AI must be more than a checkbox
Otherwise, we're all wasting our time.
We’ve all heard plenty of stories about the proverbial CEO banging her fist on the table and insisting, “We need an AI strategy!” As an AI startup, this is generally good for us — market demand for AI solutions is a tailwind that leads more potential customers to engage with us.
It also leads to poor incentives — for both buyers and sellers. For startups who are desperate to sell in order to scale their business, it can lead to wasted time and low-quality revenue that’s likely to churn. For customers who’re looking to fill a checkbox with an AI feature, it can lead to poor results, wasted money, and embarrassing customer interactions.
In both cases, AI needs to be treated as more than a checkbox.
Startups get lost in the fray
A noisy market is dangerous for a startup for many reasons, but one of the most insidious ones is that it can lead you to spend time with prospects who aren’t serious customers — folks who are just interested in an AI product for superficial reasons. Time is of the essence at a startup, so spending time with leads who aren’t serious about buying your product can be incredibly painful.
For technical founders (like us!), learning to let go of low-propensity leads is difficult in the best of circumstances (see Founding Sales to learn more). When those low-propensity leads are coming from brand name companies that you’re excited about working with, it’s all that much harder. Unfortunately, when every enterprise in the world needs to have an AI strategy, there are many low-propensity leads casting wide nets across many AI products. Anecdotally, we’ve spoken to folks whose goal is to understand how AI can improve every aspect of their business — and yet, they’re unable to articulate which areas are the highest priorities.
This is general sales advice, but especially important in today’s market: You need to think carefully about who fits into your target market and how you can quickly weed out anyone who’s not serious.
You might think you’ve moved past this issue if you’ve begun to close deals with customers. That’s certainly a good step, but it’s unfortunately not enough in this market. We’ve spoken with a number of enterprises who have paid for solutions only to realize they had very little reason to have that product. In all likelihood, they’ll churn after their contract is up — either because they don’t have that problem or to find a better-suited solution for their team.
To again share generic advice that is amplified in this market, you need to think very carefully about what kind of value you’re creating for your customers. Simply having them on board only for them to realize there’s nothing they’re getting from your product means your churn rate is going to be miserable in a year. This type of churn is particularly painful in a market where there will be plenty of fast-followers who’ll smell blood in the water.
Why are we buying this again?
Things might look safer from the other side of the table — after all, the worst case scenario is that you buy a product and get rid of it in a year. What could be so bad about that? While the jeopardy is certainly lower than that of a startup trying to get off the ground, adopting AI products without a clear understanding of the value proposition can still be painful.
The most obvious risk is that you’re spending money on something without a clear understanding of why you’re buying it, especially in a market where budgets are still tight. This may sound ridiculous, but we’ve seen plenty of examples of AI products get purchased without a clear understanding of the ROI. Ostensibly, this is has been to check the AI box.
More importantly, buying products without a clear understanding of the value is a waste of your time. You’re evaluating, procuring, and deploying something that you’re not actually deriving value from. In the worst case scenario, this leads to bad customer interactions: We’ve seen a public example of a large company purchase a product, make it publicly available to its customers, and then ignore the results (which quickly went off the rails). It might have been an initial win for both sides to buy/sell this solution, but the current state of affairs reflects poorly on everyone.
None of this means that AI solutions are bad or that customers shouldn’t be engaging with them. In our experience, this boils down to is a lack of understanding of what you’re trying to accomplish with AI. We often get asked questions like, “Why would I use GitHub Copilot and RunLLM both?” To us, this suggests an AI-checkboxing mentality— yes, both are developer-oriented, AI-powered tools, but they accomplish very different things. Copilot is for code generation in the IDE, and RunLLM answers developer questions specific to a product. To us, this is like asking why you would buy both a CRM and a support ticketing tool, since both track the interaction with customers.
Everyone in this market needs to be thinking carefully about how to quantify the ROI of AI products. In turn, quantifying ROI requires understanding how good individual AI products are and how well a product fits your use cases. This burden is on both sides of the table. Most critically, product builders need to move past vibes-based evals or even raw accuracy metrics and start quantifying the value proposition of their product. On the other hand, the most successful sales conversations we’ve had have been with teams who show up with a clear evaluation framework.
Without rigor on both sides of the equation, AI startups simply won’t succeed.