Catch the rule an answer breaks before your client does.
AFA Assurance judges one already-written advice answer against a configured legislation and ethics matrix. The verdict comes back as a status with findings that cite the matrix, and the sources the check used.
Pilot. It runs, and it is not yet a self-serve product.
Where it runs today: your team reads it in the Emable workspace, signed in and your own systems call it over the REST API.
What changes
Who inside your organisation is better off, and what they can show for it.
Compliance Lead at an adviser network
Sampling after the fact becomes a pass on every answer before it is sent. Each finding names the matrix entry it came from, so a review note is a citation, not an opinion.
the emable capability catalogue, afa-assurance: validate_answer returns a status plus grounded findings citing the matrix, and the declared source is the legislation and ethics assurance matrix configured in the AFA validator.
AI Product Lead at an insurer
You can put your own assistant in front of advisers without hoping it behaves. The check is one REST call inside the agent loop, made before the reply leaves your stack.
It runs in the Emable workspace and over REST. The same tool is namespaced validate_answer on the Emable bundle connector, scope afa:validate.
Head of Advice at a distributor
An adviser sees a status and the passage the answer was judged against, before using it with a client. A flagged draft is a first-class outcome, not a silent send.
The verdict returns a status plus findings that cite the matrix entry they came from. Pilot, Czechia.
Head of Internal Audit at a bank
The judgement is recorded at the moment the answer was made, together with the sources the matrix used, instead of being reconstructed from logs months later.
Context: the AI Act enters general application on 2 August 2026 and DORA has been in application since 17 January 2025, reliability B, per /answers/what-is-ai-assurance-for-financial-advice.
Two ways to get it
Have it delivered, or let your own AI call it.
As a service
We deliver the assurance step for you. We configure the legislation and ethics matrix for your product lines, wire the check into the point where your answers leave, whether that is an assistant, an adviser desktop or a review queue, and agree what happens to a flagged answer before anything goes live. We hand over the configured matrix, the wired check and the verdict format your compliance team reads. This is a pilot, so scope is agreed first and nothing runs unattended. Write to us with the channel your answers leave through, the product lines they cover, and who signs off a flagged draft today.
Tell us what you needOne message, no form. We answer with what it would take.
Through your own AI
Your own assistant can call the check directly. Connect the Emable bundle connector at https://api.emable.ai/mcp-emable with the afa:validate scope, and the tool validate_answer appears in your client. Your assistant then submits the client question, the answer it or an adviser wrote, and whether to check legislation, ethics or both. It gets back a status and findings that cite the matrix, with the sources the check used. It judges an answer that already exists and never drafts one. Pilot, and the only configured coverage is the Czech matrix, so results carry that label. There is no self-serve run on this page.
- Connector
- https://api.emable.ai/mcp
- Tools it adds
validate_answer
What your AI can answer
Connect it once, and your assistant answers these.
Questions in your own words, each one landing on a published tool. Ask us and we answer the same ones for you.
- 1Our assistant just drafted this reply to a client. Does it break a rule before it goes out?
validate_answer - 2This answer is legally correct. Is it acceptable on ethics as well?
validate_answer - 3A client with no investment experience asked about a structured product. Does the answer we wrote hold up?
validate_answer - 4Check this one against legislation only. Which rule does it fail?
validate_answer - 5What exactly was flagged in this answer, and which passage of the matrix says so?
validate_answer - 6We rewrote the flagged sentence. Does the answer come back clean now?
validate_answer - 7Judge these three wordings of the same answer and tell me which one passes?
validate_answer - 8Which matrix and which sources was this answer judged against, so I can put that in the review record?
validate_answer
AFA Assurance answers these through the tools above. A question outside them is refused, not guessed.
What it rests on
The evidence, and the edges.
Evidence
- Status pilot, exposure public, provider emable, country CZ, one tool named validate_answer. Source: the emable capability catalogue.
- Grounded in the legislation and ethics assurance matrix configured in the AFA validator. That is the only declared source.
- Three check modes, from the input schema: legislation, ethics, full. Question up to 5,000 characters, answer up to 20,000.
- Charging unit: one answer check within the account quota. Zero credits at the connector rail, because the account AI-job quota meters it.
- Callable from an AI client today as validate_answer on https://api.emable.ai/mcp-emable, Emable OAuth, scope afa:validate.
- Verified usage: Not enough verified runs yet. Public statistics appear only after 20 verified runs across 5 different accounts.
- Regulatory context, reliability B, from /answers/what-is-ai-assurance-for-financial-advice: the AI Act enters general application on 2 August 2026, DORA in application since 17 January 2025.
The response cites the assurance sources used by the configured matrix.
What it does not do
- Coverage is limited to the configured Czech assurance matrix.
- A passing check does not constitute legal advice or regulatory approval.
- It judges an answer someone else already wrote. It never drafts one.
- A check cannot span a market outside the matrix configured for the Czech market.
- It does not certify an adviser and it does not approve a product.
- Pilot, not live. The matrix and the citation format stay under test until they are confirmed outside the pilot cohort.
- No self-serve run on this page. A trial is one run, and it opens through us or through a connected client.
- Judging one answer says nothing about whether the advice suits a client, because suitability needs the client's facts.
Verified usage
Not enough verified runs yet
Public statistics appear only after 20 verified runs across 5 different accounts.
The contract, for the people who will wire it
Status, permissions, tools, schemas and where it runs. The same values the API returns, so nothing here can drift from what your client will actually get.
Status, access and price
- Status
- Pilotpublic
- Provider
- emable
- Manifest version
- v1.0.0
- Countries
- CZ
- Data class
- derived
- Access
- read
- Permissions
account:readafa:validate
- Charging unit
- One answer check within the account quotaPricing model: included. No separate charge per run.Credits are the platform's metering unit, not a price. We quote the engagement.
Tools and schemas
Validate answer
validate_answer
Check a financial-advice answer against the AFA legislation and ethics matrix so you can catch a gap before the answer reaches a client.
Use it when
- An advisor-facing answer already exists and needs a compliance-style grounding pass before it is shown or repeated.
- The user asks whether an answer holds up against legislation, ethics, or both.
Do not use it when
- No answer exists yet — this tool judges an existing answer, it does not generate one.
- The check needs to span a market outside the assurance matrix currently configured for the Czech market.
Returns
A status plus grounded findings.
Arguments
| Name | Type | Required | Description |
|---|---|---|---|
| question | string | yes | The client question the answer responds to. |
| answer | string | yes | The AI or advisor answer to judge. |
| check | string | yes | Which assurance matrix to judge against. |
Example: Check an answer against both matrices
{
"question": "Can I recommend this structured product to a retail client with no prior investment experience?",
"answer": "Yes, as long as the product's risk category is disclosed.",
"check": "full"
}{"status": "flagged", "findings": [{"matrix": "ethics", "citation": "..."}]}
Errors a caller must handle
invalid_api_keyThe credential is absent, invalid, expired, or revoked.Stop and reconnect with a valid key or OAuth grant.missing_scopeThe identity lacks the `afa:validate` scope this tool requires.Request `afa:validate` and reconnect.422Input does not match this tool's schema.Correct the input using the schema returned with this tool.capability_runs_disabledExecution is temporarily off for this capability.Keep discovery available and retry later; this is not a permanent failure.
Input schema
{
"type": "object",
"properties": {
"question": {
"type": "string",
"minLength": 3,
"maxLength": 5000,
"description": "The client question the answer responds to."
},
"answer": {
"type": "string",
"minLength": 1,
"maxLength": 20000,
"description": "The AI or advisor answer to judge."
},
"check": {
"type": "string",
"enum": [
"legislation",
"ethics",
"full"
],
"description": "Which assurance matrix to judge against."
}
},
"required": [
"question",
"answer",
"check"
],
"additionalProperties": false
}Output schema
{
"type": "object",
"properties": {
"status": {
"type": "string"
},
"findings": {
"type": "array",
"items": {
"type": "object"
}
},
"sources": {
"type": "array",
"items": {
"type": "object"
}
}
},
"required": [
"status",
"findings"
],
"additionalProperties": false
}Surfaces and connectors
No standalone MCP URL is deployed for this capability yet.
Specialist workspace, sign in requiredSources, limits and support
Sources
The legislation and ethics assurance matrix configured in the AFA validator.
Limitations
Coverage is limited to the configured Czech assurance matrix.
A passing check does not constitute legal advice or regulatory approval.
Dependencies
- Configured AFA legislation and ethics assurance matrix