AI-Augmented Product Discovery at a Fortune 500 Energy Company

The Challenge

A Fortune 500 energy technology company had a 300+ person product organization that was doing discovery the right way. Customer interviews, hypothesis-driven experiments, structured test-and-learn cycles. They had the mindset. They had the people. What they didn't have was a way to make it all scale.

Every customer interview generated hours of notes that someone had to process manually. Insights lived in scattered documents, slide decks, and people's heads. When a product manager needed to know what customers had already told them about a problem, the answer was usually "ask around and hope someone remembers." Connecting what the team learned from customers to what they actually built required manual effort at every step, from transcript to insight to roadmap to backlog.

For products already in production, the problem was worse. Customer signal wasn't just coming from interviews. It was buried in usability test recordings, defect reports, CRM notes, field team feedback, and industry research. All of it was relevant. None of it was connected. Teams were making product decisions with a fraction of the evidence they already had.

The organization was growing fast, adding new product lines and new teams. Every new team meant another silo of customer knowledge that nobody else could access. Leadership knew the discovery process was right. They also knew it was breaking under its own weight.

Key pain points:

• Customer interview transcripts took hours to process manually, creating a bottleneck between learning and action
• Insights were trapped in individual documents with no way to search, connect, or build on them across teams
• Product decisions were informed by whatever evidence a PM could find quickly, not what actually existed
• Usability tests, defect data, CRM notes, and field feedback never reached product teams in a usable form
• Roadmaps were disconnected from the customer evidence that should have driven them
• Rapid team growth compounded every problem. More teams meant more silos, more duplicated research, more missed connections.

The Approach

The solution was not more process. These teams already had good process. The solution was AI that could handle the grunt work between "we talked to a customer" and "here's what we should build," so product managers could focus on judgment, strategy, and the conversations that actually matter.

A single consultant designed and implemented the system across two phases, working directly with product teams on real products, not theoretical exercises.

Phase 1: New Product Discovery

For products being conceived and validated, the AI system supported the full discovery lifecycle:

• Interview planning: AI helped teams design interview guides, identify the right customer segments, and frame hypotheses worth testing. PMs still decided what to ask and who to talk to. The system eliminated the blank-page problem and ensured experiments had clear success criteria before they ran.
• Transcript processing: After every customer conversation, the system automatically extracted key themes, pain points, quotes, and signals. What used to take hours of manual note review happened in minutes.
• Insight synthesis: Individual interview findings were connected across conversations, surfacing patterns that no single PM would catch on their own. When three different customers in three different interviews described the same friction point, the system flagged it.
• Experiment tracking: Structured test cards and learning cards kept every hypothesis, method, result, and implication traceable. Teams could see exactly what they'd tested, what they'd learned, and what it meant for the product.
• Strategy connection: Validated insights linked directly to strategic initiatives, product roadmaps, and feature priorities. The path from "a customer told us this" to "that's why we're building this" was visible and auditable.

Phase 2: Continuous Discovery at Scale

For products already in production, the system expanded to synthesize customer signal from far beyond just interviews:

• Multi-source ingestion: The system pulled from usability test recordings, defect and issue triage logs, CRM notes, field team reports, and industry research. All of it fed into the same insight framework.
• Cross-source pattern detection: When a usability test revealed the same friction that customers described in interviews and that showed up in support tickets, the system connected those dots automatically. Product teams saw the full picture instead of fragments.
• Backlog intelligence: Discovery outputs fed directly into team backlogs, epics, and feature decisions. Evidence didn't sit in a report somewhere. It showed up where teams were already doing their planning.
• Portfolio-level visibility: Leadership could see discovery activity and insight patterns across all product lines. Gaps, overlaps, and unmet needs that were invisible at the team level became clear at the portfolio level.

The AI system was powerful, but only because the people using it understood discovery, experimentation, and evidence-based decision making. The tooling didn't replace the discipline. It made the discipline scale.

The Results

The most immediate impact was speed. Processing a customer interview went from a half-day task to something that happened automatically after the conversation ended. Product managers got their time back for the work that actually requires a human: deciding what the evidence means and what to do about it.

The deeper impact was connection. Insights from interviews, usability tests, defect logs, and field reports all lived in one place, linked to the strategic initiatives and product decisions they informed. When a PM updated a roadmap, they could point to the specific customer evidence behind every priority. When leadership reviewed the portfolio, they could see which product decisions were backed by data and which were still assumptions.

Teams that had been making decisions based on whoever argued loudest in the room started making decisions based on what customers actually said. The framework was described internally as "simple, learnable, executable," and leadership recommended expanding it to additional product lines.

One consultant designed and delivered the entire system. Not a team of twelve. Not an enterprise software implementation. One person who understood both the product discovery discipline and the AI capabilities needed to scale it.

What's Next

The foundation built here extends naturally. The same AI-augmented approach applies to competitive intelligence synthesis, automated win/loss analysis, and predictive roadmap prioritization based on accumulating customer evidence. The organization now has the infrastructure to make every product decision more informed than the last.