I love it when a plan comes together
I'm a big team person. I have played and coached sports all my life. I have built teams both small and large in multiple settings. I regularly sign up to be the team parent for my daughter's sports (much to her chagrin, sorry not sorry). Maybe this is why I have long been fascinated by team dynamics at work.
Last year, I looked at startup performance data and concluded that team size matters. In particular, startups with teams of about 5 people per manager perform best. Shortly thereafter, I wondered whether the rise of AI agents means that organizations won't need as many managers.
It's a year later now, so it's time for some new data. The t;dr: Team composition is changing, particularly inside startups, and the trends would inform my decisions if I were starting a tech career right now.
Startup teams are changing
I recently dove into team composition data from 68 private AI companies that I consider to be best in breed. 34 companies are early-stage and 34 are already scaling. The data is a combo of what is available publicly and stuff I collected when I looked at investing in these companies earlier this year.
I want to highlight four major insights.
Insight #1: AI companies that are scaling have sizable span of control
The early-stage companies are mostly small with few managers. Most of them are seed or pre-seed with limited revenue. These teams will evolve quickly and their composition will change, but their current state gives a view into what modern AI projects look like when they first begin.
As AI companies scale, they are adding ICs at a faster rate than managers. The later-stage AI orgs are flatter, with 8.9 employees per manager.
Insight #2: Early-stage companies are mostly devs
Early-stage startups have always been dev-heavy, so this is no surprise. The typical pattern is a couple of technical founders build something and then whoever is CEO ends up building early GTM and ops. I lived this pattern.
However, teams are taking longer than a few years ago to add non-dev roles, especially product, design, and ops. Since these are AI-first companies, in their own operations as well as their products, are they using AI to delay hiring non-engineer specialists? I suspect the answer is yes.
Insight #3: The role mix is changing as companies scale
Later-stage companies are adding as many researchers as designers and PMs. This is not just happening at model companies. It is also true for companies doing applied AI.
Ops hiring is leaner than in traditional SaaS. Five years ago, typical companies at scaling stage would have been about 15% headcount across HR, finance, legal, and ops. Now it's closer to 8%. On the other hand, it is common to have 1 FDE (Forward Deployed Engineer) for every 5 product engineers. This ratio was higher than I was expecting, and there is some growth point beyond which this will not scale linearly.
Interestingly, the patterns below hold almost regardless of the company's industry e.g. the ratios at dev tools companies don't look that different from vertical SaaS.
Insight #4: Companies are ramping up revenue / employee way faster than traditional SaaS
65% of the companies have been able to cross $250K ARR per employee in their first couple of years of commercialization. This is astounding.
The big moneymakers like Anysphere and Replit are getting a lot of chatter because they're easily clearing $1M ARR / employee. But it's not just a few outliers. The headline here is that most of these companies are significantly ahead in revenue / employee compared to traditional SaaS companies.
Yes, this is because they're selling a lot. But it's also because the teams are smaller and flatter. There's just less cost (at least on the headcount side; I suspect many are spending more than the headcount savings on compute).
Here to stay or flash in the pan?
This is data from 68 companies still in their first few years of existence, and startup teams have never been stable isotopes. Enterprises may follow a different course. Some roles will go away and new ones will emerge.
Still, there's insight in the patterns. The First, Second, and Third Industrial Revolutions created more specialist roles in most industries. We're living through the Fourth Industrial Revolution now with the rise of AI. For the first time in centuries, generalists who can use the tools rule.
What's your take?
Kieran
If you liked this story, why not subscribe to nerd processor and get the back issues? Also, why not learn to tell data stories of your own?
My latest data stories | Build like a founder | nerdprocessor.com