Many have argued that we’ve stagnated in the world of atoms. This is true in the world of bits, too.
We’ve sold software to businesses for decades under the premise that it will improve their productivity. Do more with less, we said. But after capturing the initial benefits of digitization, we’ve achieved very little that’s truly transformative. Every business app is still a CRUD interface with a custom database. We’ve improved the design and made it easier to update and access. But ask yourself: what has fundamentally changed in an ERP or your favorite category over the last two decades? We promised Jarvis but delivered JavaScript instead.
Meanwhile, businesses spend billions annually on software, nearing 10-20% of total revenue. While budgets keep inflating, the ROI has not improved commensurately, leading to a frustrating stagflation for customers.
What would a non-linear jump from this status quo look like?
How about software that answers your phones, prospects, recruits, or closes your books? How about software that passes the Turing test and is not used by humans at all? You could take any role that looks like structured input <> human reasoning <> structured output and create software that nearly automates it.
We could have businesses that operate with far fewer people or, with the same number of people, produce vastly more. There are enormous implications for society as this unfolds over the next decade that deserve separate attention.
Impact on startups/big tech
While software cost inflation has been good for some tech giants, arguably it’s bad for tech in the long term. There are only so many ways you can sell digital transformations. We’ve flogged that S-curve to death.
It’s time to add more value or expand into other domains.
First, on adding more value. Think about something as simple as answering phones for a small business. We couldn’t solve that problem so we sold software to manage your phone, added remote contact centers, sold more software to make them more efficient, IVR for some basic functions, and pushed folks into async channels like text/email. Basically, we did everything but solve the problem.
Secondly, on expanding into other domains. For all the trillions in enterprise software market cap, the TAM is stuck at around 10% of all business spend. The rest is largely contractors, external services and internal payroll. If software needs to break out of the IT spend category, it needs to eat into outsourced services or internal payroll. This year, we are starting to see fully automated digital workers for SDRs, copywriting and recruiting. If capabilities continue to increase and steadily expand into other roles (see graphic below for other ideas), in a decade enterprise software TAM could grow to 20-30% of business spend which is hundreds of billions of new market cap.
Startups have an advantage here and will likely be a big beneficiary: if you’re augmenting or replacing the work of a person, especially outsourced work, there’s no existing software incumbent with a distribution advantage who can copy your tech before you get their distribution.
Business model: seats to services
There’s another interesting nuance in this brave new world. Let’s define SaaS as improving productivity by 20%, co-pilot as boosting that to 40-60% and autopilot as doing 100% of a job.
If you sell SDR SaaS, you’re probably stuck at <$1000/user/year. If you are a SDR co-pilot, you might command $10,000/user/year. If software replaces a SDR on autopilot, you can charge close to what the business pays for SDRs, ~$50,000/yr. That is a 10-50x jump in market size because it’s tied to value for the first time. This opens up a lot more billion dollar opportunities for startups that were previously considered small markets because of per seat pricing.
Who wins?
It’s pretty clear that businesses win dramatically as they gain access to co-pilots and autopilots.
If you’re a founder building a co-pilot or autopilot, the key question you should ask is: where are the durable moats? The answer will vary by category. Sometimes it’s in the app layer because of some unique data advantages, sometimes the underlying infra is hard and has scale/quality moats. Often it’s not obvious, but the first one to race out and bloody their nose on the wall will figure it out.
This is the most exciting era of enterprise software yet, with the possibility of moving from SaaS stagflation to software acceleration (s/acc?).
