There's an uncomfortable conversation happening in boardrooms across the world right now.
The question: "So... what are we doing with AI?"
The answer, for most companies, is increasingly awkward. According to PwC's 2026 Global CEO Survey, 56% of CEOs say their AI investments have yet to produce any meaningful financial benefits. Only 12% report both cost efficiencies and revenue gains.
Let that sink in. We're three years into the generative AI era, companies have poured billions into the technology, and more than half of them have nothing tangible to show for it.
What's going wrong?
The Shiny Object Problem
The pattern is familiar. A new technology emerges. Leadership gets excited. Budgets get allocated. Teams get formed. Pilots get launched.
And then... nothing much happens.
KPMG's 2026 Global Tech Report found that only 10% of organizations report high levels of AI maturity. The vast majority are stuck somewhere between "we're experimenting" and "we're not sure what we're doing."
Meanwhile, Deloitte's 2026 State of AI in the Enterprise reveals an even more telling statistic: only 25% of organizations have moved 40% or more of their AI experiments into production.
Three out of four companies can't get their AI projects past the pilot phase.
This isn't a technology problem. It's a direction problem.
The Missing Ingredient
Here's what I've observed working with product teams: the companies struggling with AI ROI share a common trait. They started with the technology and worked backwards toward problems.
"We need to use AI for something."
"What can we automate with AI?"
"How do we add AI to our product?"
These questions put technology at the center. But technology is never the center. Customers are.
The companies seeing real returns from AI? They started with a different question: "What do our customers actually need?"
This isn't revolutionary thinking. It's the oldest principle in product development. But in the rush to "do AI," many companies have forgotten it entirely.
The Customer Insight Gap
Consider what happens when a product team doesn't have clear visibility into customer needs:
They build features no one asked for. AI-powered this, intelligent that. Impressive demos, minimal adoption.
They automate the wrong things. Efficiency gains that don't translate to customer value don't translate to revenue.
They solve imaginary problems. Without real customer feedback, teams rely on assumptions. Assumptions are usually wrong.
They can't prioritize. When everything feels equally important (or equally uncertain), nothing gets the focus it deserves.
The result? Pilots that never graduate. Projects that stall. AI investments that produce impressive PowerPoint slides but disappointing P&L impact.
Why Customer Insights Are Harder Than They Should Be
If customer insights are so crucial, why don't more companies have them?
Because gathering, organizing, and acting on customer feedback is genuinely difficult.
Your customers are talking—in support tickets, sales calls, user interviews, NPS surveys, social media, product reviews, community forums. The signal is there. But it's scattered across dozens of channels, buried in unstructured text, and growing faster than any human team can process.
Most product teams face one of two situations:
Situation A: Data Drowning They have access to feedback, but there's too much of it. Customer calls pile up. Survey responses go unread. Support tickets get closed but never analyzed for patterns.
Situation B: Data Scarcity They know they should talk to customers more, but the infrastructure isn't there. Scheduling interviews is painful. Insights live in someone's Notion page. The team operates on vibes and HiPPO (Highest Paid Person's Opinion).
Neither situation leads to good AI investments. Because good AI investments require clarity on what problems are worth solving.
The Insight-Driven Approach to AI ROI
Companies that are succeeding with AI share a common playbook:
Step 1: Understand the actual problem
Before touching any technology, they develop a clear picture of customer pain points. Not assumed pain points—documented ones. With quotes. With frequency data. With severity assessments.
Step 2: Prioritize ruthlessly
Not every problem deserves an AI solution. The best teams identify where AI can create outsized value—typically in areas with high volume, pattern-based decisions, or tasks where speed directly impacts customer experience.
Step 3: Build with feedback loops
The AI project doesn't end at launch. Customer response shapes iteration. Usage data informs refinement. The solution evolves based on real-world impact, not internal assumptions.
Step 4: Measure what matters to customers
Not just "did we reduce costs?" but "did customers notice?" "Are they happier?" "Do they stay longer?" The ROI equation includes customer value, not just internal efficiency.
This approach isn't slower. It's actually faster—because you're not building things that nobody wants.
The Voice of Customer as Strategic Advantage
Here's the counterintuitive truth about AI: the companies winning aren't necessarily the ones with the best models or the biggest compute budgets.
They're the ones with the best understanding of what to build.
In Fortune's recent coverage of AI innovation challenges, Ronan Harris of Snap described the pressure facing executives: "The board has gone from being interested to being demanding when it comes to AI. I need to show up with demonstrable results."
Results come from solving real problems. Real problems come from listening to customers.
This makes Voice of Customer (VoC) infrastructure strategic. It's not a "nice to have" layer on top of product work. It's the foundation that determines whether AI investments pay off.
Think about it this way: every AI project is a bet. You're betting that the problem you're solving matters enough that customers will pay more, stay longer, or recommend you to others.
What's the quality of information feeding that bet?
If it's gut feelings and assumption chains, you're gambling. If it's systematically collected, analyzed, and prioritized customer insights, you're investing.
Practical Steps for Product Teams
If you're a product leader wondering how to connect AI investments to customer value, here's where to start:
Audit your feedback channels. Where is customer voice currently captured? Sales calls? Support? NPS? User interviews? Make a list. Identify gaps.
Assess your synthesis capability. How quickly can you go from "customer said X" to "we should build Y"? If the answer involves months of manual analysis, you have a bottleneck.
Connect insights to roadmap. Can you trace any current initiative back to specific customer feedback? If not, you might be building what you think customers want rather than what they've told you.
Instrument your AI projects. For every AI feature in development, what's the customer signal that justified it? What metrics will tell you if it worked?
Create feedback velocity. The faster you can understand customer response to changes, the faster you can iterate. Slow feedback loops mean slow learning.
The 44% Who Will Win
If 56% of companies aren't seeing AI returns, that means 44% are seeing something.
That minority isn't smarter about technology. They're smarter about customers.
They've built the infrastructure to systematically understand what users need, prioritize accordingly, and validate that their solutions actually work.
As boards get more demanding and AI budgets face more scrutiny, this capability becomes existential. You can't justify AI investments with demos. You justify them with customer outcomes.
The companies that survive the AI ROI reckoning won't be the ones who adopted the most sophisticated models.
They'll be the ones who never lost sight of the only question that matters: what do customers actually want?
Building something customers actually need starts with hearing them clearly. Pelin uses AI to surface insights from customer conversations, helping product teams prioritize what matters and skip what doesn't.