Do Things That Don't Scale (With Documents)

Paul Graham's advice to "do things that don't scale" is one of the most useful ideas in startup wisdom. But most people only apply it to customer acquisition. Go door to door. Send personal emails. Do things that a big company would never do.

The same principle applies to documents, and almost nobody talks about it.

When you're starting a company, there's enormous pressure to "professionalize" your document handling from day one. Get a CRM. Set up automated contracts. Build a proper invoicing system. Use templates for everything.

This is a mistake.

The manual phase

In the early days, you should be creating every contract by hand. Not from a template. By hand. You should be personally emailing every invoice. You should be reading every NDA before sending it. You should be manually tracking who's signed what in a spreadsheet.

This sounds inefficient. It is inefficient. That's the point.

When you manually create a contract, you're forced to think about what should actually be in it. When you personally send an invoice, you learn how customers react to your pricing. When you read every NDA, you discover which clauses cause friction and which ones nobody cares about.

Automation hides this information. If you set up automated contracts on day one, you'll automate whatever you copied from a template. You'll never learn which terms actually matter for your specific business.

What manual document handling teaches you

A founder who personally handles the first 50 contracts learns things that are impossible to learn any other way:

Which clauses cause customers to push back. This tells you where your terms are out of step with market expectations.

How long each step actually takes. You can't optimize a process you haven't done yourself. Timing each step manually gives you the data to know what's worth automating.

Where errors creep in. When you manually copy data between documents, you'll make mistakes. Those mistakes show you exactly which fields need to be auto-populated later.

What customers actually read. You'll notice that some sections generate questions and others are ignored. This tells you what to emphasize and what to simplify.

The graduation point

There's a moment — different for every company — when manual document handling starts to break. You'll feel it. Contracts are falling through the cracks. Invoices are going out late. You can't remember which version of the NDA you sent to which client.

This breaking point is good. It means you have enough volume that automation is justified. And because you've been doing things manually, you know exactly what to automate and how.

The company that automated from day one doesn't have this knowledge. They automated a generic process. You're automating your process — the one you built through experience.

Common objections

"But manual processes look unprofessional." Maybe. But customers care about the content of your documents, not the tool you used to create them. A personally written contract that addresses the client's specific concerns beats a slick automated one that feels generic.

"We don't have time for manual processes." You don't have time to set up automation properly either. The choice isn't between manual and automated — it's between quick-and-manual and quick-and-badly-automated. At least manual gives you learning.

"What about compliance?" Fair point. Some document processes have legal requirements that mandate certain controls. But even here, you can often meet the requirements with manual checklists before investing in automated compliance systems.

The automation transition

When you do automate — and you should, eventually — the transition is smoother because you understand the process deeply. You can write better specifications for your tools. You can evaluate document software based on whether it handles your actual workflow, not some generic demo.

The best document systems at great companies almost always started as messy manual processes. The founders did things by hand until they understood them, then built or bought tools that codified that understanding.

The worst document systems are the ones set up by someone who read a "best practices" blog post and implemented everything on day one. They look professional but don't match how the company actually works.

Do things that don't scale. Including your documents. Especially your documents.