4 min read

How Sesame Street Helps Sell Machine Learning and Learnings for All Startups

One of the hardest parts of bringing any new technology to market is customer acquisition. Inherently, people are reluctant to jump into an unknown technology so having a clear and articulate way to convert interest to understanding in your funnel and a compelling sales pitch is critical to converting your leads into paying customers. Machine Learning (ML) companies face this challenge head on every day. When we brought together a group of AI and ML experts from our Core Community to discuss their businesses overall, how they effectively articulate their value proposition to attract customers turned into a lively conversation with an unlikely set of analogies and metaphors.

In addition to the regular challenges of selling a B2B product (identifying a segment, targeting prospects, educating them, showing measurable or even potential ROI, etc.) the ‘black box’ nature of machine learning makes it tough for people to buy into. Alan Ringvald, CEO of Relativity6, offered some advice on how he was able to better explain the technology.

During our first few months out in the field pitching Relativity6’s technology to potential customers, whenever the inevitable “so how does this actually work?” question would rear it’s head, my CTO and co-founder Abraham [Rodriguez] would roll up his sleeves and start explaining how our convolutional neural network worked or how we built a unique approach to feature learning based on space and time.

The room of non-technical people would nod their heads in unison signaling that they understood what he was saying, but we didn’t get one buyer in those first few months. Something was definitely wrong.

One fateful day the lightbulb turned on. Abraham and his young son were watching an episode of Sesame Street called “Hungry Games: Catching Fur” which is a parody of the hit series Hunger Games. In this episode, the cookie monster and crew were forced to predict the next item in a fairly straightforward pattern: Banana, Apple, Banana, Apple, Banana, etc.

Pretty simple, right?

But eventually things got trickier. After finding the obvious pattern, the cookie monster was faced with predicting the next item in a series of items featuring: cheese, cookie, sandwich, pizza, cracker….

Not so obvious.

[Spoiler Alert] Turns out, the answer wasn’t a certain item of food, but instead the shape of the food. The cookie monster predicted a round item would be next, and he was right!

I don’t think Sesame Street or it’s intended audience was aware, but this was actually a very elegant way to explain non-obvious pattern recognition which was similar to how our technology spotted similarities between customers based on non-obvious contextual features.

We showed this video to a non-technical customer and they instantly understood what we were talking about. This lead us to start coming up with machine learning analogies related to all kinds of mundane things like shopping for fruit to blowing bubbles in the air.

Real-world ML analogies are now a standard part of our pitch, which has lead to way less confusing pitch meetings and some significant sales.

As our session continued we opened up the black box further and inevitably came to the human vs. machine discussion.

Matt Osman, CEO of Legit Patents offers another perspective on how he humanizes the technology to make it easier for buyers to digest.

So far, we’ve found the message of human intelligence augmentation far more powerful than the message of replacement.

[Identifying the persona is key here.] We just happen to think that in the patent space the human that should be augmented is the inventor not the attorney.

I think when selling software that performs functions formerly reserved for human cognition you need to humanize the technology but only to an extent. The pitch has to be human-enough to breed trust but not emphasizing superiority in the manner of a ‘robot overlord’.

A technique that works with a management level buyer is suggesting that the software is doing the job of a junior employee — one of the reasons that digital assistants (bots used to schedule meetings and automate basic tasks) are such a natural initial application of AI (at least from a marketing perspective). For example, we’ve found the best way to explain our first product is as ‘a patent agent that lives in your browser’.

People are comfortable with a digital servant but not a digital master.

One thing we could all agree on is that for now AI/ML is a complex enough new technology domain that it will be difficult for most buyers to understand. The black box nature of data trained models or neural nets and deep learning may present unique challenges to explain, but this is not unique to early markets. Odd as the analogies that surfaced may be to explain AI/ML, the idea of using analogies and metaphors to simplify and help a broader audience relate and understand is a powerful tool for startups to consider.

Thanks to the many great Boston ML companies that have contributed to our discussions and for their fun and enlightening work to explain their technology.

What’s the best way you’ve found to boil complex ideas down to a straightforward explanation? Share your analogies and stories with us in the comments below.