This was a question during a Serotonin monthly all-hands whose answer inspired me to publish as an article. Thank you, @byKatherineRoss!
I want to preface this by saying that I have some serious social concerns about AI, or rather, about the groups of humans creating AI to serve their very human interests. Namely, my concern is that the goals of these human creators are misaligned with us, their users. If it served their interests, these actors would happily use powerful AI to ensnare our attention, siphon our wealth, and warp our perceptions. They are already using relatively weak AI in these ways, often content to serve a false feeling of productivity to keep us as subscribers.
While some of us feel joy using these tools, others feel fear: of getting left behind or losing jobs. The AI companies benefit from your fear. Otherwise, they would market their product as glorified autocomplete rather than harbingers of doom.
My other fear is that LLMs are already changing our psychology. When movies and then television first emerged, critics worried that millions of people seeing the same film would overly standardize their behavior, making society more homogeneous. Arguably, they were correct. With LLMs more deeply pervading our lives than movies and TV - and the LLMs themselves converging, as they have been found to, on standard speech patterns - that homogenizing effect is exponentially stronger. If you or the culture feels stuck, get ready, this is only the beginning. If you want to preserve your creativity, or even your individual identity, you need to come up with a plan, now. These things are getting in our heads.
But this article isn't about my personal concerns about AI (though now you know them). It's about how I am thinking about AI in the context of the company I founded and help lead, Serotonin. The above disclaimers aside, the picture is far rosier.
At Serotonin, we are already seeing demand for new services, such as helping companies implement AI tools to make them more efficient. This is a service we now offer, after testing it within our own organization. Our approach has always been to learn from customers what they want and staff accordingly. We pass information about the evolving shape of demand onto our team so they can continually upskill in response.
What being, let's say, a social media manager, required in 2023 is not the same as in 2026. If a person woke up from a coma after three years and somehow came to work the next day, they would need to study up on how the job had changed before they became an effective teammate once again. I can't predict what that person's day will look like in 2029, but I am betting their job will be there.
I think a lot of technology companies are going to survive and thrive: ones that produce physical infrastructure and products, and some that produce digital products, such as stablecoins and yield protocols. But a lot of technology companies are also going to go bust, specifically ones that built their business around selling subscriptions to software designed to scale at near-zero marginal cost per user, but that people can now build themselves with LLMs cheaply (more on this in point #3).
Also, a lot of the companies whose products are people are going to go bust, starting with the ones where the people's skills are most fungible, like outsourcing and customer support centers. Look at the Infosys share price if you don't believe me. But, if you're not producing physical infrastructure or products, or a small subset of digital products that are going to succeed, instead of sitting on the fence, the best thing to do is make the asymmetric bet in the opposite direction and bet big on specially skilled people.
We are going to employ people (who use AI tools) for a wide variety of functions for a long time. Thanks to these tools, a single person is going to be able to do more work and produce more economic value than ever before. And now that they can be so productive, companies will insist on having the best person, who can now realistically serve them all. If we can attract and retain the very best people who already have the best discernment and pattern matching (see the recent commentaries on the increasing value of taste), continually update their abilities based on market demand, and use the most powerful tools, we will be a mighty organization indeed.
The truth of this is emerging as some companies choose the opposite route, disgorging their staff of highly skilled and not highly skilled (according to my definition of skilled above) employees alike, in favor of becoming software products that aim to scale at 0 marginal cost. The problem with this is that no one wants to subscribe to your software, because they can produce it themselves for less than it costs you to make, or at least, they aren't willing to pay enough that your investors will get paid back a 10x multiple. Betting against highly skilled people in favor of software you just built with an LLM spells doom.
Remember to say the quiet part out loud: on the OpenAI SuperBowl ad for Codex, their coding platform, they wrote: "You can just build things." I would finish that sentence differently. You can just build things, but so can everybody else. Think about how you are different, and that answer may well be people. If it is, don't just accept it; optimize around it. It's the asymmetric advantage you were looking for, hiding in plain sight.
I'll share the first tenet with you: do not raise VC that expects a 10x return. Don't raise any at all. Services companies should be profitable from day 1, or very lightly bootstrapped. Hire into actual booked revenue, not speculatively. Being the best matters. Margins matter.
If you are operating in a structure that raised from VCs expecting a 10x return, but are realizing you actually should be a services business, that is going to be a problem. Give them their money back if you can, and start a proper services business, with the right expectations.
One positive thing about LLMs is that they are going to bring about the end of friction and the end of rent seeking, so people will actually need to work for their supper. Welcome to services. The water is warm if you're willing to swim.
Written by @amandacassatt
Reactions and replies to this article.