From professional interest to an agreed bank pilot
Turn one bank review question into an explicit test: the scope, usable sources, competent reviewers, evidence limits and acceptance criteria. Validate the workflow alongside existing controls before agreeing operational use.
The Brief is a private walkthrough of ten prepared model situations. Access is free for invited visitors using a valid shared audience code or personal invitation. No email registration is needed for a shared code. New access requests are reviewed manually and do not grant access or create a pilot commitment.
Two starting points
The Brief is a private walkthrough of ten prepared model situations. Access is free for invited visitors using a valid shared audience code or personal invitation. No email registration is needed for a shared code. New access requests are reviewed manually and do not grant access or create a pilot commitment.
Request access to The BriefUse an Evidence Readiness Check to discuss the case, available records, source suitability and the person responsible for review.
Discuss my review questionProduct maturity / complete history
The problem and responsible decision boundary were defined.
The intended Evidence Pack format and review grammar were made visible.
Separate controlled-case demonstrator in the final stage of development, with a live demonstration in October 2026; institutional validation remains to be completed.
Contracted institution-specific configuration and calibration after agreed gates.
Recurring assisted workflow within an institution after agreed acceptance.
Current product state: guided prototype, with a separate controlled-case demonstrator in the final stage of development, with a live demonstration in October 2026. Earlier completed stages are product history, not a bank onboarding sequence.
Joint work with the bank
We are designing the paid pilot as joint work between 3BrainAI and a bank team drawn from the relevant departments. It is intended to include configuration and calibration for that bank’s specific environment. The scope of the pilot and the specialist capacity required will depend on the tasks the solution is intended to support within the bank.
The Brief explains the workflow, its purpose and the intended outputs. The demonstrator is intended to show how that workflow can operate in detail with the departments involved on the bank’s side. The paid pilot is intended to cover configuration, calibration and evaluation for the specific bank environment.
Depending on the task, collaboration may involve property monitoring, risk or corporate banking teams, alongside IT, AI and security specialists. Data handling, access rights and integration with existing systems also need to be assessed. Any independent review required for bank-process controls and AI governance should have a separately agreed scope and clear responsibilities.
Value is to be assessed against agreed criteria, such as preparation time, manual effort, traceability, rework and costs. Time or cost savings are not assumed in advance. The focus is on supporting professional review.
Preparing the first professional review
We are preparing a professional review of CRI with representatives of one bank operating in Czechia. The first agreed meeting will focus on the usability of the complete workflow: do participants understand the steps and intended outputs, and how would the workflow fit the work of their departments?
The technical methods and their limits are to be explained as part of that discussion. A separate assessment of analytical correctness requires suitable reference data and defined criteria. We are continuing to refine the scope and conditions of the first review; details of participation are available on request, subject to confidentiality.
A separate institution-specific axis
Scope one review question, usable context and a responsible reviewer.
Test a bounded workflow with controlled institution context.
Configure, calibrate and evaluate against agreed criteria.
Move to recurring assisted use within the institution.
Each institution begins with its own question and agreed gate. The public chronology above does not restart for each bank.
Institution-specific engagement
| Stage | What we do together | Concrete output |
|---|---|---|
| 1 / Evidence Readiness Check | Clarify the professional question, scope, available inputs and owner. | Continue, refine or stop with a reason. |
| 2 / Bank demonstrator | Walk through the full process with bank reviewers: scope, baseline, suitability, calibration, a scoped reply and its review. | A defined workflow and bank input list. |
| 3 / Agreed preparation | Confirm rights, authoritative records, reference truth, reviewers and measurement conditions. | A bounded scope for configuration and calibration. |
| 4 / Contracted paid pilot | Configure, calibrate and evaluate an agreed sample alongside existing controls. | Measured results, limitations and an acceptance discussion. |
The Brief is a private walkthrough of ten prepared model situations. Access is free for invited visitors using a valid shared audience code or personal invitation. No email registration is needed for a shared code. New access requests are reviewed manually and do not grant access or create a pilot commitment.
The bank’s contribution
Identify the asset, phase or unit; supply the authoritative milestone record, version, date and responsible owner.
Identify available records and permitted uses. Agree access, retention and distribution for the selected engagement.
Nominate the process owner and competent reviewers. Agree what will count as a supported answer and who confirms it.
Define the current preparation and review process before measuring time, rework or usefulness.
Technical validation contract
Agree the evidence required for each gate before evaluating results.
| Gate | Evidence retained | If the basis is insufficient |
|---|---|---|
| Scope & baseline | Asset or phase identity, revision, question and responsible reviewer. | Refine the question before analysis. |
| Source suitability | Rights, provenance, dates, coverage, resolution and known exclusions. | Record limited evidence; request another source or stop the inference. |
| Candidate & calibration | Versioned analytical proposal, reason codes, uncertainty and agreed reference truth. | Withhold an unsupported conclusion; keep the review pending where a response is still needed. |
| Human acceptance | Competent reviewer, reviewed basis, accepted scope, exclusions and version. | Continue, refine or stop with an explicit reason. No automatic bank decision. |
What a PoC evaluates
Evidence readability
Can a reviewer quickly distinguish observation, bounded conclusion and uncertainty?
Review usefulness
Does the output help the reviewer identify the right next question?
Uncertainty handling
Are insufficient evidence, data limitations and interpretation boundaries visible?
Governance fit
Can the Evidence Pack fit the institution’s roles, controls and decision ownership?
A PoC evaluates workflow fit. It does not promise performance outcomes in advance.
Measurement before promises
| Question to measure | Evidence retained |
|---|---|
| Are the inputs available and suitable? | An inventory of required, available, missing and unsuitable sources with reasons. |
| Do scope, date and version match? | Recorded matches, mismatches and unresolved questions. |
| Is the next step clear and usable? | Reviewer feedback, ownership and fit with existing controls. |
| How does the work compare? | Preparation, active review, waiting and rework measured against the agreed baseline. |
Classification performance can be evaluated only where suitable reference truth exists. No time saving, accuracy or service level is assumed in advance.
Shadow-mode boundary
Existing review process – unchanged
CRI output reviewed in parallel
Configuration and calibration
Define the use case, responsible roles, review expectations and ownership.
Structure project context, milestones, risk windows and evidence cadence.
Agree the relevant data handling, security and configuration scope.
Run an agreed sample and evaluate the workflow against explicit gates.
Different engagements
A focused conversation or reference-case walkthrough to assess relevance.
A separately scoped exercise using a defined period and permitted historical records.
Institution-specific configuration, calibration and evaluation with agreed roles, criteria and terms.
Earlier study durations are not a standing commercial offer. Shadow mode means review alongside existing controls without affecting live bank decisions.
Product maturity
| Stage | Status |
|---|---|
| Idea and product concept | Completed foundations. |
| Illustrative Evidence Pack and prepared walkthrough | Available examples of the record and review logic. |
| Controlled-case Proof of Concept | Separate controlled-case demonstrator in the final stage of development, with a live demonstration in October 2026; institutional validation remains to be completed. |
| Contracted paid pilot | Next institution-specific gate, subject to agreement. |
| Recurring operational use | Target after relevant acceptance and operating conditions. |
Before the conversation
Banking professionals. An external technical adviser may participate within the bank’s process; the bank determines the roles and authority.
We make the limitation explicit, consider suitable documentary or other evidence, refine the scope or stop. Missing observation is not evidence of no change.
No. Access is free for invited visitors with a valid shared code or personal invitation. A new request is reviewed manually after you send it; preparing the email draft does not send it, issue a code or grant access.
No. Add the question you want to discuss yourself. Do not include confidential client information in the initial enquiry.
Professional request
Choose a private walkthrough or a discussion about a specific review question.
Already have an invitation code? Enter The Brief.
The Brief is a private walkthrough of ten prepared model situations. Access is free for invited visitors using a valid shared audience code or personal invitation. No email registration is needed for a shared code. New access requests are reviewed manually and do not grant access or create a pilot commitment.
Direct contact: cri@3brain.ai
3BrainAI