# Finding the Right Problem & Proving Value *You can execute flawlessly on the wrong problem and create nothing. The highest-leverage work a forward-deployed engineer does is choosing what to build and proving, on real data, that it was worth it. This post maps the discovery-to-value theme of the field guide.* Third in [The Forward Deployed Field Guide](/blog/posts/fdfg-01-understanding-customers-and-driving-ai-adoption.html). Having understood the customer and knowing how to build, the question becomes *what* to build — and how to prove it mattered. This theme sits between understanding and shipping, and it decides whether the engagement produces value or just output. ## The topics in this theme - **Discovery.** Turning a vague "AI could help here" into a specific, real problem — surfacing the actual pain, the people who have it, and (for AI) the fears that come with it. Discovery is the most underrated skill in the whole discipline. - **Use-case selection.** Treating candidate problems as a portfolio and choosing by business value against honest feasibility — not by what demos best. A flawless build on the wrong use case is wasted effort. - **Scope.** Defining the smallest thing that is genuinely valuable and genuinely real — a first win that would be worth having even if nothing else got built. Scoping well is half the engagement delivered. - **Prototypes and acceptance.** Building the thinnest thing that proves value on real data, and agreeing *in advance* what "good enough" means — the acceptance criteria (task, bar, sign-off) that turn a hopeful trial into a decision. - **Value.** Quantifying the outcome in the customer's terms — hours saved, cost avoided, revenue unlocked, risk reduced — so the win is a business case, not a vibe. ## The new AI topics reshaping it - **Eval-driven scoping.** For AI, "acceptance criteria" increasingly means an evaluation set — the customer's own labeled examples that define pass/fail. Defining the eval is now part of scoping, and it's one of the most valuable artifacts the engagement produces. - **The POC as the crux.** For AI products the proof-of-concept is where deals and projects are decided — and where they die if success was never defined. Scoping the production success criteria before the trial is the single highest-leverage move. - **Value measurement for probabilistic systems.** Proving value means showing the result against an agreed bar *and* that the feared risks (hallucination, cost, security) are retired — value and risk-reduction together. ## Where to go deep This theme maps closely to the pre-sales and delivery craft: [AI-Era Technical Sales](/blog/series/ai-era-technical-sales/) covers AI discovery, the POC, and the technical win; [Sales & PreSales for Engineers](/blog/series/sales-presales-for-engineers/) covers the fundamentals; and [The Forward Deployed Consultant](/blog/series/the-forward-deployed-consultant/) covers scoping an engagement to a first win. The [AI Production Readiness assessment](/ai-readiness/) is a concrete tool for part of this, and the [Forward-Deployed Playbook](/playbook/) frames the whole qualify → scope → build → prove loop. ## Why this theme is the pivot Everything upstream (understanding the customer) and downstream (building, shipping) is wasted if this theme is done poorly. Choosing the right problem and defining what proof looks like is where judgment matters most — and where an experienced forward-deployed engineer earns their keep, often by saying "not that one, this one" or "not yet." It's the difference between an engagement that produces an impressive artifact and one that produces value the customer can point to. ## Key takeaways - The **highest-leverage work** is choosing what to build and proving it mattered — you can execute flawlessly on the wrong problem and create nothing. - The topics: **discovery** (vague ask → real, specific problem + the AI-era fears), **use-case selection** (portfolio by value vs honest feasibility, not demo appeal), **scope** (the smallest genuinely-valuable, genuinely-real first win), **prototypes & acceptance** (thinnest proof on real data + agreed success criteria in advance), and **value** (quantified in the customer's terms). - **New AI dynamics:** **eval-driven scoping** (acceptance = an evaluation set; the eval is a prized artifact), **the POC as the crux** (where AI projects are decided or die — scope success criteria before the trial), and **value measurement for probabilistic systems** (result against a bar *and* risk retired). - Go deep via AI-Era Technical Sales, Sales & PreSales for Engineers, and The Forward Deployed Consultant; use the [AI Production Readiness assessment](/ai-readiness/) and [Playbook](/playbook/) as tools. - This theme is the **pivot** of the whole discipline — upstream understanding and downstream building are wasted if the wrong problem is chosen; judgment here (including "not that one" / "not yet") is where an experienced FDE earns their keep. ## Further reading - [Proof of concept — the trial that, for AI, decides whether the problem was worth solving](https://en.wikipedia.org/wiki/Proof_of_concept) - [Minimum viable product — scoping the smallest genuinely-valuable first win](https://en.wikipedia.org/wiki/Minimum_viable_product) --- Author: Pratik Dhanave (https://pratikdhanave.com) | Published: 2026-10-30 Canonical: https://pratikdhanave.com/blog/posts/fdfg-03-finding-the-right-problem-and-proving-value.html Attribution: please link the canonical URL when citing or quoting this post.