When shipping is cheap, focus is everything: A product leader's guide to the AI era

Throughout my 20-plus years in the tech industry, I’ve seen the product world from multiple angles. I’ve worked in research and development (R&D), served as an information systems engineer, and incubated next-generation products in large corporations. I then switched to the go-to-market side and helped scale a startup from scratch to over ten million in revenue.

One thing was constant everywhere I went: the friction and misalignment between product teams and their stakeholders. That persistent challenge is why I co-founded Bagel AI, a product intelligence platform designed to be the glue that connects product management with the rest of the business, helping you make decisions that truly drive growth.

Today, that friction is being amplified by a new force: the incredible speed of AI-driven development. It’s a challenge every product manager (PM) is feeling.

The double-edged sword of AI-driven development

The toolset we have today allows us to develop and iterate at an insane speed. The proliferation of internal AI initiatives is amazing, and it’s a huge advantage for companies that can build with an AI-native foundation.

Just think about the automations you can build in Slack with a simple prompt. We've seen companies build agents that can query all internal customer interactions or even turn Slack into their main ticketing system. At Bagel AI, we can build a new integration for a feedback source in about an hour – something that was impossible a year ago.

This speed is incredible, but it also creates a trap.

It’s tempting to fall into a feature factory mindset: building more and more without a clear purpose. There’s immense pressure from management to increase product velocity, and the ease of building makes it easy to say yes. If you don't navigate this carefully, you risk building things nobody cares about, hurting your core business and product quality in the process.

I was recently listening to a podcast with Jenny Wen, the Head of Design for Coda's AI offering, and she put it perfectly. She said that while we can do wonderful things with AI, someone will still need to decide what to build and why it matters. That core skill of a product manager isn’t going away; it’s becoming harder and more critical than ever.

Why feature velocity is a misleading metric

For too long, teams have measured success by how quickly they ship features. But in this new environment, we need to optimize for time-to-impact. The goal is to maximize the impact you deliver while minimizing the time to market. That’s the new formula for success.

This shift requires us to look at a new set of metrics that go beyond simple efficiency and focus on alignment and true business impact. As a product intelligence platform, we have the privilege of seeing how the most innovative product teams are rethinking their measurements.

Moving from time to value to time to impact

As a product manager, the expectation is that you will become an even deeper domain expert. You now have more time to do the job you were hired for: understanding customer pain points, what drives value for them, and how that translates to growth for your company.

In the past, connecting qualitative feedback with quantitative analytics was a heavy lift, often requiring a dedicated data science team. Today, you have the power to do real A/B testing and truly understand what moves the needle for your business. Your job is to identify that needle.

Successful companies know what drives their business, and it’s rarely a simple metric like ARR or logins. It might be transaction volume at a specific point in the customer journey or something else unique to your model.

Once you understand what to optimize for, you gain the focus we've been talking about. You'll know what to make ten times better and what to leave for others to solve.

A new metric to measure alignment: Rework ratio

One of the most insightful metrics we've seen our customers adopt is the rework ratio. This metric tracks how many features come back to development for enhancements, bug fixes, or other changes in a subsequent quarter.

A high rework ratio tells you a lot about the quality of what you're shipping and, more importantly, whether you actually solved the pain you intended to. If you didn’t properly synchronize with your go-to-market team or truly analyze the customer’s problem, you’ll end up building something based on a gut feeling or because one region shouted the loudest. It’s inevitable that this work will come back to you like a boomerang.

Shipping half-baked solutions or features that create more friction than they remove can tarnish your brand. Feature overload is a common issue; it creates noise and can cause busy users to bounce quickly. The rework ratio is a powerful way to keep your team accountable for delivering real, complete solutions.

The build vs. buy dilemma in the age of AI

Deciding whether to build a capability in-house or buy an external solution is not a new problem. However, it’s amplified today because it’s so much easier to build. The core principles for making this decision remain the same, and you can evaluate it using a simple three-part framework.

1. Is it your core business?

Every hour and resource you spend on an internal tool is an hour you’re not spending on your core product. You have to consider the opportunity cost.

Should you really spend mental focus on something that isn't your core business? This will require ongoing maintenance, enhancements, and a dedicated team to support it.

2. Can you rely on the system?

When you build a tool internally, especially one for analytics or intelligence, you have to ask: Do I trust the data? Can I confidently make decisions based on its output?

Building a reliable system requires deep domain expertise to preprocess data correctly and deliver accurate, contextual insights to the right people at the right time.

3. What is the true cost?

Right now, many companies have an open checkbook for AI experiments. The mandate is to run fast and figure it out later. But in a few months, leaders will start looking at their hosting and AI vendor bills. They’ll see how many tokens they’re using, and because optimizing that usage isn’t their core business, the costs will be high.

This will force a difficult conversation. Ultimately, you have to decide if bearing that cost and complexity makes sense, or if you’re better off partnering with an expert who can do it more efficiently.

This doesn't mean you should buy everything. Many things can and should be built internally now. What I am saying is that you need to be deliberate when you make that decision and continuously assess whether it’s the right one for your business.

Rethinking your roadmap strategy for a high-velocity world

The good news is that modern product intelligence tools give us the ability to build much better, more accurate roadmaps. By analyzing unstructured data from across the business, we can get a clearer picture of customer pain points and priorities. The key is to avoid creating “context debt” – getting so blurred by the noise that you lose sight of your strategic commitments.

Shortening your planning horizon

The faster you can ship, the smaller your units of work need to be. Think of it like a CI/CD approach for product strategy. To validate ideas quickly, you need to define a concise scope with clear acceptance criteria.

This speed also means that long-term roadmaps are becoming obsolete. As Jenny Wen also noted, you shouldn’t plan more than six to twelve months out. It’s almost impossible to plan further ahead with any accuracy. This makes your focus on the here-and-now and the mid-term even more important. Everything you do must feed that strategic north star.

Applying the fundamentals of good strategy

While the tools and timelines are changing, the fundamentals of product strategy are more relevant than ever. I highly recommend the book Good Strategy, Bad Strategy by Richard Rumelt, which outlines the three core kernels of a good strategy:

Diagnosis: First, you must understand and define the core problem or opportunity you’re going after.
Guiding policy: Once you know the problem, you need to develop a guiding policy for how you’ll address it – the right solution for the right audience. This is all about focus.
Coherent actions: Finally, you must develop a set of coherent actions to execute that policy. This involves continuous prioritization, validation, and optimization.

This framework has been around for ages, but today we need the mindset and the systems to run through this cycle much, much faster.

The blurring lines between product and engineering

One of the biggest challenges – and opportunities – facing product leaders today is the blurring of roles between product and engineering. We see more companies with a CPTO (Chief Product and Technology Officer), merging the two functions. PMs are writing production code, and engineers are getting more involved in product decisions.

The challenge is defining the new skill sets and figuring out who does what. PMs at Bagel, for example, are becoming more technical, but they can’t build highly sophisticated systems. Engineers can build those systems, but often lack the deep business context to know why they’re needed in the first place.

Building the right practices to support this new, hybrid model will be critical. The goal is to create a structure that accelerates company growth, not one that creates more noise, less quality, and instability.

When shipping is cheap, focus and judgment are everything. You need to trust your data and have clear ways to measure what’s relevant to your business. And as silos break down, building a common language and establishing trust across the organization is more important than ever.

The goal is to drive your company toward more impact, faster. Technology now enables this, and it’s our job as product leaders to implement the practices that will help our teams and our businesses thrive.

When shipping is cheap, focus is everything