Leader Spotlight: Spearheading Crunchbase’s AI overhaul, with Megh Gautam

Megh Gautam is Chief Product Officer at Crunchbase. He began his career as an engineer at Microsoft before joining product management at Stanford University, where he concurrently got his Master of Management Science and Engineering, to monitor energy use across the campus. From there, Megh transitioned to Pivotal as a product manager before becoming a lead product manager at Finxera. Before his current role at Crunchbase, he served in various product and leadership positions at Hearsay Systems, Dropbox, and Twilio.

In our conversation, Megh talks about how Crunchbase’s product got a major redesign with the implementation of AI tools and features. He talks about how his team approached change, both internally and for longstanding customers. Megh also discusses the evolution of product management and how he foresees it changing alongside shifting consumer expectations.

---

Bringing AI into Crunchbase’s product

What was the core insight or driving force behind the AI overhaul of Crunchbase’s core product?

It all starts with the customers. A lot of our customers said, “Hey, I have a great deal of trust in Crunchbase. I want to use it in my day-to-day — in my processes and my work stream." As we were seeing the developments happening outside of us, like horizontal models and the frontier LLM companies, we thought we could do more with the unique and differentiated dataset we provide.

That led us to look at what was happening in the market and apply it here. From there was born this idea to use best-in-class LLMs to generate insights. For example, it can help users understand if a company is growing or not and whether it’s worth their time. That led to a lot of close co-development with our customers who were fully bought in, and are now implementing it. We’re working with them to figure out the operational parts of how the data touches every aspect of their operations.

How did you collect feedback from customers when evaluating this idea and driving this initiative?

Crunchbase operates through a self-serve method, which means users can go to our site and sign up for a subscription. We have sales customers who assign deals with us, as well as partnerships. We also run a Crunchbase Venture Program. There are many different modes of reaching the market and meeting people where they are. So, for each of them, it is vastly different.

On the self-serve model, you could run surveys. With direct sales, the most value comes from 1:1 or Zoom conversations. And when you get into tricky workflow conversations, that lends itself best to our whiteboard tool. That makes the most sense for the channel and time investment.

Navigating and introducing change

What was your process for deciding what parts of the Crunchbase experience should be reimagined through AI versus left untouched?

It started with good proofs of concept. We began with a search at the very beginning, and it's great to look at natural language. It breaks down the barrier of entry for a lot of people. You don't have to fiddle around with knobs and buttons — you can just ask the system what you want to do.

Gradually, it became apparent that this new wave of AI could help every single part of Crunchbase. We started with the way that we collect our data, how it’s represented, and how we can plug in our data through APIs into our customer sources. By the end, we changed the whole Crunchbase experience, from how people interacted with it and found it to pricing and packaging. It was a complete retooling of the experience.

What have been some of the most difficult challenges in making Crunchbase's AI as useful as possible, and how did you overcome them?

First and most importantly, it always starts with the customer. A lot of our customers were used to a particular way of doing things. If someone has interacted with the website in a particular way for the better part of the decade, we knew they might resist changes. There were a lot of sentiments like, "Can you make sure that the things that I used to do are not fundamentally broken? How do I find them, and can I get to them as easily as I could before?" That's the one kind of perception behavior change.

Second, we made a direct effort to ensure we could tell our users and our customers that there’s more that they can use Crunchbase for. It doesn't have to be to look up a company. They can build a list or start a search. They can use it across the whole company, and that was a huge impetus that we had in our mind when we came forward with this design.

Lastly, there were technical constraints. The token price has gone down 10x in the last year, so we wanted to pass down the benefits of the macro environment to our customers. That was great.

Can we follow the moving target and shifting behavior, and make people aware that they can use Crunchbase for all of these different use cases? All the grounds were shifting. The technology is moving, behavior is moving, and packaging is moving, so we had to make sure they all made sense.

It’s clear education was necessary in terms of the new user experience. Did you find that there were certain ways of getting that information out there to customers that worked best?

Totally. We did a lot of customer discovery to understand what words would make the most sense to land the messaging, because at that point, everyone was trying to build their own AI agents. We made a concerted effort not to call it a copilot because a lot of our customers would use it for entirely different reasons — everything from scouting and investment to figuring out the right kind of structure for your deals.

To get tied in with the copilot messaging would have been a disservice. We tested it with the widest variety of users that we could. We talked about it a lot more as it became real, and we did a lot of focus studies. Having open, honest, up-front conversations was great. Still, when you serve such a large population, a lot of people will wake up one day to a totally different user experience. We wanted to make sure we managed that instead of allowing them to be surprised in a bad way.

Surfacing context and having strong data

At a high level, how have you worked to maintain data integrity and user trust throughout this AI overhaul?

Your foundation of data dictates how good the AI signals are on top of it. We continue to invest in having the right guardrails in place. If there is a way for users to interact, we want to make sure we're responsive — and that we track the responses — and discrepancies are addressed over time. We take all the efforts possible to focus and double down. We pride ourselves on the integrity of our data. There are various data points, so we also want to ensure we have a lot more high-fidelity ways of capturing the data.

A classic example is saying, “Hey, this company had a funding round." But say it’s not a traditional equity round and was done through crowdfunding or other means. We needed to make sure that we captured the fidelity and the details of what happened across the board.

The other piece was the explainability of where Crunchbase gets its numbers from. Our highest-viewed help desk article is how we calculate a growth score. Customers want to trust and believe in this, and they want to be taken along for the ride. They often say, “We would have done it differently, but it's great to see your point of view." And then at the end of it, they look to integrate it best in their business processes.

What have been some of the most valuable benefits of transforming your product into an AI-based product?

A great benefit of the redesign has been the surfacing of contextual events that we also have data on. For example, a funding prediction would have said that this company is going to fundraise in the next six months with a 60 percent probability. Before, people would wonder what went into that analysis, but we would have never had the opportunity to have that conversation. People were like, "Ugh, funding happened. Now I have to do the legwork myself." It’s all about bringing a little bit more of the thinking about a user to the fore, and that’s fundamentally changed the way people interact with Crunchbase.

Further, we knew we were selling signals to people who wanted to get a signal. If you're a salesperson, you're like, "OK, I want to get the top 10 companies that are growing, because I know that they're perfect." We then got into this other class of customers who say, "I already know which companies are growing, but I want to know if they are growing here or outside, or if they have a financial expansion versus a geographic expansion." They were consuming parts of our signal in a way that would supplement and augment their own models.

Finding this out has been a really helpful learning experience across the board. We now have a way to reach them that’s more compelling — they can use us as one of the core data points to help make decisions.

The evolution of product management

From your perspective, how do you see the PM role evolving in this new AI world, and what does this look like in practice?

The PM and the design function are changing quite a bit. A lot of what we are trying to do is take information from our customers' pattern matching and make sure that we have the right solutions for our customers. There are parts of it that have become almost ritualistic. I wonder if the PM discipline, specifically as a function, shifted too much to being extremely recipe-driven. Now, we're going back to the era of things that help generate a recipe.

AI is really good at it. It can brainstorm metrics with you well, but the creativity of figuring out which one to choose is lacking. Those are all decisions that are grounded in strategy and your understanding of a customer's scenario. AI allows PMs to do what they should have always done, which is to make calls, be accountable for the results, and push as close to the customer as they can.

What emerging trends — both good and bad — do you think AI will surface for the PM role?

The people who end up taking responsibility for the entire life cycle of the product, from conception to launch, will do well. Creative people will do well, same with people who like spending time with their customers. However, if they are narrowly focused on their particular part of the puzzle, those blinders will cause a lot of struggle. There's going to be erosion in that world, both from other PMs and from people using AI systems. My advice is to stay aware of that change and set yourself up to make good calls that you are accountable for.

Further, as a PM, you need to have the right mindset. Fundamentally, you’re accountable for the product all the way, and you’re basically conducting an orchestra to make sure that all the functions work properly. But in the end, you also have to be at the frontier of what's changing.

At the same time, you just don't want to get lapped by things that don’t make sense, like getting an AI certificate, for example. Just understand where things fit into your particular workflow, and play around with them.

Everchanging consumer expectations

As consumer expectations continue to shift, especially with so many new technologies coming out, how do you re-evaluate product-market fit?

There aren't nearly enough words devoted to losing product-market fit. Product-market fit is a constant thing — you find it and you have to hold on for dear life to keep getting it.

The market moves very quickly, and now, if your competitors make changes and if the market shifts, there's not a lot you can do. You either keep up or you're going to get obsolete. And so, the way to go and rally around product-market fit is higher prices, more customers, enduring value, and better retention. And that isn't always a durable organization. It's not what got you to product-market fit, and the zero-to-one rule is going to let you scale. The thing that scales will look very different than the thing that brought you into the market the first time.

So many people may argue that AI is becoming table stakes. How are you thinking about your product road map and the future of your offerings at Crunchbase accordingly?

It’s one of the areas that we evaluate everything that we do. Is it helping our customers become faster, better at their jobs? Are we using AI to go do it? And then, fundamentally, how is the organization using AI to get to those particular outcomes?

We don't do PRDs. We have an audit checklist of things and a prototype. It’s just easier for people to poke around and say, "OK, yes, this makes sense.” Or, “I would change this aspect of it." These are higher-fidelity conversations, and AI has to be part of it across the board. Ideally, it's being used to both ship what you're building as well as the product itself that your customers are using.