Now is a good time to start a service business

· Zach Ocean

Now is a good time to start a service business. Maybe you should try it!

A service business is one where your company sells its work output directly instead of selling a product. Dentists, plumbers, design agencies, pool cleaner, consultants: these are all service businesses.

The opposite of a service business is a product business, where your company develops a product and sells it to customers. Your dentist pays for a practice-management software product and a bunch of teeth cleaning products like drills, mirrors, and suctions, but you pay for the service of a biannual dental checkup.

Why you should not start a service business

Conventional wisdom is that technologists should avoid service businesses if they want to build valuable companies. It’s easy to see why. All of the big name-brand technology companies are product companies. All of the recent Silicon Valley wins are product companies. If you rank the largest companies in the world by market capitalization, you have to go all the way down to number 71 before you find a services business (Accenture; I’m not counting big banks like JP Morgan Chase, though they do make a bunch of money from services).

Part of what makes a company valuable is great margins; late-stage product companies of all types can develop great margins via economies of scale, and software products in particular have beautiful margins out of the box because software has such low marginal cost.

So conventional wisdom is that if you’re a technologist starting a business, you should start a software product business. The approach is basically:

  1. Talk to users to figure out what product build
  2. Build
  3. Sell the product to customers
  4. As much as possible, perform steps 1 - 3 concurrently to reduce risk and maximize feedback.

In the best of times, this algorithm has a high failure rate. Most products people build are useless, so barely any customers buy them. It’s so hard to build products that aren’t useless that Y Combinator’s motto is “Make something people want”, which is a more gentle way of saying “make something that’s not useless.” Even for very successful software companies, it often takes years until things finally start working.

And for software product companies, it is no longer the best of times. The main thing that’s changed is the amount of competition. Parker Conrad has a great talk where he says that back in 2010, you could start a SaaS company in basically any narrow vertical and build a pretty successful company. Nowadays every narrow vertical has multiple successful competitors so the approach doesn’t work anymore.

Maybe you should start a compound startup

Parker Conrad says that the solution to saturated single-vertical markets is to build a “compound startup”, a company that builds multiple complementary products in parallel.

This has obviously worked well for Rippling, which is worth $14 billion. But a compound startup is expensive and operationally difficult: you have to do O(N) as much work at the beginning, where N is the number of products you’ll offer out of the gate. The first two years of Rippling were Parker Conrad and 40 engineers, mostly pre-launch. Not a lean operation!

But seriously, think about the service business

Service businesses are a solution to the software product saturation problem that works even if you can’t raise $10M out of the gate from your name and idea. There is no R&D phase: assuming you can do the thing that you’re selling, you can start collecting revenue as soon as you’ve found the first willing customer.

How do you find that first customer? Talk to the people you would have recruited as your first users if you were starting a product company. Instead of convincing them to use your product to solve their problem, just solve their problem yourself. Charge more than you would’ve if you’d sold them a product!

How to grow your service business

Now you are running a service business, but you’re also selling your time for money. Since time’s the one thing you can’t make more of, you’ve got just a few ways to make more money:

  • Charge more (sell your time for more money)
  • Hire (sell other people’s time for money)
  • Productize (turn a subset of your service into a product, decoupling money and time)
  • Automate (sell less of your time for the same money)

The classic ways to grow a service business are to charge more and hire; plenty of people have talked about those elsewhere.

Productize means that you boil your service down so that it looks and feels like a product. If you squint, it looks like you started a product company from the outset and just got the engine turning by doing things that don’t scale. Zenefits, Parker Conrad’s first really successful company, got their first users by signing new companies up for benefits by hand; they collected everyone’s data through a pretty website, which made the service look like a product.

And now for the bit about AI

The most interesting reason why now is a good time to start a service relative to other times is automation enabled by AI. Progress in AI has been impressive over the past few years, but it has ultimately enabled only rudimentary automations. LLMs can do most digital services tasks poorly, and very few things well. Research labs like OpenAI and Anthropic are focused on leveling up the fundamental capabilities; well-funded product companies are building on top of the labs’ core LLM offerings to build autonomous workers like Cognition’s AI software engineer.

The most likely outcome in the near term is semi-autonomy: AI workers that can do work with some supervision by humans to ensure the quality bar stays high. If AI gets better at a similar pace but doesn’t reach AGI escape velocity, semi-autonomy is what we’ll get.

Semi-autonomy is a great fit for a service business: you can approach product margins while still keeping the end-to-end-solution characteristics of the service business model.

In conclusion

Service businesses have always been the easiest to start because there’s no R&D phase. But they’ve historically been tough to grow because the main ways to grow were raising prices, hiring, or pruning your service into a product. If AI keeps getting better (a guarantee) but doesn’t quite get to superhuman abilities (big question mark), there will be lots of automation opportunities that let you grow your service business beyond the traditional constraints.

Hacker News discussion