There’s a moment, when you hear a drip under the kitchen sink and you think: I should fix that. So you drive to the hardware store. You come back with plumber’s tape, a new faucet, a wrench, and approximately zero understanding of where the water is actually coming from. You tighten and replace things, but the drip continues. Eventually, you get on your knees with a flashlight and discover it’s a loose fitting you could have fixed with your fingers, which is somehow both a relief and an insult.

I keep watching teams do this with AI.

They know it can make them more productive. They’ve seen the demos, read the case studies, maybe even spent a Thursday afternoon getting ChatGPT to rewrite their meeting notes in the voice of a true crime series (we’ve all been there). The enthusiasm is genuine. The direction is not. So they grab the shiniest tool and jam it into whatever process hurts the most, and then six weeks later, the drip is still dripping, and there’s a new faucet nobody asked for. The problem was never the tools. The problem was skipping the part where you get on your knees with the flashlight.

That part is the audit. An AI workflow audit is a quick tally of what you do, what it costs, and where it gets stuck, so you automate the right step, not the noisiest one. It sounds tedious. (Nobody got into creative work to draw process maps.) But I've watched enough AI implementations go wrong to know the diagnosis matters more than the fix. I’ve seen teams speed up drafting and still ship nothing faster because the real leak was approvals, not writing, and nobody wanted to admit it.

You can do a first-pass version with a spreadsheet and 45 minutes. Pick one workflow that happens often. List every step in plain language. Then comes the part most people skip because it forces uncomfortable clarity: for each step, decide whether it’s pattern work or judgment work. Pattern work has structured inputs and predictable outputs like scheduling a meeting. Judgment work requires taste, empathy, politics, or accountability (the stuff that makes your job actually hard and occasionally interesting). That distinction will tell you where AI belongs.

Now score each step on two dimensions: how automatable it is and how much value automating it would create. This forces you to chase value instead of convenience. Most teams go sideways right here, automating the equivalent of the faucet while the fitting drips on.

After doing dozens of these with clients, I realized the structure was repeatable even when the plumbing wasn’t. So I built a tool with Claude Code that walks you through the audit and turns your answers into a workflow map. It labels each step in plain English: automate it, assist it, support a human doing it, or leave it alone. I’m particularly attached to the human-only category. (The human-only label is the one most teams need permission to use.)

Here’s what it looks like in action:

Everything I’ve described so far is repair: tighten the fittings, stop the leaks, speed up the slow steps, and replace the parts that corrode. That alone can give you back hours every week. Replace is a different move. It’s when the workflow itself needs to change. New AI tools are emerging so fast that some processes are starting to look like elaborate workarounds for an older world. When an AI agent can handle an entire research-to-brief pipeline that currently spans five team members and three weeks, or when your versioning and feedback loops collapse from twelve steps to three, you’re redesigning the system. The tool helps with repair. Replace is a human call: someone has to decide what stays, what goes, and what gets rebuilt.

My platform handles the repair. The replace work requires a human in the room who knows where the pipes actually go and where the water needs to end up next year. Someone who can look at a freshly optimized workflow and say, “this is faster, and it’s still the wrong architecture.”

If you want to do this with me instead of a spreadsheet, I'll walk you through the platform live. Pick a workflow, we'll run the audit together, and you'll leave with a roadmap you can hand to your team on Monday. The redesign work, honestly, is the part I like best. There’s something satisfying about finally finding the fitting that’s been dripping this whole time.

— Lauren Eve Cantor

thanks for reading!

if someone sent this to you or you haven’t done so yet, please sign up so you never miss an issue.

I’ve also started publishing more frequently on LinkedIn, and you can follow me here

if you’d like to chat further about opportunities or interest in AI, or this newsletter, please feel free to reply.

banner images created with Midjourney.