Buy, Build, or Just Fix the Process?

One of the first things I do with any business is figure out which of three buckets a problem falls into. It saves real money, because the expensive mistakes are usually building something you could have bought, or buying something when the real issue was the process.

1. Fix the process

Sometimes no tool is needed. The problem is that three people do the same job three different ways, or nobody knows who's supposed to follow up. Automate that and you just get confusion faster.

Signs it's a process problem:

• People disagree on how something is supposed to work
• Things fall through the cracks between people, not inside a tool
• The same question gets asked over and over because the answer isn't written down anywhere

2. Buy something

If a well-known tool already does the job, buy it. Scheduling, invoicing, email marketing, e-signatures: these are solved problems. Someone else keeps the tool updated, fixes the bugs and ships the new features.

The catch is setup. Most tools are only as good as how they're configured, and that's where a lot of businesses stall. Buying is the easy part.

3. Build something

Build when the job is specific to how you work and nothing off the shelf fits. That doesn't have to mean a big custom app. Usually it's small:

• A simple script that moves information between two tools that don't talk to each other
• An assistant that drafts quotes using your pricing and your wording
• A one-page dashboard that pulls numbers from three places
• A form that sends the right follow-up based on what someone answered

Where AI fits

AI is one more option in the toolbox, not a category of its own. It's great at reading messy input (emails, notes, photos of receipts) and drafting things for a person to check. It's not great at decisions that should belong to you. I use it where it saves real time and leave it out where a simple rule does the job just as well.

A quick way to decide

1. 1Write down the job in one sentence
2. 2Ask whether everyone agrees on how it should work. If not, fix that first.
3. 3Look for a tool that already does it, including the ones you already pay for
4. 4If nothing fits, build the smallest thing that solves it