A Playbook for Integrating GenAI into Your Product Roadmap

Your product roadmap will never be the same, but that doesn’t mean it needs to change all at once

We’re 18 months into the generative AI era, and the hype and attention paid to this remarkable technology — by pundits, by VCs, by your mom — has hardly died down. The pressure is on founders, chief product officers, and their teams to create dazzling new customer experiences, or at least sprinkle enough GenAI features into their products so it doesn’t look like they are laggards — and to do it without causing a hiccup on sales of existing products.

We’ve watched enough young companies navigate this tricky problem to have a point of view on the best way forward. By all means, lean in to GenAI now, but do so thoughtfully and with patience in the short term, even as you think boldly about how to use it to reshape your products and your market in the long run.

The fear of being left behind because your products aren’t already packed with chatbots and GenAI agents is understandable. But rushing to show customers, the market, and investors that you’re AI-enabled is not necessarily the answer. Instead, take a deep breath and stay focused on what your customers need today, and at the same time get to work on developing the enthusiasm, expertise, processes, and technology infrastructure to set you apart from competitors in the years ahead. Your North Star should be to build an overall competency now so you are ready to take full advantage of GenAI when game-changing opportunities arise. 

As NEA venture advisor Francois Ajenstat, chief product officer at Amplitude and before that at Tableau, puts it: “AI will be an ingredient in every product within a few years. So it’s time to start building this muscle now.”

Testing the waters

Amplitude provides a fine example of this balanced approach. The data analytics company has used machine learning and other AI techniques since it was founded in 2012, and it quickly began working with GenAI when ChatGPT was released in late 2022. In early 2023, the company rolled out a chatbot called Ask Amplitude. Like so many other companies that rushed ChatGPT-like chatbots to market, Ask Amplitude was not a game changer. But rather than start over, the company built on the effort by making sure the team had the necessary skills and infrastructure to apply GenAI more effectively. 

At a recent hackathon to explore ways to use GenAI, engineers came up with a feature called Session Replay AI that lets Amplitude’s customers quickly analyze thousands of videos of their customers interacting with products to spot patterns. Without AI, this would take countless hours to do. “This was a lightbulb moment for how we could make qualitative insights scalable,” Ajenstat says, “It came together in two days, and we launched Session Replay AI a month later. It was incredible.” 

Ajenstat says Amplitude’s progress stems primarily from a lean-forward mindset about GenAI’s long-term impact. “I see way too many people, especially at bigger companies, taking a wait-and-see attitude on GenAI because the technology is still far from perfect,” he says. “Guess what: The first iPhone wasn’t great either, but you didn’t want to be the last in your category to do a mobile app,” he says. “Don’t put your head in the sand. This is happening.”

Harvest the low-hanging fruit

That said, there’s no reason to tear up your product roadmap or quickly retrofit every product with GenAI features. First and foremost, continue to attend to the basic blocking and tackling of creating great products, delivering on promised upgrades, integrations, security fixes and the like. As you do so, find out customers’ appetite for GenAI, and how it could potentially add value. Are they asking for GenAI? Do they trust it? Even if they do, do they have the data sources required to make GenAI effective? Chances are, there’s not as much of a rush as you might think. According to Miku Jha, director of AI/ML and GenAI at Google Cloud, only 15% of companies have a clear idea of what they want to do with GenAI. He estimates that only 5% will come up with ways to add GenAI to projects in a way that leads to significant monetization.

Of course, there are cases where GenAI can quickly make a big difference. Since it was founded in 2016, Sana Labs sold a powerful “learning management system” (LMS) where companies could store data on their people, processes, and other useful information. Historically, these systems were mostly unseen parts of a company’s software infrastructure, used primarily by administrators, HR people, and training experts responsible for developing curricula. 

With the advent of GenAI, Sana gave its LMS a major facelift that has made it more useful to far more people. It developed a proprietary technology that lets it ingest a broader range of information, including traffic from apps like Slack and Google Drive, and an easy-to-use AI assistant that any employee can use to get answers to questions. In other words, more of the customer’s employees have a tool that lets them learn far more, increasing the value of the stored data. The new system can even create some of the curricula and other content itself. We spoke to dozens of users before investing in Sana, and we were amazed to hear people praise the software as they would a beloved consumer app.  

GenAI is often more effective in small doses. Instead of rethinking entire products, many of the most successful implementations have been more restrained. When Google integrated its Gemini LLM into Gmail, for example, it no doubt could have redesigned the app to focus on GenAI capabilities. Rather than see your emails, perhaps you might have been presented with a summary of a conversation and action items. But that might have forced loyal, satisfied users to go five clicks deep to use email the way they’re used to. Instead, Google folded in useful aids like a “Help me write” feature to quickly compose messages. It may sound elementary, but we’ve seen a surprising number of firms release GenAI “upgrades” that actually downgrade the user experience.

Plan for the future

Assembled is another company that has successfully threaded the needle in its use of GenAI. The company has used machine learning and other AI methods since it was founded in 2015 to help users predict how many customer service people they will need to have on hand at any point. Leveraging GenAI, it first introduced a copilot called Assembled Assistant that lifted the product from a resource-management tool for administrators into a service that helps customer reps do their work by providing data on past interactions or even suggesting diplomatic ways to respond to angry end users. 

In 2023, Assembled rolled out a service that solves a problem for customers that was never solvable before: how to guarantee they have enough customer reps to handle the call load in case of a natural disaster or some other unforeseeable event. Rather than race to find more agents, customers can turn on the service, which has been trained to answer common requests, freeing the agents who are on hand to deal with more complicated, disaster-related queries. 

Over time, GenAI could have a far more sweeping impact. It may well wipe out demand for existing product categories and create vast new ones. What if GenAI-powered voice interfaces get so good that we stop spending our days typing on GUIs? Why create that next version of Word, Slack, or other keyboard-centric apps? Why create infrastructure technology that just provides operating system–level functionality, when your competitors are making products (think Sana Labs) that simply provide the information and insights users seek? Rather than build functionality that allows humans to do work more efficiently, the best GenAI apps will provide answers so those humans don’t have to. 

Eyes wide open

We hear a lot of discussion in the entrepreneurial community about the need to tap into large budgets that companies like Bank of America, JP Morgan, and McDonald’s are setting aside to investigate GenAI. For founders, it’s a tempting target, and by all means you should try to win these deals and get your foot in the door at these massive accounts.

But do so with your eyes wide open. Don’t count on the revenue from them to be around for the long haul. Many of these budgets are experimental and may go away in a year or two. The customer may well be kicking the tires on a lot of promising technologies, not just yours, and in the end they might choose to create their own. So approach these wins as beachheads that give you an opportunity to search out champions within business units who will be your actual long-term customers. 

Also, don’t feel the need to jam GenAI into your pitch decks to impress venture capitalists. Yes, we love to hear how companies are using GenAI internally to run more efficiently. But in terms of creating products, GenAI is one of many ingredients that you can use. And it’s fine not to use it if it doesn’t differentiate those products. You should be looking for investment partners that support you because of your vision and your dedication to solving customer problems — not because you’ve checked the GenAI box. Indeed, if you’re optimizing your products to win VC funding, you’ve already lost.

Lastly, don’t forget to have fun. In the hands of great entrepreneurs, product visionaries, and developers, GenAI is indeed a game-changing technology. This is a great time to be thinking about how to make the most of it.