Pricing that ships outcomes
Four ways to work with kraftbyte, ending where we believe all AI services end: paying for finished work. Every engagement includes human approval gates and full observability, because trust is part of the deliverable.
Workflows mapped across teams, automation candidates ranked by effort and impact, an ROI model and a sequenced roadmap you can execute with anyone. Credited against :ship// within 60 days.
One workflow taken prototype to production for real: human approval gates, observability and handover included. The FDE loop, applied once. Price your exact workflow in the estimator.
A senior crew embedded in your operation, shipping automations continuously and operating what they build. Forward deployed engineering as a subscription.
After a pilot establishes baselines, move to per-unit pricing: per document processed, ticket resolved, invoice reconciled. Pay for finished work, not hours.
Price your exact workflow
192 workflows rated for complexity, agent roles, timeline and tool stack. Pick yours, watch the estimate assemble, stress-test the payback with your own numbers.
Frequently asked questions
Why fixed-scope instead of hourly?
Hourly billing rewards slowness. A fixed-scope build makes the deliverable explicit before we start: the scope document names the workflow, the systems we integrate, the accuracy bar, the approval points and the handover artifacts. One number, one timeline, and the delivery risk sits with us, where it belongs.
How are the workflow estimates calculated?
Every use case carries a delivery rating: complexity on a five-point rubric, cooperating agent roles, weeks to production and the tool stack involved. Worked example: KYC compliance rates Complex (4 of 5), needs 3 agent roles across document processing, compliance controls and approval gates, and takes about 8 weeks. Those inputs produce the from-$37,500 estimate you see in the estimator. Same rubric, every workflow, so the number is explainable rather than plucked from the air.
What does outcome-based pricing look like, concretely?
Take invoice processing. A :ship// pilot runs for a few weeks and establishes the baselines: monthly volume, straight-through rate, exception rate. Pricing then flips to per finished unit. An illustrative shape: a fixed platform fee of $1,500 per month, plus $0.75 per invoice processed straight through, with exceptions that need human review billed at a lower rate because your team finishes those. If accuracy drops below the agreed bar, those units are free. You pay for completed work; the observability we build in is what makes every unit verifiable.
Who pays for the AI tokens and infrastructure?
You do, and you see all of it: workflows run in your cloud and your model accounts wherever possible, so there is no markup hiding in usage. Typical workflows land between $150 and $1,000 per month depending on volume and model mix, and we keep it low by routing routine steps to small models and reserving frontier models for judgment calls. The observability layer shows cost per completed unit, so run cost is a dashboard number, not a surprise. The estimator includes an adjustable run-cost line in its ROI calculator, and under outcome pricing it is baked into the per-unit fee.
What is included in every build?
Human approval gates on consequential steps, full observability (every input, decision and action logged), documentation and handover, and a 30-day stabilization window after go-live where fixes are on us. Trust is part of the deliverable, not an add-on.
What happens if scope grows mid-build?
The scope document is the contract. If we discover the workflow is genuinely bigger than rated, we tell you before building further, re-rate it against the same rubric and you decide: trim to the original scope or approve the delta. No silent overruns, no surprise invoices.
What if my workflow is not in the list?
The library covers 192 patterns but the list is where we start, not where we stop. Book a free readiness call and we will map your actual workflow, rate it against the same rubric and price it the same way.
Rather have a human price it with you? One call, no obligation, and you leave with a written AI readiness report either way.
Get your free AI readiness report