November 18, 2026

Did I Build the Right Thing? 6 Lessons from Ravi Mehta on Product Management in the Age of AI

Jessica Jess
Content Strategist, Voice of Customer

AI has made building software easier than ever. Former Tinder CPO Ravi Mehta explains why customer understanding, conviction, and knowing what not to build matter more than ever.

AI has made it faster and cheaper than ever to build software.

So why doesn't product management feel any easier?

That question framed a conversation between Ravi Mehta and Enterpret CEO Varun Sharma. Ravi is a product advisor who was previously an EIR at Reforge, CPO at Tinder, and a product leader at Facebook, Tripadvisor, and Xbox.

His answer: AI hasn't made product management easier. It's gotten teams to the hard part faster.

The hard part was never writing specs or drawing wireframes. It was understanding customers, turning that understanding into a strategy, and bringing an entire organization along with you.

Watch the full webinar on demand →

Here are six lessons for product teams navigating that shift.

1. Speed is about latency, not velocity

Annual planning and quarterly OKRs were built for a slower world. Ravi argued that they can now hold teams back.

Technology is advancing faster, switching costs are falling, and competition is coming from unexpected directions. For the first time, an online therapy company and a data analysis company can find themselves competing with the same product: ChatGPT.

Ravi learned some of his most important lessons about speed at Tripadvisor, where he came to distinguish between two things:

  • Velocity: How much work a team can accomplish.
  • Latency: How long it takes to go from a hypothesis to knowing whether it was right.

When most teams hear "move faster," they think about shipping more, hiring more engineers, or working longer hours.

But Ravi argued that the real opportunity is reducing latency.

Imagine having an idea today, getting it into the product tomorrow, and knowing whether it worked the day after.

The difference compounds. Review your goals monthly and you have 12 opportunities to adjust course each year. Review them weekly and you have roughly 50. Shorten that feedback loop even further and you create hundreds of opportunities to learn.

The goal isn't simply to do more work. It's to learn faster.

The takeaway: Shorten the time between a hypothesis and an answer. The faster your team learns, the more opportunities it has to build the right product.

2. Conviction comes from customer signal, not more data

When anyone can build almost anything, what actually sets great product teams apart?

For Ravi, it's how deeply they understand their customers and how much conviction they have in what to build.

And despite everything AI has changed, the foundation of that conviction remains surprisingly familiar: talking to customers.

Product teams often dismiss individual customer conversations as a sample size of one. But Ravi challenged that thinking, arguing that a single, in-depth conversation can reveal things that even large datasets miss.

Data can tell you what's happening. It rarely tells you why.

Most product managers have experienced this firsthand. A feature isn't performing, the dashboards aren't explaining it, and nobody can agree on the problem.

Then you spend 30 minutes with a customer and suddenly understand what's going wrong, why it matters, and what you might do differently.

The challenge is maintaining that proximity as companies grow.

At an early-stage startup, founders might answer support tickets themselves. Ravi described meeting the CEO of a product he uses because that CEO personally responded to his support request.

At a larger organization, a PM might speak directly with a customer only once a quarter.

That distance makes it harder to build conviction.

Ravi described conviction as a process of triangulation. Customer conversations are essential, but they're one part of a broader picture that includes experiment results, competitive behavior, customer workarounds, and patterns from other industries.

He also challenged the way product teams sometimes rely on A/B testing.

A statistically significant result can feel like certainty. But when teams rely on experiments alone to make decisions, they risk outsourcing their product judgment to the data.

An experiment should strengthen your understanding, not replace it.

The takeaway: Don't confuse statistical confidence with product conviction. The strongest decisions come from connecting customer conversations, behavioral data, experiments, and product judgment.

3. Build more, ship less

For years, engineering capacity was one of the biggest constraints in product development.

Building something required significant time and resources. So once a feature was built, teams naturally felt compelled to ship it, or at least run an A/B test.

AI is changing that equation.

When building becomes dramatically cheaper, the constraint shifts from what teams can create to what they should actually put in front of customers.

Ravi argued that the best product companies have always understood the importance of curation.

They don't automatically ship everything they build. They evaluate the experience again before launch and ask whether it deserves to reach customers.

If it does, they ship it. If it doesn't, they hold it back.

That discipline becomes even more important in the AI era.

Customers don't have unlimited attention for new features. Products don't have unlimited traffic for experiments. And a product assembled from dozens of individually successful A/B tests can eventually feel like a collection of disconnected ideas rather than a cohesive experience.

Varun challenged Ravi on this point.

If a feature is already built, wouldn't a team learn more by shipping it and seeing whether it works?

Ravi's response was that lowering the bar for shipping ultimately lowers the bar for customers.

Customers bear the cost of experiments that shouldn't have launched. Those who don't find value get a worse experience, while those who do find value may lose something they enjoyed when the feature is eventually removed.

Ravi shared how he approaches experimentation, emphasizing the importance of having a clear hypothesis, understanding what the team hopes to learn, and identifying which customer segments are most likely to respond.

The point isn't to experiment less. It's to make sure each experiment is grounded in a meaningful question rather than simply testing what's possible.

Varun added that the strongest hypotheses often originate from customer signals in the first place.

AI gives teams the ability to build more than ever. That doesn't mean customers should experience everything they build.

The takeaway: Use AI to expand what you can explore, not simply what you release. Building more should lead to better products, not more features.

4. Design for the human behind the agent

One of the biggest strategic questions facing software companies is what happens when customers stop interacting directly with their products.

Increasingly, software is being accessed through AI agents, APIs, and MCP servers rather than traditional interfaces.

Varun shared how Enterpret is seeing this shift firsthand, with customers increasingly accessing customer intelligence through agents and MCP rather than relying exclusively on a traditional web interface.

It's an example of a broader change facing software companies: the experience customers value may no longer happen entirely inside the product itself.

For product leaders, this creates a difficult choice.

Open your product to agents and you may lose some of the advantages of bringing users into your own interface.

Keep the walls high and customers may choose a more accessible alternative.

Ravi offered a useful way to approach the decision: there's still a human using your product, even if that person is several steps removed from the interface.

The job is to understand what that person is trying to accomplish.

He pointed to Descript's AI agent, Underlord, which can be accessed within the product, through an MCP server, or through an API.

The underlying capability may be similar, but the experience is different.

Inside the product, a person might direct an edit themselves. Through an API, a marketing team might automate the process of turning one long video into several shorter clips.

The human behind that automation doesn't want to supervise every step. They want the finished result.

Ravi also argued that the right approach depends on what makes a product difficult to replace.

Consider travel companies.

A platform with unique inventory, rich imagery, and a differentiated discovery experience has a stronger reason to bring customers into its own environment.

A platform selling inventory that's widely available elsewhere may have more to gain by making its capabilities accessible to agents.

For many companies, Ravi believes opening up is increasingly the safer choice.

AI agents are becoming capable enough to navigate existing interfaces anyway. Making a product accessible through APIs and agent-friendly experiences can create a better experience than forcing those interactions through the traditional UI.

It also changes the build-versus-buy decision.

When a product can serve as a building block in a larger workflow, customers can create experiences tailored to their needs without rebuilding the underlying capabilities themselves.

The takeaway: Design for the human behind the agent, not just the interface in front of them. Understand the job they're trying to accomplish and what makes your product worth using, regardless of how they access it.

5. Price the way your customers get value

AI isn't just changing how products are built and used. It's forcing companies to rethink how they charge for them.

Ravi argued that pricing should reflect how customers experience value, rather than simply passing a company's underlying costs along to them.

He shared the example of an AI audio company expanding into audiobooks.

Its costs may increase with every minute of audio consumed, but that doesn't mean customers want to pay by the minute.

People are accustomed to buying a book, not being charged each time they listen to it.

That distinction matters because usage-based pricing can unintentionally discourage the very behavior a product is designed to encourage.

Ravi offered a framework for thinking about AI pricing through two types of value.

Autonomy: The customer hands off a task and expects a completed result.

In these cases, outcome-based pricing can make sense. The customer is paying for work accomplished, much like hiring someone to complete a job.

Amplification: AI helps a person do their existing work faster or better.

Here, flat or tiered pricing may be a better fit. Customers can use the product freely without calculating the cost of every interaction.

Varun raised the challenge from the vendor's perspective.

With flat pricing, AI usage can increase dramatically while the company absorbs the additional cost. Historically, seat-based pricing naturally expanded as more people adopted a product. AI agents complicate that model.

As Ravi observed, agents are increasingly taking seats too.

The opportunity is finding a pricing model that allows customers to use the product in the way that's most valuable to them while ensuring the business benefits as that value grows.

The takeaway: Start with how customers experience value, not how your infrastructure incurs costs. The best pricing model encourages adoption while allowing your business to grow alongside customer success.

6. Make decisions by conviction, not consensus

The final audience question brought the conversation back to one of product management's most persistent challenges.

If building is getting easier, but stakeholder alignment remains difficult, how can large organizations make decisions faster?

Ravi started with culture.

Large organizations often default to consensus. Teams work toward decisions everyone can accept, but in doing so, they can lose the sharp thinking that made an idea compelling in the first place.

Ravi challenged the assumption that strong decisions require universal agreement.

In large organizations, consensus can sometimes come at the expense of conviction. Teams gravitate toward decisions everyone can accept, even when a more ambitious direction might create greater impact.

His point wasn't that disagreement is inherently valuable. It's that product leaders need the confidence to make well-reasoned decisions, even when not everyone agrees.

Making that work requires more than encouraging people to disagree.

Decision owners need genuine authority, executive support for controversial choices, and accountability for the results.

Ravi shared an example from his time at Tinder.

A young product leader championed Swipe Night, an interactive video experience that aired live on Sunday evenings.

It was an unusual bet for a dating app, particularly because its success depended on users participating at the same time.

But the decision wasn't made blindly.

The product leader had developed conviction through earlier experiments showing that when more people were online simultaneously, they swiped more, matched more, and started more conversations.

That customer understanding gave the team a reason to pursue a bold idea.

Ravi emphasized that executives have a responsibility here too.

If leaders give teams the authority to make risky decisions but punish them when those decisions don't work out, they'll quickly teach the organization to avoid risk altogether.

The result is more consensus, slower decisions, and fewer meaningful bets.

The takeaway: Give decision owners the authority and evidence to make bold choices. Reward thoughtful decision-making and learning, not just successful outcomes.

The hard part is still the job

AI has made it easier than ever to build software. But building faster doesn't mean building the right thing.

Throughout the conversation, Ravi returned to one idea: the teams that stand out won't necessarily be the ones shipping the most.

They'll be the ones who understand their customers deeply enough to know what matters, have the conviction to make difficult decisions, and the discipline to say no to everything else.

Because when anyone can build anything, deciding what to build is the job.

Want to hear more? Watch the full conversation with Ravi Mehta and Varun Sharma, including the story behind Tinder's Swipe Night, how AI is changing product strategy, and what it takes to build products customers actually want.

Watch the webinar on demand →

This article summarizes key themes and insights from a November 17, 2026 conversation between Ravi Mehta and Varun Sharma, hosted by Enterpret. It is an editorial recap, not a verbatim transcript. Learn more about Ravi's work at ravi-mehta.com.

‍

Related Blogs
See all blogs
Announcements
Nov 17, 2026
Product reviews from Walmart, Target, Best Buy, and more, now in Enterpret
Announcements
Nov 10, 2026
Close the Loop: Automations built around your changelog
Voice of Customer
Nov 9, 2026
Insights Are Cheap. Decisions Are the Bottleneck: 5 Lessons from Zack Hamilton on the Future of CX
Product Insights
Nov 6, 2026
The Fundamentals Don't Change. The Training Does.

AI That Learns Your Business

Generic AI gives generic insights. Enterpret is trained on your data to speak your language.

Book a demo

Start transforming feedback into customer love.

Leading companies like Perplexity, Notion and Strava power customer intelligence with Enterpret

BOOK A DEMO