The Operator’s ROI Model

MODULE_04 · ROI_MATH

The Operator’s ROI Model

How to value AI agent deployments in dollars, time, and pipeline. The exact model we use inside ZIP AI to decide what to build first — and the business case template that gets it approved.

~30 min · 5 lessons FREE · NO SIGN-UP

// LESSON 01The three currencies

5 min · Read

Most ROI calculations for AI fail because operators try to express everything in dollars. AI agents create value in three currencies, and only one of them is dollars. Treat the other two as second-class and you’ll either overspend on low-impact builds or underbuild because the dollar number didn’t look big enough.

$
Currency 01
Dollars
Direct margin impact. Easiest to measure, easiest for finance to approve. Usually the smallest of the three in the first six months.
◷
Currency 02
Time
Hours of operator + team time returned. Not “saved” — redirected. A founder’s 6 hours/week is worth more than a $20k contractor.
↗
Currency 03
Pipeline
More qualified opps, faster response, better follow-through. Shows up next quarter, not this one. Operators systematically undervalue this.
Key Takeaways
  • Three currencies: dollars, time, pipeline. All three count.
  • Dollars are easiest to measure, often smallest in months 1–6.
  • Pipeline shows up next quarter. Don’t discount it just because finance does.

// LESSON 02The agent valuation model

7 min · Read · Interactive

Here’s the formula we use to value any proposed agent. One equation. Put it in a spreadsheet, run every potential agent through it, and you have a defensible priority list in 30 minutes.

// AGENT_VALUE.FORMULA Agent Value = (Cost Saved + Pipeline Lift × Margin% × Confidence) − (Build + Run) ÷ Risk

Run the math on your own agent

Plug in your numbers. Year-1 net value updates live. Green = ship it.

Cost Saved$26,000
Pipeline Lift (raw)$187,500
Pipeline × margin × conf$78,750
Total Value$104,750
− Build − Run−$31,400
÷ Risk1.5×
Year 1 Net$48,900
GREEN LIGHT · SHIP IT
Key Takeaways
  • One formula, every agent, ranked side-by-side.
  • Be conservative on confidence. Be honest on risk.
  • Year-1 net > $25k = green light.

// LESSON 03The priority matrix

6 min · Read · Visual

Once you’ve valued every candidate agent, don’t just sort by value. Sort on a 2×2 of Value vs Time-to-Ship. The agent worth $200k that takes 9 months loses to the agent worth $50k you can ship in 14 days.

Year-1 Net Value →
Time to Ship →
Q1 · SHIP NOW
Q2 · ROADMAP
Q3 · LEARNING
Q4 · CUT
Q1: High value, fast Q2: High value, slow Q3: Low value, fast Q4: Low value, slow
The Prioritization Rubric
  • Year-1 net > $25k AND ship < 30 days → SHIP NOW
  • Year-1 net > $50k AND ship < 90 days → SHIP THIS QUARTER
  • Year-1 net > $100k AND ship > 90 days → ROADMAP (build OS first)
  • Year-1 net < $25k AND ship < 14 days → LEARNING SHOT (only if it unlocks Q1)
  • Everything else → CUT
Key Takeaways
  • Sort on Value × Speed, not just value.
  • Build the fast wins first — they fund the slow, high-value ones.
  • Have the courage to cut Q4. Most operators don’t.

// LESSON 04Building the business case

6 min · Read · Template

Even when you’re the operator who decides, you usually still need to sell the build internally. The business case is a one-page document. Anything longer doesn’t get read.

business_case.md
01 · Problem
Specific pain, in numbers. e.g. “SDR team spends 12 hrs/week qualifying inbound leads; 38% of A-leads aren’t contacted within 5 min.”
02 · Proposed agent
What it does. Where it slots in. What’s the blast radius.
03 · Value (the formula output)
Cost saved · pipeline lift · build · run · net · payback period.
04 · Risk + rollback
What could go wrong. How we’d catch it. How we’d roll back.
05 · Ask
Budget. Time. Sign-off.
Include Every Time
  • A specific 12-month dollar number (not a range).
  • The payback period in months.
  • The named human whose time gets returned (not a generic role).
  • The single number you’ll watch weekly.
  • An exit/rollback plan that doesn’t require trust.

// LESSON 05Re-scoring quarterly

6 min · Read

Your initial ROI model is wrong. Every operator’s first model is wrong. The math is right, but the inputs are estimates. The model becomes accurate only after you have real production data — and that takes a quarter.

Every 90 days, take your shipped agents and re-run the model with actuals: hours actually replaced, pipeline actually generated, run cost actually incurred. Update the confidence discount based on how off your estimates were. Then re-rank the backlog with the corrected estimation skill.

Quarterly Review Checklist
  • Pull actuals for every live agent — cost, value, time.
  • Compute variance % vs forecast.
  • Update default confidence (most operators are 30–50% too optimistic).
  • Re-rank the backlog with the new discount applied across the board.
  • Kill any agent whose actual net < run cost.
  • Document lessons in a permanent “agent ledger” you’ll reference for years.
Key Takeaways
  • First model is always wrong. Re-score quarterly with actuals.
  • Most operators are 30–50% over-optimistic on first estimates.
  • Kill agents that underperform — keeping them open costs you compounding focus.
Scroll to Top