A future no one is asking for
Why most companies have no real use for AI technology but will invest in it anyway
Welcome to I’m Late to This, a newsletter about things I haven’t stopped thinking about.
If someone sent this your way or you haven’t signed up yet, subscribe below so you never miss an issue. We publish every Sunday morning.
If you agree, disagree, or just want to weigh in on today’s newsletter, reply to this email or hit me up on Twitter @MylesUdland. Feedback is always welcome and highly encouraged.
And now: the future. Sort of.
My favorite Twitter joke format is:
No one:
Absolutely no one:
Not a single soul:
(Something no one was asking for)
To me, this joke captures what we as consumers and workers are most often presented — a new thing literally not a single person was looking for. Enjoy!
In the business world right now, the punchline is that every latest and greatest idea is some version of an AI/AR/ML strategy.
Short for artificial intelligence, augmented reality, and machine learning — we could also throw NLP (natural language processing) and VR (virtual reality) in there, too — these are the innovations that are supposed to disrupt everything we know about work, how we make a living, and how human progress moves from this stage to the next.
Of the business blob’s collective truisms right now, none are as ubiquitous as pushing AI/AR/ML as the no doubt future of all businesses. This idea was even potent enough to carry Andrew Yang’s presidential run for more than two years. Imagine what it can do in the hands of an MBA in the corporate strategy department.
See: it’s right here in the picture.
Every company has, is developing, or wants a strategy to make their business either prepared to address the challenges of the AI/AR/ML world or wants to convert their core functions to be conducted by AI/AR/ML techniques so they thrive in this world. A world, again, that is apparently inevitable.
In the AI/AR/ML future, everything mundane will be automated. The computers running a business’s core functions will approach something like a profit maximizing singularity that accelerates cost-cutting trends, gross margin expansion, winner-take-most market dynamics, and social and corporate inequality.
The rich — be they businesses or citizens — will get richer as they accumulate more money to invest in better technology that improves exponentially and thus accelerates the amount of money available to the best companies, these companies’ remaining workers, but most of all the owners of both capital and these businesses.
If this sounds bad, it is. Elon Musk is worried about it. Though I think the obvious badness of this future makes the entire dystopian outline far less likely than the Musks and Yangs of the world believe.
But so the obvious reason why these acronyms are popular in the business world today is they offer the perfect mix of an assumed-true future that is rendered completely meaningless in the present. If your company spends millions of dollars each quarter to invest in “AI technology” or “AR experiences” there is basically no way to account for whether this investment is good or not. In the echo boom of the Amazon supernova that justifies investments of any quality, nothing does more for departments seeking bigger budgets than a concept that elegantly explains a theoretical future no current executive will ever see that also cannot be effectively P&L’d in the present.[1]
Of course, various pieces of software that touch our lives everyday already fall under the broad umbrella of artificial intelligence and machine learning, like when Gmail offers you sanitized corporate-speak to finish your sentence in an email. The retort to why business leaders love talking about this stuff is that it’s already happening. Therefore, it will happen more.
But the planes on which we discuss the future of these technologies and our actual experience with them are so far apart that the open lane for grift, alarmism, and bad business ideas is so large it often seems like the best area for an entrepreneur or executive to pursue.
Writing on a16z last month, Martin Casado and Matt Bornstein outlined the economics of AI companies. And to what I assume is the chagrin of many readers of this piece, Casado and Bornstein argue that AI businesses are in a different category from the traditional software-or-service dichotomy that predominates today.
Their outline is that AI software businesses don’t have the magic of your pure play SaaS company that creates the money-printing set it and forget it model so coveted by investors right now: sell software one time to a customer who never cancels and you make money forever while your costs stay basically flat. But AI also doesn’t quite have the problems of the traditional one-off services model in which scaling cannot be conducted at anything other than linear speed because every transaction can only happen once.
And so on the one hand, new business models are exciting. On the other hand, gross margins of 50%-60% instead of 60%-80% are less exciting. If software is eating the world, I think many folks believe AI/AR/ML will just make the meal go down easier.
But in Casado and Bornstein’s outline of why AI is good and AI is a challenge, the biggest takeaways are that 1: AI requires a lot of humans (which nukes another one of the most pervasive and laziest tropes in business right now that robots are coming for your job) and 2: AI’s biggest challenge — and opportunity — is that the technology is in fact novel.
In fact, one of Casado and Bornstein’s recommendations for founders is, “Choose problem domains carefully – and often narrowly – to reduce data complexity. Automating human labor is a fundamentally hard thing to do. Many companies are finding that the minimum viable task for AI models is narrower than they expected.”
As an outsider to the AI space, I came away from Casado and Bornstein’s piece feeling quite excited about what these applications and this software could do for the business world at large.
If one of the themes of this newsletter is the uniformity of the business world today, it seems that increasing adoption of artificial intelligence could create a more diverse set of answers to similar problems.
But AI might only get there through the backdoor. Because from the front, it all sort of looks the same right now.
As Casado and Bornstein write:
In the AI world, technical differentiation is harder to achieve. New model architectures are being developed mostly in open, academic settings. Reference implementations (pre-trained models) are available from open-source libraries, and model parameters can be optimized automatically. Data is the core of an AI system, but it’s often owned by customers, in the public domain, or over time becomes a commodity. It also has diminishing value as markets mature and shows relatively weak network effects. In some cases, we’ve even seen diseconomies of scale associated with the data feeding AI businesses. As models become more mature – as argued in “The Empty Promise of Data Moats” – each new edge case becomes more and more costly to address, while delivering value to fewer and fewer relevant customers.
This does not necessarily mean AI products are less defensible than their pure software counterparts. But the moats for AI companies appear to be shallower than many expected. AI may largely be a pass-through, from a defensibility standpoint, to the underlying product and data.
And so if the VC community’s struggle with AI businesses is that they look less like Salesforce and more like McKinsey, this tech and business model fits well in the deployment age it appears we’re headed towards.
But the road between here and there will get rockier before it gets smoother.
Because if AI/AR/ML technologies and businesses appear to have one flaw right now it’s that they are deeply misunderstood.
Making the industry’s primary challenge less related to whether the technology actually works or not, and more about trying to be anything other than the punchline in a joke about who is looking for this and why.
[1] Amazon, of course, invested all their free cash back into the business because the original retail concept was a great idea and had an absolutely massive TAM but required a ton of capital. Investing into a business is good so long as the idea isn’t flawed, a statement so obvious it feels silly to write. But you’d be amazed at what’s out there.