The Hidden Pitfall of AI-Powered MVPs

Focusing on Technology Over Viability

With nearly $50 billion invested in AI startups in 2023, the race to market is intense (Crunchbase News). For veterans of the boom and bust cycles in Silicon Valley, we have seen this story before.

Not hard to predict that the streets of Silicon Valley are going to be littered by AI startups that will stumble.

When I get the call from a hopeful, funded founder who wants advice, I hear the same theme - really smart, very dedicated people in love with a technology solution. That love leads them to prioritize technological sophistication over creating viable products that address real needs.

As a lover of startups and a veteran of multiple successful early stage companies, here is what I say and how I guide these CEOs.

“V” Still Means Viable, not Version

I have written about this topic before (here) and my co-authors and I cover the topic in our upcoming book Sail to Scale. (Sail to Scale will be available on August 22. Sign up for our email list at https://www.sailtoscale.com)

But the lesson needs to be retold.

In the race to bring a product to market, we conflate ‘Viable’ with ‘Version.’

Where I get the call is when an investor or a founder needs an external advisor to help a founding team distinguish between ‘Version’ and ‘Viable.’ The symptoms are the same - the technical founder has fallen in love with some advanced AI capabilities that they built and decided to call a product. But those AI capabilities do not address enough minimums to be a viable product. Bolting on AI doesn’t guarantee that you aren’t building a technology in search of a problem to solve. It can be a tough love conversation.

As an advisor, investor and a board member, I look for CEOs that are willing to examine, and re-examine, their hypotheses behind their plans for an MVP.  Like my co-author, Maria Fernandez Guajardo says in Sail to Scale,

You can only really know how your users will react to your product when you put it in front of them, and they use it in the context it’s intended to be used in.

To ensure the viability of your new AI-powered idea, validate your hypotheses for your MVP against the following 5 minimum thresholds.

Ensuring Viability of AI MVPs

Step 1. Minimum Pain

It’s often repeated and yet rarely internalized. It’s hard to get customers to pay for cool technology – not impossible, but not something I’d bet my money on … again. Yes, I speak from experience.

In 2013, I helped found and launch Viblio. We had a ridiculously amazing AI solution that could “understand” video content, well before today’s rash of video AI solutions. We loved our technology. And our target customers, consumers with out of control home video files, thought it was really cool also. They could use Viblio to automatically search through their home video collection for videos of past kids’ birthday parties, or soccer games, or create highlight reels of their beach vacations or daring ski jumps. (And yes, some users used it to find their more R-rated videos.) It was a nice product. And it used AI. But we weren’t solving a pain point - just offering something fun to play with. Even the coolest AI capabilities don’t create viable products.

Today, as an advisor, I talk with a number of founders, and innovation teams within larger companies, who are falling into the same trap - their love for their newest idea for applying AI overtakes honest assessment of whether there is a critical customer need that this new idea solves for.  There has to be a minimum amount of pain your new product is relieving. Take time to invest in user research and analysis, not to see if people like your idea, but to see if the problem you think they have is generating a minimum level of pain for them that they want solved in some way. Prioritize solving real world problems over showcasing AI capabilities.

Step 2. Minimum Number of Customers

Because of my former role as Chief Corporate Strategy Officer at UserTesting, I get to advise various startups entering the user experience space. One such startup offered AI-based prototype testing only for existing mobile apps. Now companies building mobile apps definitely need to test their prototypes.  There was sufficient minimum pain and budget to relieve it. But only large enterprise companies regularly design new prototypes for their existing mobile apps. The feature limitation of the startup’s MVP limited their target audience to large mobile app businesses. And while there is no shortage of large mobile app businesses, there are very few that rush to work with brand new startups in an already crowded market. There just weren’t enough initial customers to keep the startup afloat, AI or not. Minimum pain is not enough for viability.  That minimum pain has to be experienced by enough customers in your serviceable obtainable market. (Here’s a good LinkedIn article that breaks down TAM, SAM and SOM if you need it.) Do your market research to ensure your SOM is large enough to sustain your new business through its launch phase.

Step 3. Minimum AI

You have spent the time validating that there are enough customers out there with the same deep pain point and budget to spend to relieve their pain. Your next step is to truly understand the minimum set of features AI capabilities you need to build the pain relieving AI-solution.

We have confused the message from Eric Reis’ book The Lean Startup. Let go of the perceived pressure to release a product version fast. Replace this with an urgency to release the minimum set of features, necessary to actually relieve your customer’s pain. If you are building an AI solution, this may mean your MVP roadmap is a lot longer than you think. Ask IBM. In an effort to make IBM Watson a revenue generating investment, IBM rushed to launch Watson for Oncology aimed to revolutionize cancer treatment with AI.  And by “rushed”, I mean “nearly six years of painstaking work by data engineers and doctors to train Watson in just seven types of cancer, and keep the system updated with the latest knowledge”.

Watson for Oncology is an extreme example of a product whose minimum technical viability was more ambitious than most mature enterprise products. IBM shut down Watson for Oncology in 2020.

Before launching a version of your product, make sure you have completed a new round of user research, this time focused on what minimum capabilities your users need for them to use your product in the context of their workflow.  For any technology, but particularly for AI-based products, there will be minimum capabilities that go beyond your core technology in order to make a product viable within your user’s workflow.

Step 4. Minimum Feedback:

I’m often told by startups that they have been collecting tons of feedback from their early customers.  They are looking at their website analytics and product analytics dashboards and examining how their customers are using their product. But dashboards and data never moved a creator to reconsider their creation.

But dashboards and data never moved a creator to reconsider their creation.

At UserTesting, our most successful product implementations were the ones where our customer success team practically lived with a customer for the first few months of their adoption (sometimes virtually using our own product).

When your customer success team hears first hand feedback from your customers, you won’t be able to sweep it under the rug like an analytics dashboard. If your success team isn’t engaged with your customers, your customers aren’t engaged in giving you feedback.

Step 5. Minimum Planning for Scale

Launching your MVP can’t be all about today. While a startup shouldn’t invest in scaling at the start of its journey, it definitely needs to consider its minimum plan for architecting for scale. AI solutions often require significant computational resources. If your AI MVP is not designed considering future scale without compromising performance, you may later find yourself with a level of technical debt and scaling issues that is too expensive to overcome.

Conclusion

In my journey guiding startups integrating AI into the MVP process, I've seen firsthand how a laser focus on product viability over quick product versions spells the difference between failure and success. Which “V’ are you focused on?

For more actionable advice, check out my book website Sail to Scale book at https://www.sailtoscale.com.