Cantilever-Engineering.AI
Book a demo
PlatformSolutionsIntegrationsPackagesResourcesCompany
Get started
Book a demo Contact us Architecture review
Trust
Security overview Data handling
Book a demo
Value model

Put your own numbers in. See what the Cantilever catalog gives back.

A parametric model, not a benchmark. Enter the provisioning, change, incident and audit volumes you already run, adjust every assumption behind them, and watch capacity, cost and cycle time recalculate as you type.

engineer-hours returned each year
FTE-equivalents released to delivery
gross value at your blended rate
realisable value after the haircut*

* Realisation accounts for ramp, imperfect capture, and capacity reinvested in delivery rather than removed from budget. Methodology is set out under the questions evaluators actually ask.

The model

Your volumes. Your assumptions. Your number.

Pull one quarter of ticket, change and incident data and multiply by four. Rough is fine — this is about order of magnitude, not decimal places.

01Annual volumes

What your organisation does in a year. These are the only figures that have to come from you.

02Hours returned per event

The hours leaving your process, set to platform reference values. Override any of them with a number your own team will defend in a room. Changed fields are outlined in brass.

03Adoption and realisation

Two discounts, kept separate on purpose. Adoption is how much work actually flows through the catalog. Realisation is how much freed capacity a finance team will recognise as value.

Share of eligible events executed through the catalog rather than a ticket or a console. Leave at 100% for full potential; set your year-one target for the near-term case.

Ramp, imperfect capture, and capacity reinvested rather than removed. Anything above 80% will not survive a finance review.

Salary, benefits, overhead and contractor mix across the engineers whose time this frees.

Drop to 1,700–1,800 for FTE equivalents net of leave, training and meetings.

Capacity returned each year
engineer-hours, after adoption coverage is applied
FTE-equivalents
Gross value
Realisable value
Where the hours come from
Stream by stream
StreamVolumeHrs eachHours backValue
Total

Value column is gross, before the realisation haircut.

Run trace

The same model, as Cantilever would report it.

Every governed run on the platform emits an evidence trace. So does this model — the difference being that this one is an estimate, and says so on its last line.

value model · roi-estimate
Why the number is this big

Three costs that never appear on an invoice.

Nothing in this model comes from new capability. It comes from removing work that exists only because the toolchain does not answer to a single policy.

01

Work that waits

A standard environment takes three to ten business days, most of it queueing behind a platform engineer who has to translate a ticket into engineering work. Catalog self-service removes the translation step, not the engineering.

02

Work that repeats

The same Day-2 change, the same triage path, the same new service pattern built from scratch by a different team. Reuse compounds: the marginal cost of the next automation falls every time one is published.

03

Work that reconstructs

Audit evidence assembled after the fact from logs, threads and retroactively completed change tickets. When lineage is a by-product of execution, weeks of senior engineering time stop being spent on archaeology.

By role

The same number reads differently depending on what you own.

Capacity persuades engineering leadership. Lead time and provable control persuade everyone else in the approval chain.

CISO

Self-service your team can actually approve

The audit line in the model is the smallest number on the page and the easiest one to defend. Roughly 400 hours a year stop being spent reconstructing evidence, because lineage is produced as a by-product of execution rather than assembled afterwards.

What holds the estimate up
  • Identity federated through your existing provider on every run
  • Just-in-time credentials — no standing access left in pipelines
  • Your standards enforced as a gate, before anything changes
  • Permanent lineage mapped to SOC 2, FedRAMP and ISO 27001 evidence
Cycle time

The argument that usually lands harder than hours.

These targets follow from the same design as the capacity model, and they are the ones worth writing into a pilot. They are also the ones a business stakeholder feels without being shown a spreadsheet.

Standard environment lead time3–10 days → under 1 day
New capability time-to-catalog3–6 weeks → 3–5 days
Access grant1–3 days → minutes
“What do we run, and who owns it?”2–5 days → minutes
Audit evidence package2–4 weeks → 1–2 days
Mean time to restore4–8 hrs → 2–4 hrs
Questions

The things evaluators actually ask about the model.

Are these numbers measured or estimated?
Estimated. This is a parametric model: it turns your volumes and your assumptions into a defensible figure, and it does not produce evidence. Treat every output as a hypothesis to test in a pilot, and say so when you present it. Any vendor quoting you a percentage without showing the arithmetic behind it should be asked the same question.
Why are adoption and realisation two separate discounts?
Because they fail for different reasons and get fixed by different people. Adoption is a product and change-management problem: how much work actually flows through the catalog. Realisation is a finance question: how much freed capacity turns into recognised value rather than being reinvested in delivery. Blending them into one number hides which one is hurting you.
Does this double-count anything?
It should not. Each stream counts a distinct class of work, and an hour saved appears in exactly one line. The one place to be careful is incidents: if your change-induced incidents fall because of policy gating, that shows up as fewer events in your own volume input, not as a larger per-incident saving. Reduce the volume rather than inflating the rate.
Why is risk-avoidance value excluded?
Fewer incidents, smaller blast radius and the removal of standing credentials are real value — they are simply not credible when priced from an industry average. Model them separately as incidents avoided multiplied by your own incident cost, drawn from your own history. An industry breach-cost figure in a business case invites an argument you will lose.
Which default should I replace first?
Hours per provisioning request, then hours per standard change. Together they carry roughly 80% of the modelled total, so the whole estimate moves with them. The audit and tag-remediation lines are small enough that arguing about them is not worth the meeting time.
What should we govern the programme with?
Adoption coverage — the share of infrastructure and Day-2 change executed through the catalog rather than through a ticket or a console. Cantilever instruments it, and it belongs in your pilot success criteria. If it climbs, everything in this model follows. If it does not, nothing here does.
Take it with you

Send me this model

We will email the numbers exactly as you have set them, alongside the assumptions sheet your finance team will ask for. No drip sequence — one email, from a person.

Next step

Bring your volumes. We will pressure-test the model with you.

One quarter of provisioning, change and incident data is enough. In thirty minutes we will populate this with your numbers, agree the adoption metric that will govern the programme, and scope a pilot that settles it either way.

30 minutes · tailored to your stack · no slideware