I get a lot of calls that start the same way: "We want to add AI to our workflow." Great. I love that conversation. But nine times out of ten, the real issue isn't a lack of AI - it's a process that was broken long before anyone mentioned machine learning.
Here's what I mean.
The copy-paste pipeline
A company reaches out because their order handling is slow. They want an AI agent to speed things up. So I sit down with the team and look at the actual workflow. Turns out, orders arrive by email, get manually copied into a spreadsheet, then copied again into the ERP system. Three people touch the same data. No one trusts the numbers.
That's not an AI problem. That's a plumbing problem. And no amount of intelligence - artificial or otherwise - will fix bad plumbing.
Automate the right thing
The temptation is to slap AI on top of whatever exists today. But if the underlying process is messy, all you get is a faster mess. AI amplifies what's already there. If your data is inconsistent, AI will be inconsistently wrong. If your workflow has unnecessary steps, AI will just do unnecessary things faster.
The fix is almost always the same:
- Map the actual process - not what the manual says, but what people actually do.
- Remove what doesn't need to exist - redundant approvals, duplicate data entry, manual steps that exist "because we've always done it."
- Then decide if AI adds value - sometimes it does. Sometimes a simple integration or automation is all you need.
The unsexy truth
The most impactful projects I've worked on didn't start with AI. They started with sitting down with the people doing the work and asking: "Walk me through your day." That's where the real bottlenecks live - not in a pitch deck about machine learning.
Once the process is clean, AI can do incredible things. Classify incoming requests. Extract data from unstructured documents. Predict issues before they happen. But it only works when the foundation is solid.
So before you call someone like me
Ask yourself: if we had a perfect intern who never made mistakes and worked 24/7 - would our process actually make sense? If the answer is no, fix that first. Then we can talk about AI.
The companies that get the most out of AI aren't the ones who adopt it first. They're the ones who know what problem they're solving before they pick the tool.