Assure
Discover to evidenceAI Product Readiness Review: one feature, reviewed before it ships
One bounded feature, four to six weeks. What your role is under the regimes that apply, how the feature fails and who that affects, what has to be disclosed, what a human must see before an output reaches a customer, and the release controls that make the decision to ship defensible afterwards.
- Published range
- $6,000–12,000 / engagement
- Timing after intake
- 4–6 weeks
- Quoted in India
- ₹3,00,000–6,00,000 / engagement
Fixed scope, not hourly. Quoted before we start, with the exclusions written down. The India figure is a separate price for a separate market, not a conversion.
How this is actually delivered
Discovery, assessment, design, implementation and the evidence register. Everything except the ongoing maintenance that follows it.
The scope is fixed before the work starts, which is why the price is a published range rather than an hourly estimate. Anything outside it is quoted separately rather than absorbed quietly or billed by surprise.
- Why companies buy this
- Features ship. Disclosures, evaluations and the decision about what a human must see before an output reaches a customer tend to be assembled afterwards, in a hurry, when a buyer or a regulator asks.
- What we actually do
- One bounded feature, over four to six weeks. We map your role for that feature under the applicable regimes and the reasoning behind the classification, work through foreseeable failures and who they affect, set the disclosure and transparency requirements, define human review proportionate to consequence and reversibility, design the permission and escalation model where the system can take actions, and build the release checklist and test evidence that goes with the decision to ship.
- Where it stops, and who takes over
- Any substantive legal opinion on classification in a named jurisdiction is scoped separately and assigned to qualified counsel. Technical assurance - evaluations, security testing, model validation - belongs to a qualified specialist, and we say so rather than implying we have done it.
- What you can show afterwards
- A signed release decision with its reasoning, test and evaluation records, the disclosure set as implemented, and the escalation and rollback procedure with an owner against it.
What you receive
One AI feature reviewed end to end: your role under the applicable regimes, how it fails, what has to be disclosed, and the release controls to put around it.
One bounded feature. Any substantive legal opinion on classification in a named jurisdiction is scoped and quoted separately with qualified counsel. Penetration testing, formal model validation and independent certification sit with the appropriate specialists and are not included.
- A role and applicability map for this feature, with the reasoning written out
- A foreseeable-failure review: what goes wrong, who it affects, how reversible it is
- Disclosure and transparency requirements as they apply to this feature
- Human-review design, set by consequence and reversibility rather than applied uniformly
- Permission and escalation design where the system can take actions outside a chat window
- A release review checklist with the test evidence that has to accompany it
- An incident, escalation and rollback procedure with named owners
- A signed release decision record, including accepted residual risk
What it costs, and what it replaces
Both figures are published ranges for the same unit of work. Ours is fixed before we start; if our process gets faster, that is our gain and your price does not move.
How engagements are structured →Published range for this engagement, against the span of the whole card. In India the same scope is quoted at ₹3,00,000–6,00,000 — a separate price for a separate market, not a conversion. Where in the range a quote lands is set by scope, entities and how much usable evidence already exists.
Midpoint $9,000 per engagement
How the work runs
01
Bound the feature
We agree exactly what is in scope. A review of 'our AI' is not a review; a review of one named feature, with its own purpose and users, is.
02
Role and applicability
Intended purpose, your role, territorial scope and the relevant exceptions. We do not assume every AI system is high risk or that every company is a provider - that classification is reasoned, recorded, and where it turns on law in a named jurisdiction, assigned to counsel.
03
Failure review
How the feature fails, who is affected, how visible the failure is and how reversible. This drives everything downstream, because review requirements should follow consequence rather than habit.
04
Controls and disclosures
What users must be told, what a human must see before an output reaches anyone, and what permissions and approvals apply where the system can act. We test the approval and rollback paths rather than describing them.
05
Release decision
The checklist, the test evidence and a signed decision with its reasoning and any accepted residual risk - so the question 'why did you ship this' has a documented answer.
What we need from you
- The feature itself: a walkthrough, access to a staging environment, or both
- Its intended purpose, the users it affects, and the markets it will be available in
- Model and vendor details, including versions and the configuration in use
- Any evaluation or testing already done, however informal
- The product owner and an engineer who can answer questions about behaviour
What we check before delivery
- The classification reasoning is written out and checked by a second reviewer, not asserted
- Legal conclusions link to dated primary sources or are escalated to qualified counsel
- Technical assurance claims are made only where a qualified specialist has done the work
- Controls recommended are tested in your environment before they are signed off
When firms send us this
- A customer-facing AI feature has a launch date
- A buyer asked what happens when the model is wrong
- An agent in your product can take actions with external consequences
- You are entering a market whose requirements you have not mapped for this feature
Questions about ai product readiness review
Can you certify the feature as compliant?
No. Nobody in an advisory role can, and certification against a standard is issued by an independent certification body, which we are not. What you get is a reasoned, documented position, the controls to support it, and the evidence that the decision was taken properly.
Will you test the model?
We review what testing exists, what it covered and what it did not, and specify what else is needed. Running evaluations or security testing is specialist technical work, and we bring in a qualified specialist rather than implying we have done it ourselves.
What if the review says we should not ship?
Then it says so, with the reasoning and the options. The point of commissioning this before launch is that the answer still has somewhere to go.
Firms who send us this usually also send
This engagement
Scope it in fifteen minutes.
AI Product Readiness Review at $6,000–12,000 per engagement, delivered in 4–6 weeks after a complete intake. Scoping costs nothing, and we will say if a smaller engagement would serve you better.