If you’re evaluating an AI Office engagement on year-1 ROI, you’re watching the wrong number. Year 1 is honest, focused, capability-building work — the first workflow ships, the team learns, the roadmap takes shape. ROI is usually positive but modest: 1-2X cumulative is normal.
Year 2 is where the math gets interesting. The team has built operating discipline. The workflow library is growing. Each new workflow ships faster than the last. The workflows already in production are producing value at full run-rate. The compounding kicks in.
Here’s what year 2 actually produces, drawn from the clients who’ve now been with us 24+ months.
The composite
The picture below is composited from 4 clients in services, healthcare admin, and light manufacturing who’ve been with us 24+ months and reached a similar shape of year-2 maturity.
The company is a $30M services firm, $5M EBITDA, 100 employees. It started at AI Office Operator tier ($5,000/mo) in year 1, moved to Embedded ($10,000/mo) at month 7, and has continued at Embedded through year 2.
What changed between year 1 and year 2
End of year 1:
- 4 workflows in production
- ~$330K cash spend
- ~$680K annual run-rate value
- Internal AI Operations Lead newly hired
- multi-quarter program kicked off late in the year
End of year 2:
- 9 workflows in production (5 new ones shipped in year 2)
- 1 multi-quarter program completed — the deeper operational change that started late in year 1
$520K year-2 cash spend ($850K cumulative)- ~$1.8M annual run-rate value (compounding from the year-1 base plus new workflows)
- AI Operations Lead now running most of the day-to-day; the partner role shifted toward senior strategy and Sprint execution
The ratios:
- Year-2-only ROI: 3.5X year-over-year incremental value vs. year-2 spend
- Cumulative ROI across years 1+2: ~2.1X cumulative value / cumulative spend
- Run-rate ROI at end of year 2: ~3.5X annualized vs. cumulative spend
What got built in year 2
New workflows shipped (5):
- Customer service AI for Tier 1 (started late year 1, in production by Q1 year 2)
- Dispatch optimization (a multi-quarter program piece, shipped Q2 year 2)
- AP automation with vendor matching (shipped Q3 year 2)
- Internal knowledge search across 20 years of past projects (shipped Q3 year 2)
- AI-drafted board materials and investor updates (shipped Q4 year 2)
Existing workflows refined:
- The original quoting workflow got a v2 build — expanded to a second product line and integrated with the dispatch system
- The field-to-office reporting workflow expanded to a second division
- The leadership briefing workflow added a quarterly variance analysis component
Operating capability built:
- AI Operations Lead promoted to a director-level role
- A second internal AI person hired (a junior data analyst working alongside the AI Operations Lead)
- Governance posture matured; first internal AI audit conducted by the CFO
- Internal prompt library grown to 200+ documented prompts and templates
Where the value compounded
Three sources of compounding.
1. Workflow throughput. Year 1 shipped 4 workflows in 12 months. Year 2 shipped 5. By itself that’s modest acceleration — but each year-2 workflow shipped in 4-6 weeks vs. 6-10 weeks in year 1. Engineering velocity compounded while strategic discovery time stayed roughly the same.
2. Workflow value. The first 4 workflows produced their full annual run-rate in year 2 (they only partial-year’d in year 1). The 5 new workflows produced partial run-rate. The portfolio value compounded.
3. Cross-workflow benefits. Year 2 surfaced something we hadn’t predicted: workflows that fed each other. The quoting workflow’s output started feeding dispatch optimization. The field reports started feeding AP automation. Each new workflow made the others more valuable. We call this the “integration dividend” — value that isn’t visible until multiple workflows are in production together.
What the partnership looked like in year 2
The shape of the work shifted.
Year 1:
- Heavy engineering — most of the build was new
- Heavy strategic discovery — the roadmap was being built
- Senior strategist carried a high working-session load
Year 2:
- Engineering still substantial, but more pattern-driven
- Strategic discovery lighter — the roadmap was mature; new candidates were incremental
- Senior strategist’s role shifted toward executive interface, deeper strategic questions, and quality oversight
- Internal AI Operations Lead handling more of the day-to-day
By month 18, the partnership felt like a senior advisor with engineering capability rather than a build team. The client team owned more; we provided what the internal team couldn’t justify hiring for.
The CFO conversation at end of year 2
This is an end-of-year-2 review we sat in on.
CFO: “Two years in, $850K cumulative. What’s the math?”
CEO: “Run-rate annual value at end of year 2 is $1.8M. That’s the year-3 baseline before any new workflows ship. Year-3 spend will be roughly $400-500K. New value in year 3 will probably add another $500-700K to run-rate.”
CFO: “So by end of year 3, we’re at $2.3-2.5M annual run-rate value against $1.3M cumulative spend. ROI ratio approaching 2X cumulative.”
CEO: “Right. And we have an internal capability now that compounds without proportional partner spend.”
CFO: “Defensible. Continue.”
That’s the version of the conversation that survives scrutiny. The numbers are honest. The trajectory is real. The capability is genuine.
Where the partnership ends up in year 3
For most of our 24+ month clients, year 3 looks like:
- AI Office tier drops from Embedded to Operator, or stays Embedded with reduced hours
- Internal AI Operations Lead handles 60-70% of what the partner was doing
- 2-3 Sprints per year on the deeper builds the internal team can’t tackle alone
- Quarterly strategic review continues
- The partnership feels like a senior advisory relationship more than a build relationship
Total partner spend in year 3 is typically 40-60% of year 2. Total internal AI investment is higher. Value produced is materially higher than year 2.
This is the steady state — though not the only one. Some clients stay at Embedded for years. Some bring everything internal and we stay involved at Sherpa for periodic deeper engagements. The shape varies. The pattern of compounding doesn’t.
What this means for the owner reading this
If you’re considering an AI Office engagement and you’re focused on year-1 ROI, you’re focused on the wrong metric. Year 1 is foundation. Year 2 is where the math gets compelling. Year 3 is where the structural advantage starts being visible to competitors.
The clients who’ve gotten the most value from us treated year 1 as an investment in capability, not a hunt for quick wins. They didn’t get all the wins they hoped for in year 1 — they got all the wins they hoped for and more by the end of year 2.
That’s the version of this we’d want you to evaluate us on.