When teams evaluate AI automation, they compare the cost of the tool against the cost of the manual process. The problem is that most teams dramatically underestimate the cost of the manual process.
The time nobody tracks
Ask a sales rep how long prospect research takes. They’ll say “a few minutes.” Time them, and it’s 20-30 minutes per prospect. They don’t count the tab-switching, the Slack interruption in the middle, the context switch back, or the formatting of their notes.
Ask a support manager how long ticket triage takes. They’ll say “it’s quick.” But add up the time across 200 tickets a day — reading, categorizing, prioritizing, assigning, sometimes re-reading because someone categorized wrong — and it’s 8-10 hours of team time daily.
This invisible time is the real cost of manual work. It doesn’t show up in any budget line item. Nobody tracks “hours spent copying data between tabs” or “time lost re-reading a ticket because triage was wrong the first time.”
A framework for calculating ROI
Step 1: Identify the task
Pick a specific, repeatable task. Not “marketing” — that’s too broad. “Repurposing a blog post into social media posts, newsletter content, and email copy.” That’s concrete enough to measure.
Step 2: Measure the real time
Time the task end-to-end for a week. Include:
- Active work time — Actually writing, researching, formatting
- Context-switching time — The 5-10 minutes it takes to get back into flow after an interruption
- Rework time — Fixing errors, redoing work that didn’t meet standards
- Coordination time — Asking questions, waiting for approvals, aligning with others
Most tasks take 2-3x longer than the “active work” estimate suggests.
Step 3: Calculate the current cost
Multiply the real time by the fully loaded cost of the person doing it. A team member earning $80K/year costs roughly $50/hour including benefits and overhead. A manager at $120K costs roughly $75/hour.
Example: Prospect research
- 25 minutes per prospect × 50 prospects/week = 20.8 hours/week
- 20.8 hours × $50/hour = $1,040/week = $54,000/year
That’s one sales rep spending half their time on research instead of selling.
Step 4: Calculate the automated cost
For AI automation, the cost has two components:
Platform cost: JieGou subscription (varies by plan)
AI token cost: The actual cost of running the recipe or workflow. This varies by model and task complexity, but typical ranges:
- Simple extraction/classification: $0.002-0.005 per run
- Content generation: $0.01-0.03 per run
- Complex analysis: $0.05-0.15 per run
Example: Prospect research (automated)
- $0.065 per prospect (Sonnet for research + Haiku for qualification + Sonnet for outreach)
- 50 prospects/week × $0.065 = $3.25/week = $170/year in AI costs
Step 5: Compare
| Manual | Automated | |
|---|---|---|
| Annual cost | $54,000 | $170 + platform |
| Time per prospect | 25 min | < 1 min |
| Quality consistency | Variable | Consistent |
| Scales with volume | Linearly (hire more) | Marginal cost only |
Even with a generous platform subscription, the ROI is orders of magnitude positive. The sales rep gets 20 hours/week back for actual selling.
ROI by department
Here are realistic examples across departments:
Sales: Lead qualification pipeline
- Manual: 15 min/lead × 200 leads/month = 50 hours/month = $2,500/month
- Automated: $0.065/lead × 200 = $13/month in AI costs
- Time recovered: 50 hours/month for selling
Marketing: Content repurposing
- Manual: 3 hours per blog post × 4 posts/month = 12 hours/month = $900/month
- Automated: $0.08/post × 4 = $0.32/month in AI costs
- Time recovered: 12 hours/month for strategy
Support: Ticket triage
- Manual: 3 min/ticket × 200 tickets/day = 10 hours/day = $12,500/month
- Automated: $0.005/ticket × 4,000/month = $20/month in AI costs
- Time recovered: 200+ hours/month for complex issue resolution
Finance: Invoice processing
- Manual: 15 min/invoice × 50 invoices/week = 12.5 hours/week = $2,500/month
- Automated: $0.04/invoice × 200/month = $8/month in AI costs
- Time recovered: 50 hours/month for analysis and planning
HR: Resume screening
- Manual: 10 min/resume × 100 candidates/month = 16.7 hours/month = $1,250/month
- Automated: $0.005/resume × 100 = $0.50/month in AI costs
- Time recovered: 16.7 hours/month for interviews and candidate engagement
The compounding effect
These numbers look at individual tasks. The real impact is compounding:
-
Time recovered enables higher-value work. The sales rep doesn’t just save 20 hours — they spend those 20 hours on calls and relationships that generate revenue.
-
Consistency improves downstream quality. When every prospect gets researched with the same thoroughness, outreach quality goes up. When every ticket gets triaged consistently, resolution times improve across the board.
-
Capacity scales without headcount. Doubling your lead volume doesn’t require doubling your research team. The marginal cost of processing one more lead is $0.065, not $0.065 + a proportional share of a salary.
-
Knowledge persists. When the person who’s great at prospect research leaves, their approach is encoded in the recipe. The replacement rep produces the same quality output on day one.
Building the business case
When presenting the ROI internally, focus on three numbers:
- Hours recovered per week — This is the most tangible metric for team leads
- Annual cost comparison — Manual cost vs. automated cost (AI tokens + platform)
- Quality consistency — The hardest to quantify but often the most impactful
Avoid overpromising. The AI doesn’t eliminate all manual work — it eliminates the repetitive parts. Someone still reviews the output, makes decisions, and handles exceptions. A realistic claim is “reduces time spent on [task] by 80%” rather than “fully automates [task].”
JieGou includes time tracking per recipe (estimatedManualMinutes) that feeds into the analytics dashboard, so you can measure actual time savings against the manual baseline after you deploy.