Recap from the ProductLed Alliance webinar featuring Bagel AI CEO Ohad Biron
Every product team is shipping faster than ever. Developers are writing code with AI at a pace that would have seemed absurd two years ago. Marketers are cranking out content daily. Every silo in the organization is running at 10x velocity.
And product is becoming the bottleneck.
That’s the uncomfortable truth Ohad Biron, co-founder and CEO of Bagel AI, laid out during a recent webinar hosted by the ProductLed Alliance. The session, moderated by Nasi Rwigema (Head of Product at ProductLed Alliance), dug into what Ohad calls “context debt” and why the teams that win from here won’t be the ones that ship the most features. They’ll be the ones that know exactly which features to ship.
When shipping becomes cheap, decisions become expensive
Ohad opened with a reality check that’s easy to feel but hard to articulate. When building is fast and cheap, the bottleneck shifts upstream. The question isn’t “can we build it?” anymore. The question is “should we?”
As he put it: “The question is whether we’re becoming a feature factory just for the sake of meeting that expected velocity, or delivering the right thing that drives impact.”
Teams that chase velocity for velocity’s sake are accumulating what amounts to high-interest product debt. They’re shipping fast, but they’re shipping blind.
His framing of the new success metric was sharp: “If in the past we had to maximize impact, today we need to maximize the time to impact. We need to maximize the impact but in a shorter time.” Same ambition, compressed timeline, higher stakes on every decision.
The PM’s real moat is domain expertise
One of the strongest threads in the conversation was about what actually makes a product manager valuable in a world where AI can handle more and more of the execution work. Ohad’s answer was clear: “What will differentiate the excellent PMs versus the good ones is the domain expertise. How deep they know their customers, how deep they know the industries, how they interpret the data they see.”
Horizontal AI tools are getting better every day. They scrape competitors, summarize markets, generate analyses. But as Ohad put it: “No AI will tell us, because AI gets better and better of course, but especially the horizontal ones that look across the web and scrape our competitors… at the end they don’t know our industry like we do. That’s the job that will, at least for the next few years, remain by humans.”
It’s also the skill that most PMs are underinvesting in because they’re too busy building internal tools and playing with prompts.
The “work for work” trap
Speaking of internal tools, Ohad had a name for the growing obsession with building in-house AI systems: “If I start spending time building what I call work for work… I build tools so I can do my job better. That’s not my core business. That’s redundant.”
He wasn’t dismissing the excitement. He admitted that “taking it from zero to something tangible in an hour or less… this is really game-changing.” But he also flagged what happens next. Internal tools fail within two years because nobody budgets for maintaining them. The market moves faster than your internal system can evolve. “You go into a race of keeping up with the market versus enjoying what the market can bring to you.”
And on the emotional pull of building your own thing, he was honest: “We fall in love with it.”
Ohad’s background in enterprise systems (he started his career as an information systems engineer at Amdocs, managing internal HR systems) gave this point extra weight. He’s lived the build-vs-buy cycle. His take: leave the hard infrastructure to specialists and invest your focus where it compounds for your business. The classical buy-versus-build dilemma hasn’t changed. The speed at which you can build something has just made the trap more seductive.
Context is the moat, and it requires both qual and quant
The conversation naturally landed on voice of customer data. Ohad laid out three types of data that every product team should be watching. Analytics (how users engage with what you’ve built), external intelligence (what the market and competitors are doing), and what your current customers and prospects are actually saying.
That third category is where most teams are bleeding value. The data is already there. It’s in Gong calls, support tickets, NPS surveys, QBRs. But it’s fragmented, unstructured, and sitting in silos that nobody has the time to synthesize properly.
What makes this data powerful is context. As Ohad explained: “Whether it was said during evaluation or during onboarding or before renewal, who said it, whether it’s an important stakeholder or an end user… whether it’s by an account that we must retain or it’s not really in our ICP. All these nuances are important to at the end quantify all the pains and take the right decision.”
Ohad was emphatic that the qualitative side alone isn’t enough. Combining it with quantitative data (revenue impact, account size, usage patterns) is what gives product teams the confidence to commit to a direction and, just as importantly, to defend that decision to stakeholders later.
Justification and accountability aren’t going away
This was perhaps the most underappreciated point in the entire session. In a world where building is easy, the pressure to justify what you build and prove it mattered only increases.
Ohad referenced a recent conversation with Jenny Wen, Head of Design for Claude’s co-work product, who said something that stuck with him: “We are going through an immense transformation, but what still exists and will stay for sure in the near future is we still need people to decide what to build and why it matters.”
That’s the durability of the PM role. As Ohad framed it: “The justification part and the accountability part, that’s something that will only become tougher over time, because of all the noise and the ease of building and shipping.”
AI-native vs. AI-bolted
The final thread worth highlighting was Ohad’s honest assessment of the AI-native vs. bolt-on dilemma. He acknowledged the risks: AI-native products face real challenges around quality, validation, and regulatory trust. But the alternative is worse.
He pointed to Monday.com’s recent public admission that after two years of adding AI wrappers to their existing product, they realized they needed to rebuild from the ground up. As Ohad noted: “For two years they were building AI wrappers on top of Monday… but a few months ago they really understood that they’re doing something wrong and they have to change the whole architecture from the ground up, otherwise they will become obsolete.” A publicly listed, successful company came to that conclusion. For Ohad, that says everything about where the market is heading.
If you’re not rethinking architecture with AI at the core, someone else is. And their product will be fundamentally better, because they designed the whole system to work together in ways that a bolted-on approach simply can’t replicate.
The webinar was produced by ProductLed Alliance.
