Does an agent build actually pay back?

▸ The arithmetic, in full

build        = (development + architecture + integration)
               × domainMultiplier × (regulated ? 1.25 : 1)
development  = 25,000 + 5,000 × (variants − 1)
architecture = (variants ≥ 2 OR regulated) ? 10,000 + 2,500 × (variants − 1) : 0
integration  = (systems ≥ 2) ? 15,000 + 8,000 × (systems − 2) : 0

baseline     = people × hours × 52 × rate
released     = baseline × A1
realized     = released × A2

inference    = volume × 12 × A3
platform     = A6
review       = volume × 12 × A4 × (A5 ÷ 60) × rate
runCost      = inference + platform + review
netAnnual    = realized − runCost

cashflow[m]  = m ≤ 4 ? −(build ÷ 4)
               : (netAnnual ÷ 12) × ramp(m − 4)
payback      = first m where Σ cashflow[1..m] ≥ 0
npv          = Σ cashflow[t] ÷ (1 + A8 ÷ 12)^t   for t = 0..35

aht          = (people × hours × 52 × 60) ÷ (volume × 12)

Build spend is spread evenly across the first 4 months — the mid-point of our published 12-20 weeks delivery window — and no benefit is allowed before it ends, which is what makes a payback shorter than the build structurally impossible rather than merely unlikely. Returns are computed off the modelled build, never off a figure clamped to the published band; the band $25,000 – $150,000 is used for engagement framing only.

▸Published—VelocityMind published pricing, /pricing, 26 Jul 2026

▸ The nine coefficients you can change

A1

Automatable share of in-scope hours

Default 55% · range 10% – 80%

The largest single lever in this model, which is exactly why it is a slider. Defaults come from the industry profile and reset when you change it. Bounded above by the published 60–70% technically-automatable ceiling, which is a ceiling and not our number.

▸Assumption—yours to set(The largest single lever in this model, which is exactly why it is a slider. Defaults come from the industry profile and reset when you change it. Bounded above by the published 60–70% technically-automatable ceiling, which is a ceiling and not our number.)

A2

Share of released hours converted to budget

Default 60% · range 20% – 100%

Released hours are capacity, not a guaranteed headcount reduction. This is your call on how much of that capacity turns into budget. Our 60% default is a modelling assumption with no published source — Forrester's TEI methodology recaptures 50% of hours saved in its own composite model, so ours is on the optimistic side of that.

▸Assumption—yours to set(Released hours are capacity, not a guaranteed headcount reduction. This is your call on how much of that capacity turns into budget. Our 60% default is a modelling assumption with no published source — Forrester's TEI methodology recaptures 50% of hours saved in its own composite model, so ours is on the optimistic side of that.)

A3

Cost per item processed by the agent

Default $0.12 · range $0.01 – $2.00

An assumption unless you bring it from the Architecture Configurator, which derives it bottom-up from tokens and orchestration pattern.

▸Assumption—yours to set(An assumption unless you bring it from the Architecture Configurator, which derives it bottom-up from tokens and orchestration pattern.)

A4

Exception rate

Default 15% · range 2% – 50%

For invoice processing, best-in-class is 9% against 22% typical (Ardent Partners, ePayables 2024). That is shown as context. Our 15% default is an assumption sitting between the two, and it is yours to set.

▸Assumption—yours to set(For invoice processing, best-in-class is 9% against 22% typical (Ardent Partners, ePayables 2024). That is shown as context. Our 15% default is an assumption sitting between the two, and it is yours to set.)

A5

Review minutes per exception

Default 4 min · range 1 min – 20 min

A modelling assumption. If you time it, use your number.

▸Assumption—yours to set(A modelling assumption. If you time it, use your number.)

A6

Platform and observability, annual

Default $18,000 · range $6,000 – $60,000

Gateway, vector store, tracing, evaluation runs and on-call. A modelling assumption.

▸Assumption—yours to set(Gateway, vector store, tracing, evaluation runs and on-call. A modelling assumption.)

A8

Discount rate for NPV

Default 10% · range 0% – 20%

The de facto convention, matching Forrester's TEI 10% and three-year NPV treatment.

▸Assumption—yours to set(The de facto convention, matching Forrester's TEI 10% and three-year NPV treatment.)

A9

Regulated evidence pack, build multiplier

Default ×1.25 · range fixed — not adjustable

A modelling assumption for validation documentation, change control and the reproducible audit trace a regulated deployment requires.

▸Assumption—yours to set(A modelling assumption for validation documentation, change control and the reproducible audit trace a regulated deployment requires.)

A7

Benefit ramp after go-live

Fast (3 months to full) — 50%, 80%, 100% · Standard (7 months to full) — 40%, 40%, 40%, 70%, 70%, 70%, 100% · Slow (12 months to full) — 25%, 25%, 25%, 25%, 50%, 50%, 50%, 50%, 75%, 75%, 75%, 75%, 100%

The last value in each schedule repeats for the rest of the horizon. Nothing reaches full value on day one, and pretending otherwise is how a model produces a payback that never happens.

▸Assumption—yours to set

▸ Thresholds and cross-checks

• Negative verdict — net annual benefit at or below zero. The figures still render, as negatives.
• Slow verdict — payback beyond 24 months, or no crossing inside 36 months. The crossover headcount is solved live for your settings rather than asserted.
• Implausible verdict — first-year return above 400%. We print the threshold instead of a four-figure percentage and tell you to re-check the scope.
• Handling-time cross-check — the model divides your headcount and hours by your volume and warns when the result falls outside 0.5–240 minutes per item, because that usually means the two inputs describe different processes.
• Risk adjustment — Forrester's TEI methodology risk-adjusts benefits down 10–15% and costs up 5–10%. The conservative column uses the mid-points of those two ranges. A calculator returning one to-the-dollar figure from seven guesses is not being more precise than this; it is being less honest.
• The automatable ceiling — McKinsey Global Institute puts 60–70% of work hours in scope as technically automatable with generative AI. We use that as a ceiling on this slider, never as the number itself.

▸ Sources

1. Forrester Total Economic Impact methodology

   Risk-adjusting benefits down 10–15% and costs up 5–10%; 50% recapture of hours saved; the 10% discount rate and three-year NPV convention. Cited by name as the basis for the conservative band.

2. McKinsey Global Institute, 2023

   60–70% of work hours technically automatable with generative AI. Used strictly as a CEILING on the automatable-share assumption, never as the assumption itself.

3. Ardent Partners, ePayables 2024

   9% best-in-class versus 22% typical invoice exception rate, shown as context beside the exception-rate slider and explicitly not as our default.

   No public link to the primary document — treat this as a reference point to verify with the publisher, not as our figure.

4. VelocityMind published pricing

   Agent strategy $5K–$15K, multi-agent architecture design $10K–$30K, agent development and deployment $25K–$75K, enterprise integration $15K–$50K, and the published $25,000–$150,000 / 12–20 week implementation band.

Every weight and formula on this page is also published on the methodology page, rendered from the same module this calculator runs on, so the disclosure cannot drift from the code. Prices come from our published price list.