Stop Automating. Start Eliminating.

You see manual work. You build automation. Faster emails. Smarter scheduling. Better tracking.

But what if the work shouldn't exist at all?

Last week: Katie West rebuilding the same dashboard at three different companies.

Week before: Andrew Mewborn manually A/B testing emails in spreadsheets.

They're not asking for faster dashboards or better spreadsheets. They're asking for the work to disappear.

The difference matters. Automation makes work faster. Elimination makes work unnecessary.

The Automation Trap

Most AI tools today make work faster:

• Scheduling tools reduce back-and-forth
• AI assistants write emails quicker
• Dashboards surface insights faster

They're automation. Same workflow, less friction.

The problem: You're still doing the work. Just faster.

Katie still has to categorize customers—now with AI help.

Andrew still has to run tests—now with better tools.

Sales teams still have to follow up—now with AI-generated emails.

Better tools. Same workflows. Still manual.

What Elimination Looks Like

We spoke to Paul Irving this week. He's a VC who sees patterns across 100+ companies. Different industries, same problems.

There's so much work everybody does that they don't want to do, and it isn't actually what humans are best at. How much of that can be automated? Like, a ton.

But he's not talking about automation. He's talking about elimination.

Here's what that looks like.

Example 1: The 4-Day Email

Current workflow (automated):

Friday afternoon. You email a client for a simple answer. AI drafts a better subject line. Better body copy. More likely to get opened.

Client is 3 hours ahead. Doesn't see it until Monday morning. Their AI summarizes your email. Busy Monday, they respond Monday afternoon. You get the answer Tuesday morning.

Four days. For a yes or no.

You automated email writing. You didn't eliminate the wait.

Elimination approach:

Voice AI agent handles the back-and-forth asynchronously. Client calls Saturday afternoon while walking the dog. Answers the question. You have the answer Saturday night.

No waiting. No email chain. Work eliminated.

We've just lost 4 days, and it was something that didn't need a consultative conversation. It was a simple answer back, coordination, schedule, sign off.

Example 2: Accounts Receivable

Current workflow (automated):

Invoice sent. AI generates personalized follow-up emails. AI tracks responses. AI escalates to human when needed.

Still takes weeks. You automated the emails. You didn't eliminate the loop.

Elimination approach:

Agentic AR handles the entire conversation. Questions about the invoice? Agent answers with invoice data. Discrepancy on the line item? Agent reconciles with purchase order. Payment terms don't work? Agent negotiates within pre-approved parameters.

No human in the loop unless it's an exception. Work eliminated.

There's billions and billions of dollars stuck in AR in every single industry at any given time. Even if you shave a day off that, two days off that, by being able to make the cycles of communication and transaction faster, we're talking about trillions of dollars of value in the economy.

The opportunity isn't better invoice tracking. It's eliminating the human bottleneck entirely.

Example 3: Market Research

Current workflow (automated):

You need 40 expert calls for a research project. AI helps you write outreach emails. AI schedules based on calendar availability. AI transcribes calls. AI summarizes findings.

Still takes weeks. You automated scheduling and summarization. You didn't eliminate the sequential bottleneck.

Elimination approach:

Voice AI conducts all 40 interviews simultaneously. Experts call whenever it works for them. Saturday afternoon. Tuesday morning. Thursday at 11pm. Doesn't matter.

No scheduling. No waiting for 40 people's calendars to align. Synthesis happens in real-time.

I could have those 40 conversations at the same time. You're an executive I'm targeting for that research project. You can call me when you're walking your dog on a Saturday afternoon. It's not about scheduling it for Tuesday in between a board meeting and some other meeting.

The work doesn't get faster. It becomes parallel instead of sequential.

The Pattern

Automation: Make the existing process better.

Elimination: Remove the need for the process.

Katie West spending hours writing customer emails?

• Automation: AI helps her write faster
• Elimination: System sends personalized emails automatically based on usage triggers

Andrew Mewborn manually A/B testing in spreadsheets?

• Automation: Better spreadsheet tools
• Elimination: System runs tests automatically, advances to next variable when threshold hit

Car dealerships missing after-hours calls?

• Automation: Better voicemail system
• Elimination: Voice AI answers 24/7, schedules appointments automatically

The work doesn't get faster. It disappears.

Why Builders Miss This

You're trained to look for workflows to improve.

Someone does X manually → Build tool to do X faster.

But the real opportunity is different.

Someone does X manually → Ask why X exists → Eliminate X entirely.

Example from our interviews:

'What Katie said: "I manually write every customer email."'

• Automation thinking: Build AI email writer for CS teams.
• Elimination thinking: Why is she writing emails at all? Usage data exists. Triggers are clear (usage dropped 20%, feature not enabled, no logins). System should detect and send automatically.

'What Andrew said: "I spend 10+ hours a week A/B testing emails."'

• Automation thinking: Build better A/B testing dashboard.
• Elimination thinking: Why is he running tests manually? System should test automatically, measure results, advance to next variable when threshold hit, optimize continuously without human intervention.

If you look at the signal versus noise of everybody's day, so much is stuff that isn't actually creating value, but needs to happen.

The real question: Does it actually need to happen? Or does it need to happen this way?

The Automate vs Eliminate Framework

When you find manual work, ask three questions:

1. Is human judgment required?

• Yes → Automate (make it faster)
• No → Eliminate (remove the human)

2. Is this workflow necessary or inherited?

• Necessary → Automate (make it better)
• Inherited from pre-AI era → Eliminate (redesign from scratch)

3. Does the output require human creativity?

• Yes → Automate (assist the human)
• No → Eliminate (let AI handle it)

Examples:

Customer categorization (Katie's problem):

• Human judgment required? No. Usage data is objective.
• Workflow necessary? No. Inherited from era before real-time data.
• Requires creativity? No. Pattern matching on signals.
• Answer: Eliminate. Automatic categorization based on usage.

Invoice follow-up (Paul's portfolio company):

• Human judgment required? Not for standard cases.
• Workflow necessary? No. Inherited from pre-voice-AI era.
• Requires creativity? No. Structured conversation with clear rules.
• Answer: Eliminate. Agentic AR/AP handles end-to-end.

Scheduling 40 expert calls (Paul's research example):

• Human judgment required? No. Finding mutual availability.
• Workflow necessary? No. Inherited limitation of human schedules.
• Requires creativity? No. Calendar coordination.
• Answer: Eliminate. Async voice AI interviews, no scheduling.

Negotiating enterprise contract:

• Human judgment required? Yes. Strategic decisions.
• Workflow necessary? Yes. Relationship building matters.
• Requires creativity? Yes. Deal structure, terms, trust.
• Answer: Automate. AI assists, human decides.

The framework helps you see the difference.

What to Build

Five elimination opportunities from what we're seeing:

1. Async Communication Layers

Don't build better email tools.

Build voice AI that handles time-zone-spanning conversations asynchronously.

Client in Singapore. You're in New York. Instead of 3-day email chains, voice agent handles the conversation. They call when convenient. You get structured output.

The signal to listen for:

• I'll get back to you Monday
• Waiting on them to respond
• Different time zones

2. Agentic Transaction Handlers

Don't build invoice tracking dashboards.

Build agents that resolve invoice questions, handle standard back-and-forth, escalate only exceptions.

AR agent sees discrepancy. Pulls purchase order. Reconciles line items. Negotiates payment terms within approved parameters. Escalates to human only if outside boundaries.

The signal to listen for:

• Billions stuck in AR
• Emailing back and forth on invoices
• Takes weeks to collect

3. Parallel Work Platforms

Don't build better scheduling tools.

Build systems that do 100 things simultaneously instead of sequentially.

Market research, customer calls, expert interviews, sales outreach. Anything that's currently "schedule 40 people one at a time over 3 weeks" becomes "40 conversations this weekend."

The signal to listen for:

• Takes weeks to schedule
• Waiting on 40 people
• One at a time

4. Trigger-Based Action Systems

Don't build better email composers for CS teams.

Build systems that detect usage triggers and act automatically.

Usage drops 20%? System sends personalized email with specific data. Feature not enabled? System sends training video. No logins in 30 days? System schedules check-in call.

Katie doesn't write emails. System sends them when conditions are met.

The signal to listen for:

• I manually email customers when...
• I have to check if...
• I spend hours writing...

5. Self-Optimizing Workflows

Don't build better A/B testing dashboards.

Build systems that test, learn, advance automatically.

Week 1: Test subject lines. Hit threshold. Advance. Week 2: Test hooks. Hit threshold. Advance. Week 3: Test body. Hit threshold. Advance.

Andrew doesn't run tests. System runs them continuously.

The signal to listen for:

• I spend X hours testing
• Manual optimization
• One variable at a time

How to Know Which to Build

Listen for elimination signals in operator conversations.

'"I have to..."'

Why do they have to? Can the system do it instead?

'Katie: "I have to ask engineering to pull email lists." → Elimination opportunity: Queryable database interface. No engineering requests needed.'

"I'm waiting for..."

Why are they waiting? Can this be parallel or async?

Paul: "We're waiting for 40 people to schedule calls." → Elimination opportunity: Async voice AI. No waiting.

'"This takes hours/days/weeks..."'

Why does it take time? Is it sequential when it could be parallel?

'Paul: "Four days for a simple answer because of time zones." → Elimination opportunity: Async communication layer. Answer in hours

'"I check if..."'

Why are they checking? Can the system detect and act automatically?

'Katie: "I check if customers are logging in

'"We go back and forth..."'

Why back and forth? Can an agent handle the entire conversation?

'Paul: "Emailing back and forth on invoices." → Elimination opportunity: Agentic AR. Agent resolves the full cycle.'

The signals are everywhere. You just have to listen differently.

The Economics Changed

In 2020, elimination wasn't economical.

Building voice AI that could handle 100,000 conversations simultaneously? Required massive infrastructure. Specialized team. 18+ months to ship.

Building agentic systems that could resolve invoice disputes without humans? Same problem. Too expensive. Too complex.

The economics didn't justify it for most problems.

So we automated instead. Better tools for existing workflows. Faster, not eliminated.

In 2026, elimination is economical.

• Voice AI: $6/hour vs $25/hour human.
• LLM costs: 1000x cheaper than 2021.
• Infrastructure: Trivial to scale.
• Agents: Can be built by 2-3 people in weeks, not 20 people in 18 months.

The same problems. Different economics.

What wasn't worth eliminating before is worth eliminating now.

How much of the work people do that isn't actually creating value but needs to happen can be automated? Like, a ton.

The question isn't "can we eliminate it?"

'The question is "what are we waiting for?"'

Stop Automating. Start Eliminating.

Last three weeks:

Vol 1: The builder opportunity exists. Economic shift makes previously unviable problems worth solving.

Vol 2: Smart operators doing dumb work. Repetitive tasks that scale linearly with growth.

Vol 3: Stop making that work faster. Make it unnecessary.

The pattern:

Andrew doesn't want better testing tools. He wants testing to happen automatically without his manual intervention.

Katie doesn't want faster email writing. She wants emails to send automatically when triggers are detected.

Paul's companies don't want better scheduling. They want scheduling eliminated through async voice AI.

Automation is incremental. Elimination is exponential.

When you hear an operator describe manual work, ask:

• Can this workflow be eliminated entirely?
• What would need to be true to make this work unnecessary?
• Is this automation or elimination?

The biggest opportunities aren't in making work 10% faster.

They're in making work 100% unnecessary.

Last week: Smart operators doing dumb work.

This week: The work shouldn't exist.

Next week: Another operator. Another pattern.

The opportunities are hiding in plain sight. You just have to ask the right question.

'Not "how can we make this faster?"'

'But "why does this exist at all?"'

— Mike & Govind

📺 Watch the full conversation with Paul Irving