The Double-Edged Sword of AI Hype
10 strategies to navigate the trap of unfocused customer demand
Today, it seems like every enterprise in the world is claiming to be adopting AI in a big way, with many attempting to rebrand themselves as AI native.
Earlier this week, Apple, who has famously been an AI skeptic for years now, finally threw their hat into the ring with the announcement of their generative AI offering, Apple Intelligence (could there be anything more Apple than trying to rebrand the acronym AI itself?). If Apple is finally caving, safe to say that all enterprises are.
It’s clear that business leaders are obsessed with being affiliated with the next big thing here, but the details regarding where and how companies are actually deploying AI are a bit more fuzzy. If you read any earnings transcript, press release, or investor letter, it’s very challenging to find specifics around which AI use cases are finding the most success in organizations.
So, where is it being used? In short, it’s everywhere and nowhere today, which is increasingly becoming a problem for AI startups.
The Hammer Looking for a Nail
Most great enterprise businesses are built by solving a particular pain point in modern business. Can’t keep track of your customers? Enter CRMs. Need help managing payroll for thousands of employees? Enter HR software platforms. Pandemic forces remote work? Enter video call software.
As a nail pokes out, you search for (and pay for) a hammer to fix it.
Modern / Generative AI has a different story. Market fervor for LLMs emerged with the adoption of ChatGPT, which is an inherently consumer-facing application. The tsunami of consumer adoption and excitement led to every executive in the world scrambling to define their business’ “AI strategy”. In a scramble to not be left behind, many businesses have greenlit experimental AI budgets and encouraged their teams to go out and “adopt AI”.
Unfortunately, nobody knows what to use it for.
Today, we’re in a remarkable moment in time where enterprises want to pay for AI, but have no idea what to use it for. In short, AI offerings are a hammer looking for a nail.
The slippery slope of test and trial
Today, many enterprises are attempting to navigate the AI adoption cycle by allocating “test and trial” budgets, in which various departments can onboard multiple providers to play around with their offerings, often with no perspective on where to actually deploy them.
In some scenarios, these organizations don’t even have any budget set aside to actually commit to enterprise contracts on the other side of a POC.
In an effort to “stay on top of AI”, enterprises are willing to experiment with anything, resulting in a tidal wave of pilots and POC’s across the entire AI stack. Unfortunately, the vast majority of these POC’s and pilots will never convert into enterprise contracts, as there is no internal alignment on what a conversion would even look like.
This has become a gigantic distraction for AI founders today. These enterprises often demand a lot of time from founders, ask to be spoon fed use cases, and are extremely hesitant to hit the green light on contracts. These interactions are inherently extractive, as enterprises leverage these pilots and POC’s as a way to learn about Generative AI offerings, without ever paying for them.
In other words, they take up very valuable founder bandwidth without ever giving anything back in return.
For AI startups in which resources are finite and the opportunity cost of each hour is very high, these potential customers can become huge distractions, and can ultimately strangle promising businesses.
As an investor, signing a POC or pilot up with a Fortune 500 customer used to be an exciting signal of traction. Today I worry that it’s almost just as likely to be a noose around a startup’s neck.
How startups are navigating this period of uncertainty
While many, many startups and founders are struggling with this, we’ve had the privilege of working with a few who have crossed this adoption chasm and are now establishing themselves as the name-brand enterprise-ready providers in their respective categories (e.g., Cohere, Covariant, Hebbia, and Twelve Labs, to name a few).
Some of the takeaways that we’ve learned along the way include:
Focus on a tight offering in the early innings. For a startup, it’s virtually impossible to be everything to everyone. Horizontal platforms are hard to sell, so finding a “killer app” use case / wedge to get your foot in the door leaves room for expansion later. Less is more!
Leverage a tried and true Enterprise Framework (e.g., BANT). These sounds like oversimplifications, but there’s a reason they work so well. Budget and Authority tend to be obvious, but identifying real pain points and legitimate use cases as early as possible tends to be overlooked in AI (be specific!)
Spend the time and energy to be a bit more consultative. Buyers want to adopt AI, and we’ve found them to be very open minded to suggestions. If you have hypotheses about where AI can be deployed in their org (or have seen it successfully implemented elsewhere), don’t be afraid to gently guide them
Push hard for paid pilots / POC’s over unpaid ones, even if the dollars are small. The revenue from these will always be negligible, but forcing a prospect to pay for what you’re providing forces them to actually engage with your offering and take you seriously. This has the added benefit of weeding out orgs that are more skeptical than others
Be maniacal about feedback loops. Customer feedback is gold, especially for a startup. Large enterprises tend to move slowly and be hesitant to provide formal feedback. If you’re chasing a big customer, it usually helps to develop a strong bond with one internal champion who is willing to provide frequent and honest communication (Slack / WhatsApp > Zoom Calls!)
Always define what a successful pilot or POC conversion would look like beforehand. If you can agree to have a customer get specific about KPI’s or metrics needed to trigger a conversion, even better (in other words, get the buyer to agree to the N and T of the BANT framework ahead of time)
Leverage specific case studies as much as possible. The more you can explain what you’ve done for others / where you’ve found success, the more likely enterprises are to adopt your offering. Bonus points if you can point to work you’ve done for a competitor!
For very, very large contracts, experiment with Forward Deployed Engineers. Services work can be distracting, but for transformational customers (think high seven or even eight-figure deals), it can be worth investing the time to build your offering into a customer’s stack, which improves conversion and creates stickiness
Don’t be afraid to fire customers. If a customer is taking up cycles but isn’t providing valuable feedback or actionable guidance on a path to conversion, it’s totally okay to sunset a pilot on your end (this might even incentivize a customer to actually take action)
Revisit assumptions about your ICP and product offering at least once a month. Nobody gets everything right on the first try, but the most successful startups are the ones that iterate the fastest and learn from incremental data. Frequent communication, openness to experimentation, and a willingness to go back to the drawing board all allow startups to outhustle incumbents
This is really starting to become an area of focus for us, so as always, I’d love to hear how companies are navigating the tar sands of test and trial budgets.
All additional tips and tricks are welcome!