In every PM job I have held — at IBM, at Teradata, in the classroom — the same skill has out-lived every tool, every framework, and every wave of new technology.
It is the only PM skill I would protect at all costs.
The skill is understanding the market.
Not roadmapping. Not prioritization frameworks. Not stakeholder management. Not technical fluency. Those skills are important, and they are also taught. There are books, courses, certifications, podcasts. A motivated PM can learn them in a year.
Market understanding is rarely taught, often dismissed as "analyst work," and almost always the difference between a product that ships and a product that ships and is needed.
Every product I have launched that succeeded did so because the team understood the market before they wrote the PRD. Every product I have launched that failed did so because we did not.
What "understanding the market" actually means
It is not one thing. It is three things, held simultaneously in the head of the PM:
01The shape of demand
Who is feeling which pain, how acutely, with how much budget. A PM who cannot describe the top three jobs-to-be-done in their market, sized by who is feeling them and who is paying for solutions, is not yet operating as a PM. They are operating as a project manager for engineering.
02The competitive landscape
Who is solving each part of that pain, with what differentiation, and whose share is moving. Not the analyst-deck version — the version where you can name the actual product, the actual buyer, and the actual reason the customer chose them.
03The technology arc
What is technically possible now that was not possible 12 months ago, and what will be possible 12 months from now. This is the part PMs miss most often. The platform shifts — LLMs, agents, sovereign data, vector retrieval, governance frameworks — create new buyers and reshape old ones at quarterly speed now.
A PM who can hold all three in their head at once — demand, competition, technology — is doing the actual job. They are also, by my count, in the top quintile of working PMs.
What AI changes — and what it doesn't
I get this question more often than any other from PMs in 2026: "Does AI do this for me now?"
The honest answer: AI changes the speed. It does not change the judgment.
What scales when AI does the work
- Reading every analyst report in a category in twenty minutes.
- Summarizing a hundred customer interviews into a structured table of named pains, frequencies, and severity.
- Scraping every product page in a competitive set and turning the differences into a side-by-side matrix.
- Finding every job posting at every relevant company and inferring who is investing in what.
- Watching pricing pages and flagging the week one of them changes.
- Reading a regulatory filing and pulling out the buying signals.
What still sits with you
- Telling you which of those signals matters.
- Telling you whether a customer's loud complaint is also a paying complaint.
- Distinguishing a feature your buyer wants from a feature your buyer's board wants them to want.
- Choosing which of three plausible market shapes the next eighteen months actually delivers.
Those are still judgment calls. They sit with you. A PM who treats AI as a research assistant doubles their throughput on the first list. A PM who treats AI as a replacement for the second list loses their job within two product cycles.
AI is a force multiplier on the things that scale. Judgment does not scale. That is why it remains worth paying for.
A live example — how to analyze the enterprise AI market right now
Let me show this concretely. Here is the question I want answered:
Which enterprise AI use cases are actually creating value in 2026, vs. which are noise?
A PM with this question has to assemble four things:
- The full list of named use cases enterprises are deploying.
- The actual adoption rate and budget attached to each.
- The maturity of the technical solution — can a buyer actually deploy it without an army?
- The unsolved gap — where the pain is loud but the supply is thin.
Where AI helps: collecting and structuring (1) and (2). Where the PM still does the work: judging (3) and synthesizing (4).
Here is what good output looks like. Two axes that matter: how much pain is the buyer in, and how mature is the solution. Place each named use case in the resulting 2×2. The map reveals which markets to enter and which to avoid.
Bright dots are high-leverage markets. Wedge markets — high pain, supply still emerging — are where new PMs can earn real differentiation. Now markets are where revenue lives today, but margins compress fastest. Commoditized markets are price wars. Demo-ware is theater.
What this map actually tells you
The map reveals four things a PM can act on tomorrow.
The Now markets are where the revenue is today. Code generation, customer service AI, document RAG, enterprise search — these are deployed at scale, the buyer knows what they want, the tech works. If you are a PM at a vendor in one of these markets, you are not exploring; you are competing on execution, packaging, and price. Your differentiation is operational, not visionary.
The Wedge markets are where the next billion-dollar product categories will form. Agentic workflows, AI governance, Trusted AI / observability, sovereign and on-premise deployment — these are markets where the pain is loud and growing, but the solution stack is still being defined. These are the markets I would steer a young PM toward today. The category is unsettled, which means a sharp product can still claim a piece of the definition.
The Commoditized quadrant is where great products go to die. Marketing copy, basic sales productivity, meeting summarization — the buyer can pick any one of fifteen solutions. The pain is real but moderate. Price compresses fast. Unless you have a structural advantage (data, distribution, an existing customer base), avoid building here.
Demo-ware is the trap. Voice AI agents that almost-work, synthetic media tools that look impressive in a launch video, co-pilots for niche applications that solve a 30-minute-a-week problem. The pain is too low. The solution is unproven. The market does not exist. I see one of these get funded approximately every six weeks. Most of them fail quietly.
A PM's job is to know which quadrant their product is in — and to be honest with themselves about it.
How to do this yourself, with AI as your research assistant
You can produce this kind of map in a day, not a quarter, if you use AI well. Here is the exact prompt sequence I would run.
Step 1 · Collect
Paste the prompt below into your AI tool of choice. Use a research-capable model that can browse the web (Claude with web search, ChatGPT with browsing, Perplexity, or Gemini Deep Research).
You are a market analyst supporting a senior product manager at an enterprise software company. List the top 15–20 named use cases for AI being deployed by Global 2000 enterprises in 2026. For each, give: (a) one-sentence description, (b) typical buyer persona (function and seniority), (c) named example vendors solving it, (d) rough indication of enterprise budget being allocated this year, and (e) one sentence on the leading objection or barrier. Cite sources for any quantitative claims. Output as a structured table.
Step 2 · Score
Take the table from Step 1 and ask the model to score each use case on the two axes that matter.
For each use case in the table above, give it two scores on a 1–5 scale. (1) Buyer pain: how acute is the felt need? Use deployment rate, growth in budget allocation, and analyst urgency as proxies. (2) Solution maturity: how deployable is today's technical solution by a non-AI-native enterprise team? Use published deployment timelines, named production references, and analyst rankings. For each score, give a one-sentence justification. Output as a CSV.
Step 3 · Judge
This is the part where the model stops doing the work and you start. Read the output, push back on anything that does not match what you have heard from real customers in the last 60 days, and place each use case on the 2×2. The model can draft your synthesis; only you can sign it.
Given the scoring above, write a 300-word strategic summary for a CPO audience: where is the "wedge" market — high pain, emerging supply — that a focused team could win? Where is the commoditized space we should avoid? Flag two assumptions in your analysis that a real practitioner should pressure-test before deciding.
The third prompt is the one that matters most. The reason: it produces a draft synthesis with explicit assumptions you can interrogate. Good market understanding is not about producing an analysis. It is about producing an analysis you can defend.
What I would ask of every PM reading this
Spend one day this quarter producing the map for your own market. Not the analyst version; your version, with the use cases you have actually heard customers describe, the vendors you have actually competed against, the technology shifts you can name. Use AI to do every part of that work that does not require judgment. Use yourself for the part that does.
You will be a better PM after this one day than you were before it. And your roadmap, the next time you write one, will be defensible to anyone who asks "why this and not that?"
The PM job is not going away. The PMs who treat market understanding as the core skill will be the ones who run the products that matter. The PMs who treat it as someone else's job — the analyst's, the strategist's, the CEO's — will be running PRDs.
Come learn how to do this with us. That's the work.