Less Hours, More Revenue
The curve that matters: hours per outcome fall every cycle while revenue recurs.
The core bet is recurring revenue where the consultant hours per outcome fall over time. That is what putting AI inside X10 produces.
The curve
As AI takes on the mechanical work, hours per outcome fall every cycle while the revenue keeps coming.
| Outcome | Hours today | With AI | Direction |
|---|---|---|---|
| Cribl pipeline build | 3–6 h | 30–60 min | ↓ |
| Full Splunk/Cribl health assessment | ~40 h | < 4 h | ↓ |
| Splunk SVC reduction (Quest-style) | 96 h one-time | continuous; hours fall each cycle | ↓↓ |
These figures are estimates, drawn from internal numbers and the Quest/HEB proof points — not a guaranteed benchmark. We publish measured numbers after the first three assessments. Measuring real COGS (Cost of Goods Sold) is the highest-leverage near-term task, and it needs no customer.
Why hours fall (and don't just disappear)
First engagement: write down the method
The first time through, an engineer does the work with AI and captures the pattern. Hours are already lower than the manual baseline.
Next engagements: reuse the pattern
A saved pattern means each later customer takes a fraction of the first. HEB went from ~40 hours to a fraction once the method was worked out.
Subscription: the work becomes monitoring
Ongoing re-checks turn redoing the work into light-touch monitoring. The recurring revenue arrives with less and less labor behind it.
The shape of the business this creates
Recurring revenue
Subscriptions and renewals build on top of proven, delivered value.
Falling labor
Hours per outcome drop every cycle as patterns get saved.
Protected margin
Senior review stays the same. AI removes the grind, not the judgment.
The measured ROI baseline
The supervised agentic engineering baseline, measured against a 120-minute manual Cribl pack engineering process. Supervised means a person reviews and approves the AI's work (source: Supervised Agentic Engineering ROI Summary):
| Metric | Value | Meaning |
|---|---|---|
| Time reduction | 87.5% | The process takes 87.5% less time per run |
| Performance gain | 800% | The end-to-end process runs 8× faster |
| Throughput | 8× | Eight runs now fit in the original 120-minute window |
The pricing economics
The assessment is priced at 25% of realized annual savings — see the standing decisions. The unit economics that make the model work:
- Processing a GB at ingest time costs far less than indexing and storing that same GB later. So every GB you cut upstream saves real dollars, again and again.
- A 20% volume cut yields roughly a 10% SVC cut, and SVC is what the customer's Splunk bill is made of.
- The benchmark table turns a customer's source mix into a savings estimate you can defend. So the fee quotes itself from recorded engagement data.
The ROI baseline above and these ratios need further research and validation before they are client-facing — see the benchmarks page for the validation status of every underlying figure.
Honest guardrail
Margin is set by real engineer review time, false-positive handling, and support — not by hopeful figures. We track every hour on the first three jobs and report the true numbers. The story is "less labor over time," not "free."