The Audit Comes Before the Automation
You export a week of tasks from your project board, paste them into Claude, and ask one question: "Which of these actually need to happen?" Twelve minutes later you have a categorized list showing that seven of your twenty-two recurring workflows exist for no reason other than habit. That's not a speed improvement. That's a different kind of business.
That's the part most AI conversations skip.
The Viral Post Got It Half Right
There's a widely-shared argument making the rounds that says AI won't fix broken processes, it'll just break them faster. And that part is true. If your client onboarding takes three weeks because nobody owns the handoff between sales and operations, automating the handoff emails doesn't fix the three weeks. It just sends the wrong emails faster.
But here's where the argument leads most business owners astray. They hear "fix your processes first" and they interpret it as "wait until things are clean before touching AI." So they wait. And the mess compounds.
The actual takeaway is almost the opposite. AI is the best diagnostic tool you have right now for figuring out which processes are worth keeping at all. Not speeding up. Not automating. Keeping.
Most of Your Workflows Shouldn't Be Automated. They Should Be Deleted.
Here's something that comes up constantly when we work with businesses in the 10 to 200 employee range. They have somewhere between fifteen and twenty-five recurring workflows. Three to five of them are genuinely load-bearing, meaning they directly create revenue, retain customers, or prevent a compliance problem. The rest? They exist because someone set them up three years ago and nobody ever asked whether they still made sense.
Asana's research found that knowledge workers spend roughly 58% of their time on coordination work, status updates, and repetitive admin. That's not work. That's the residue of processes that were never questioned.
The mistake most businesses make is treating all of that as automation candidates. It isn't. Most of it is elimination candidates.
And AI can help you tell the difference in a couple of hours, not the weeks a traditional ops consultant would charge for.
The Three Questions That Sort Everything Out
Before you automate anything, run every recurring workflow through this filter.
Does it create revenue? Does it retain a customer? Does it prevent a compliance or operational problem?
If yes to any of those, it's worth rebuilding with automation. If the honest answer is "we do it because we've always done it," that's not a process. That's a habit wearing a process costume. Kill it.
This sounds simple, and it is. The problem is that most business owners are too close to their own operations to apply it cleanly. That's exactly where a well-prompted AI tool becomes useful, not as an automator, but as a neutral third party that doesn't care about your organizational politics or your team's attachment to the way things have always been done.
What the Two-Hour Audit Actually Looks Like
This is not a theoretical exercise. Here's how to run it.
Step one: Pull one week of recurring work. Export your task board from Asana, ClickUp, or Monday. Grab your recurring calendar meetings. Pull your SOP list if you have one. Even a rough brain dump of "things we do every week" works fine.
Step two: Open Claude or ChatGPT and use a prompt like this: "Here is a list of recurring tasks and workflows from my business. Categorize each one as: creates revenue, retains customers, reduces risk or compliance exposure, or internal habit with no clear business value. For anything in the last category, suggest whether it should be eliminated or could be replaced with a system-generated trigger."
Step three: Review the output. You're not accepting it blindly. You're using it as a forcing function to have honest conversations about work that's been running on autopilot.
Most business owners who do this exercise find five to eight workflows they can stop immediately. No automation needed, no new tools, just stop doing the thing.
Step four: From what's left, pick two or three workflows that are genuinely high-value and repeatable. Those are your automation candidates. Typically these are things like lead intake and routing, client onboarding steps, meeting follow-up and task creation, invoice chasing, or weekly reporting that someone is manually compiling.
That's it. Two hours. More actionable than most ops audits.
What to Actually Build With Automation
Once you've done the elimination work, the automation decisions get much cleaner. You're not trying to automate everything. You're rebuilding the three workflows that actually matter.
Here's a practical example. A service business owner does the audit above and identifies that client onboarding is genuinely load-bearing but currently a mess of manual emails, document requests, and Slack messages. The workflow itself is worth keeping. The manual execution isn't.
The rebuild looks like this. Zapier or Make connects the CRM to a document collection tool. When a deal closes in HubSpot, an automated sequence kicks off: welcome email goes out, a document request link is sent, a task is created in the project board, and the account manager gets a Slack notification. If the client hasn't submitted documents in 48 hours, a follow-up fires automatically.
That's not a complicated build. It takes a few hours to set up in Make or Zapier. But it only makes sense to build it after you've confirmed the underlying process is worth keeping, and after you've killed the five other workflows that were cluttering the picture.
The businesses winning with AI right now aren't the ones who automated the most. They're the ones who used AI to figure out what to stop doing first, then automated the workflows that remained.
The Quarterly Habit That Keeps This Working
One audit isn't enough. Businesses at the 10 to 200 employee stage change fast. A workflow that made sense six months ago might be dead weight now. A manual process that was "too small to bother automating" might be consuming twenty hours a week once you've grown.
Build the audit into your quarter. Same exercise, same questions, same two hours. Export your recurring work, run it through Claude or ChatGPT, apply the three-question filter, kill what doesn't pass, rebuild what does.
McKinsey's research on successful AI adoption consistently points to one differentiator: businesses that tie AI to specific operational outcomes and revisit those outcomes regularly outperform the ones that treat it as a one-time implementation. The quarterly cadence is how you stay on the right side of that line.
Three Things to Do This Week
First, do the export. Pull your task board, your recurring meetings, and your SOP list into a single document. Don't overthink the format.
Second, run the audit prompt in Claude or ChatGPT. Use the framing above: categorize by revenue, retention, compliance, or habit. Ask it to flag elimination candidates explicitly. The output won't be perfect, but it'll surface conversations worth having.
Third, pick one workflow to kill and one to rebuild. Not five. Not ten. One of each. Kill the habit-based one this week, no tools required. Start scoping the rebuild in Make or Zapier using their free tiers to map the logic before you commit to anything.
The goal isn't to automate your business with AI. It's to use AI to figure out which version of your business is worth building, and then build that one.
If you want a second set of eyes on your audit or help scoping what to rebuild, nextwaveharbor.com/connect is a good place to start.