Stop Pitching Automation and Start Doing Free Teardowns
Stop Pitching Automation and Start Doing Free Teardowns
Automation is a hard sell - not because it does not work, but because of how most people pitch it. Telling someone their workflow is broken triggers defensiveness. Nobody wants to hear that the process they have built and refined over years is inefficient. The negative reaction you get is not about the technology. It is about the message.
Why Direct Pitching Fails
When you lead with "your workflow is broken and my tool fixes it," you are attacking something personal. People built those workflows. They are comfortable with them. The manual steps are not bugs to them - they are features. They provide control, predictability, and a sense of mastery over their work.
Telling them to automate feels like telling them their expertise does not matter anymore. That is why the response is hostile even when the automation would genuinely help.
This is not a new problem. Sales trainers have known for decades that features do not sell - outcomes do. "Our automation tool handles 500 emails per day" does not move people. "You're spending 8 hours a week on email triage that could take 20 minutes" does. But even outcome-based pitching hits resistance if it is framed as "you're doing it wrong."
The Free Teardown Approach
The approach that actually works is offering a free workflow teardown. Instead of saying "you should automate this," you say "let me watch how you do X and show you where you might be losing time."
The difference is significant:
- No judgment - you are analyzing, not criticizing
- Their choice - they decide what to act on from your findings
- Visible value - they see the time savings before committing to anything
- Trust building - you demonstrate expertise without asking for anything in return
A teardown positions you as someone trying to help rather than someone trying to sell. The natural response to "can I watch you work for 20 minutes and give you my thoughts?" is much more likely to be yes than the response to "let me show you our product demo."
How to Run a Teardown
Step 1: Record the workflow with timestamps.
Ask to observe or screen-record a complete task from start to finish. You are watching for:
- Time spent on each step
- Context switches between applications
- Waiting periods where the person is blocked
- Repeated sub-tasks that appear in every iteration
- Steps that exist to fix errors created by earlier steps
A simple observation template:
Task: Monthly reporting for client X
10:00 - Opens email to find last month's numbers
10:03 - Copies numbers to spreadsheet manually
10:11 - Opens design tool to update chart
10:18 - Exports chart as PNG
10:19 - Opens email client
10:20 - Attaches PNG to email template
10:22 - Adjusts text manually
10:26 - Sends
Total: 26 minutes per client, 15 clients = 6.5 hours/month
Step 2: Categorize every step.
Three categories:
- Mechanical - purely repetitive, no judgment required (copying numbers, resizing images, formatting text)
- Decision-based - requires human judgment (choosing which data to highlight, writing personalized comments)
- Elimination candidates - steps that exist only because of earlier inefficiencies (correcting data that was entered incorrectly, reformatting because the original was in the wrong format)
Step 3: Build the time breakdown.
This is the artifact that matters most. A table showing time per step and annual cost at a conservative hourly rate:
Step Time/iteration Monthly Annual (at $75/hr)
--- --- --- ---
Finding last month's numbers 3 min 45 min $675
Manual spreadsheet entry 8 min 2 hr $1,800
Chart updates in design tool 7 min 1.75 hr $1,575
Export and email assembly 8 min 2 hr $1,800
---
Total 26 min 6.5 hr $5,850
Automatable portion: ~18 minutes per iteration, ~4.5 hours per month, ~$4,050 per year.
Step 4: Present it as a map, not a pitch.
Show the teardown document. Let them read it. Do not start by jumping to solutions. Let them have the reaction "I had no idea I was spending that much time on this."
When someone realizes they are spending 4 hours per week on something that could take 20 minutes, they sell themselves on automation. You do not have to push. The data does the work.
Only after they express interest do you explain what is actually automatable and how.
What Makes a Good Teardown
Focus on frequency, not magnitude. A 30-minute task that happens daily is a better target than a 4-hour task that happens quarterly. The daily task burns 10+ hours per month; the quarterly task burns 4 hours. Automation ROI compounds on frequency.
Count context switches. Moving between applications is cognitively expensive and easy to undercount. Every time someone switches from email to spreadsheet to design tool to email, that is a context switch that takes time to recover from. A workflow with six applications may take 30% longer than the raw task time suggests.
Look for error-correction steps. If someone has a "double-check everything before sending" step, ask why. Often it is because an earlier step is unreliable enough that errors have become expected. That step is a signal that something upstream should be fixed, not automated.
Document what should stay manual. Not everything should be automated. A good teardown explicitly marks decision-based steps as human-owned. This builds credibility - you are not trying to automate everything, just the mechanical parts. That distinction matters to people who are worried about losing control.
The Lesson for AI Agent Builders
This applies directly to building AI desktop agents. Do not ship features and tell users they need them. Watch how actual users work, identify the real friction points, and build tools that slot into existing workflows.
The worst version of an AI agent product is one that requires users to change their workflow to use it. The best version is one that handles the mechanical parts they already do, in the way they already do them, without requiring any behavior change.
That insight comes from teardowns, not from building in isolation.
Fazm is an open source macOS AI agent. Open source on GitHub.