Signs Your Workflow Should Be Deleted, Not Optimized

A workflow should be deleted rather than optimized when its steps exist because humans are slow — queues, handoffs, approvals, vendor waits — rather than because they create value. The test: if you designed the outcome from scratch today with an agent available, would any of the current steps survive? If not, optimization is just embalming.

This distinction is the entire thesis of Changing Workflows. The documented cases aren't "we made our process 20% faster." They're a two-day legal workflow that now takes five minutes, a business that's 100% off WordPress, a vendor queue that no longer gets used. The process didn't improve. It stopped existing. Here are the six signs yours is in that category.

Sign 1: The workflow is queue-shaped

If the honest description of a workflow includes "then it sits until someone gets to it" — a vendor's queue, your own team's backlog, an approval chain — the waiting is the workflow. Verbatim, from a weekly Optimus call transcript:

"The last two days, we've been using an agent to clean up some really bad malware that got on three of our client sites. It did it so fast that I kept asking, "Did you check all the files?" Two months ago my managed WordPress hosting company told me, "You're in our queue, it's going to be a two-week process" — and sure enough it was. This time it took me a day and a half, and it found stuff that was incredible."
— Hubert, Optimus weekly call, Jun 17, 2026

The hosting company's two-week process wasn't two weeks of work. It was a queue defending the vendor's staffing model. You can't optimize your position in someone else's line — you delete the line.

Sign 2: Most of the steps are handoffs

Map the workflow and count the steps that are transfers rather than transformations: forwarding, assigning, re-keying between systems, "circling back." Handoffs exist because one person can't hold the whole job. An agent can — which is why handoff-heavy workflows don't get faster with agents; they collapse into a single step. If more than half your steps are handoffs, you're looking at a deletion candidate, not an optimization project.

Sign 3: The workflow defends a platform dependency

Some workflows exist only to keep a platform alive: plugin updates, security patches, compatibility testing, specialist retainers. Optimizing those means maintaining your cage more efficiently. One member deleted the cage instead — a 165-page WordPress site rebuilt to static HTML in two days with one agent and three sub-agents, ending in the sentence "I'm 100% off WordPress now." Every maintenance workflow attached to that dependency died with it. Ask of any recurring workflow: does this serve the business, or does it serve a platform the business happens to sit on?

Sign 4: You'd never design it this way today

The cleanest test in the list. Describe the outcome, then ask: starting blank today, with an agent available, what's the process? If the answer shares no steps with your current workflow, the current workflow is a historical artifact — it accreted around old constraints (slow humans, dumb tools, vendor requirements) that no longer bind. The gap between "what we do" and "what we'd design" is exactly the gap agents close. The head-to-head comparison of what agents change versus what automation changes is in workflow automation vs. AI agents.

Sign 5: Optimization projects keep not sticking

If a workflow has survived multiple improvement pushes — new SOP, new tool, new hire — and keeps reverting to painful, that's not a discipline problem. It's the workflow telling you its structure is the problem. Processes that resist optimization are usually load-bearing walls in a building that needs demolition, not renovation.

Sign 6: The output is a formality

Reports nobody reads, status meetings that restate dashboards, approvals that approve 100% of the time. These workflows produce reassurance, not value. Don't hand them to an agent — that just produces reassurance faster. Kill them and see who notices.

What do you do with a workflow that shows these signs?

Not a reckless rip-out. The documented pattern is: rebuild the outcome around an agent, run it in parallel, verify until it beats the old way, then delete the old way completely — the six-step version is in how to rebuild a business process around AI agents. And before you start, put a number on what the workflow is costing you across payroll, vendors, queues, and declined opportunities — the four-part costing method is here. Deletion arguments get much easier with a dollar figure attached.

FAQ

What does it mean to delete a workflow instead of optimizing it?

Optimizing means the same process runs faster or cheaper. Deleting means the outcome gets produced a different way and the process stops existing — like legal work that took two days now taking five minutes because the two-day version is simply gone, or a business leaving WordPress entirely instead of speeding up WordPress maintenance.

How do I know a workflow is deletable?

Ask what the workflow would look like if you designed it today with an agent available. If the honest answer contains none of the current steps, the workflow exists for historical reasons — because humans were slow, tools were limited, or a vendor's process required it — and it's a deletion candidate.

Isn't deleting workflows risky?

Deletion in the documented cases isn't reckless — it follows a parallel-run phase where the agent-built replacement gets verified against the old process. The risk that actually bites owners is the opposite one: spending years optimizing processes that competitors are deleting.

What workflows should NOT be deleted?

Workflows whose value is the judgment inside them — pricing, hiring, client relationships, strategy. Those get more owner attention after the deletable workflows are gone, which in the documented cases is the point: the recovered hours go to decisions only the owner can make.