescribing the solution.
The most common failure in requirements: too technical, insufficiently specific, and quietly dictating the implementation.
→ Lesson 8 · problem space definitionYou react to disruptions instead of anticipating them.
This is a learnable behavior — knowing what to listen for in what industry leaders say publicly, and what to do with it. Transition language, the money test, and constraints treated as paradigms.
→ Lessons 10–11 · pragmatic futurism, reading the tea leavesYou’ve never estimated something this uncertain.
Single-number estimates are impossible here and ranges feel like hedging — until you notice ranges are already how the C-suite communicates with the street.
→ Lesson 9 · underperform / expected / outperformTWO CONSTRUCTS APPEAR INSIDE ALMOST EVERY FRAMEWORK IN THIS COURSE. THE WORKFLOW IS THE UNIT OF EVERYTHING — THE TARGET OF DECOMPOSITION, THE SITE OF INTERVENTION, THE BASIS OF VALUE MEASUREMENT. COMPLEXITY AND UNCERTAINTY ARE THE QUALIFICATION TEST — WHEN YOU HEAR EITHER WORD FROM A LEADER, THAT’S THE TRIGGER. WHEN YOU HEAR NEITHER, IT’S PROBABLY A JOB FOR A CHEAPER TECHNOLOGY.
Outcomes, not trivia.
- Position technology as a strategic pillar — and define discovery as moving parts of the business and operating models into it
- Qualify AI as the right technology from first principles, using complexity and uncertainty as the test
- Run both discovery modes — top-down with executives, bottom-up with the frontline
- See disruptions before your competitors do, and harvest paradigms from what industry leaders say publicly
- Assess feasibility fast across problem, data, and solution space — without prescribing solutions to your technical teams
- Estimate opportunity size as a defensible range with a floor strong enough to carry the initiative alone
- Build and defend an opportunity pipeline instead of brittle use-case thinking
- Recover a session that has gone wrong — silence, hype-chasing, blanket objections, squeaky wheels
- Data & AI strategists
- Data & AI product managers
- Technology and product leaders
- Anyone who needs a seat in the room where AI investment decisions get made
A real employer and real opportunities. Every exercise asks you to substitute your own business, your own rivals, and your own strategic goals for the worked examples — a no-risk environment to hit the barriers you’d otherwise hit live, in front of your executives.
Six units. Thirteen lessons. Nothing hidden.
Watch in order — later lessons assume the vocabulary of earlier ones. Every framework is taught twice: how it should be in a perfect setup, and how it actually is. In 13 years of practice, Vin reports never having seen a perfect setup.
UNIT IBefore You Begin · Lessons 1–2+
Why discovery is the only leverage point, and the constructs everything else rests on.
- What happens before opportunity discovery. Why your discovery process must differ from the one the business already uses · meet the business where it is · what’s in it for them · optimize for constraints, maximize ROI · the first-principles definition of AI value creation, taught with zero technical content · the cheapest-viable-technology rule · bolt-on vs. re-engineered monetization
- A new mindset and understanding. Anatomy of an Insight — a workshop that reverse-engineers a finished insight to expose everything hidden beneath it · opportunity pipeline thinking · the maturity model, introductory pass · workflow change as intervention, the unit of analysis for the rest of the course
UNIT IIThe Monetization Paradigms · Lessons 3–5+
Where the money is, before you go looking for opportunities. Skipping this unit tends to produce technically sound opportunities with no monetization path.
- Data as an asset. Four competitive-advantage criteria for screening opportunities · reframing the business as a data-generating entity · two monetization modes, direct and aligned · the alignment guardrail
- A new paradigm of monetization. Ecosystem business models — AI platforms don’t have customers, they have partners · the scaling-access shift · creativity as the new productivity, and the AI economy as an optimization economy · knowing something about the market no one else knows
- Partnership monetization. The partnership opportunity checklist · the myth of “might be” · adjacency analysis · adversarial opportunity discovery, and when the correct output is a response plan rather than an initiative
UNIT IIISourcing Opportunities · Lessons 6–7+
Top-down, bottom-up, and the on-the-fly progression.
- The opportunity discovery frameworks. The three-pillar construct — business model, operating model, technology model · top-down discovery and the four hype-resistant questions: is the technology ready · is the business model ready · is it feasible for us · are we too late · bottom-up discovery as AI product governance, vetting frontline ideas for ROI before they reach technical teams · shelve, don’t say no · assumption budgeting per release
- Real-world opportunity discovery. The four-point progression for running discovery live, or in a hallway: what’s the critical KPI · why this technology · what’s the recommendation · what information advantage is this built on. Plus downstream breakage analysis, and trust as a precondition for big recommendations
UNIT IVQualifying & Sizing · Lessons 8–9+
Feasibility and estimation — run immediately after a discovery session, before roadmaps, estimates, or commitments.
- Opportunity feasibility assessments. Problem space — translate the opportunity into something buildable without prescribing how it gets built; define success in business KPIs, never model accuracy. Data space — is there low-cost access to a data-generating source, and synthetic data’s hard limit: it amplifies signal already present and cannot create new signal. Solution space — trust the team, ask don’t tell, listen for hedge language, ask for multiple candidate solutions
- The opportunity estimation framework. Why single-number estimates get rejected and ranges are the native language of the C-suite · the three bands — underperform (~95% certain, must carry the initiative by itself), expected (~80%), outperform (~50/50, included so you’re ready for it) · old workflow → new workflow as the estimation mechanic · why being wrong is acceptable and being directionally wrong is not
UNIT VSeeing Around Corners · Lessons 10–12+
Pragmatic futurism, paradigm spotting, and big-I innovation. Lesson 10 is named the most valuable sub-framework in the course.
- Pragmatic futurism. The four phases — disruption (a new technology breaks a specific, nameable assumption) → opportunity → customer → product. If you can’t define something buildable, pass. Four taught disruptions: the AI search paradigm · the death of reporting · time travel · local AI. Plus why incumbents don’t innovate until a startup forces them
- Finding opportunity paradigms. Reading the tea leaves — listen for transition language, apply the money test, treat stated constraints as paradigms too. Worked example: data scarcity and compute scarcity, and information asymmetry as a monetizable product
- Innovation opportunities. Capital-I innovation changes an assumption sitting underneath many products. Innovation creates new behaviors — so the adoption journey is part of the opportunity. If you can’t define an adoption journey, the opportunity isn’t real
UNIT VIWhen It Goes Wrong · Lesson 13 + Capstone+
The back-pocket toolkit. You can set everything up perfectly and it will still go badly. The recurring move: figure out why, find the root cause, then redirect — without ever directly refusing.
- The silent room — restate the mission, review business goals, prompt with their pain points, have ideas in your back pocket, but don’t take over
- Prescribing technical solutions — a trust problem, not fixable inside discovery; takes three or four quarters of delivery
- “We just need to do something with AI” — the Four Rs: reassurance, root cause, redirect, restatement
- The Queen problem (I want it all) — accelerate and redirect, then pivot to what we can do and what we should do
- The impossible external roadblock — sometimes regulation genuinely justifies shelving; it’s a failure mode only when one roadblock blocks everything
- “There’s not enough data” and “none of this works” — their evidence is real, the conclusion is wrong
- Prioritization by squeaky wheel or by job title — challenge the premise and ask for proof; or let other organizations do it for you by surfacing displaced revenue
Frameworks proven inside real enterprises.
Before you enroll.
How long do I have access?+
Do I need a technical background?+
Is there support after I enroll?+
Do I need to have taken the other courses first?+
Can my employer reimburse this?+
Go deeper, live.
The instructor-led AI Product Management certification covers this territory live, with weekly Q&A, a 1:1 with Vin, and a year of office hours.
Start today. Apply it this week.
30% of students see results from the frameworks before the course is even over.