Controlled service layer

Automation with the controls your compliance team needs

Handle routine service questions and onboarding steps, with strict boundaries, audit logs and human review where required.

Built for NBFCs, wealth ops teams and fintech service desks.

Scope this with us

The problem

Customers want fast answers, but the space is regulated. Anything resembling advice, or touching sensitive data, needs careful handling and a clear trail.

In practice

A customer asks how to update their address and what documents are needed. The AI explains the process, lists the documents, starts the request, and routes it to an officer for verification.

Never

Advice

Informational and admin only — written down before go-live

Logged

Every turn

Retention and region your policy already demands

Gated

Sensitive actions

Identity check + human where required

Faster

Routine service

How-to, status, documents — without a 48-hour ticket

The shift

From a risky chatbot to a desk compliance can sign

From this

  • ✕A vendor bot that will cheerfully ‘recommend’ a product.
  • ✕KYC status questions sitting in email.
  • ✕Address changes started on a call and finished on a form nobody finds.
  • ✕No trail when internal audit asks who said what.
  • ✕After-hours customers waiting until the branch opens.
  • ✕PII in a shared inbox.

To this

  • ✓A published list of what the layer may say — and what it must never say.
  • ✓Status and how-to from the systems of record.
  • ✓Requests started in-thread, approved by an officer.
  • ✓An audit log with retention you already owe the regulator.
  • ✓After-hours answers that stay inside policy.
  • ✓PII handled with the same controls as the rest of the stack.

What's included

Service, not a robo-advisor

Policy-bounded Q&A

How to, where to upload, what happens next — from your circulars, not from a model’s idea of a product.

Application status

Where the KYC or loan file actually sits. No invented timelines.

Document collection

The list, the format, the portal link. Completeness checks; no ‘that should be fine’.

Identity step-up

Sensitive actions wait on the check your stack already uses.

Officer queue

Address change, nomination, mandate — started in chat, finished by a human.

Audit log

Who saw what, who approved what, how long you keep it.

Channel control

App, WhatsApp, web — same boundary, same log.

Kill switch

Compliance can pause a topic without waiting for a vendor’s next release.

Every role connected

Built around how each desk actually works

For compliance

The boundary is a document, not a vibe.

  • Never advice
  • Logs you can produce
  • A kill switch

For the service desk

Routine how-to leaves the queue. Exceptions arrive with the file open.

  • Fewer ‘how do I update’ tickets
  • Officer work is still officer work

Go-live

Implementation without a year-long project

  1. 1

    Week 1

    Boundary workshop

    Compliance, legal, ops: the list of allowed intents. Written down.

  2. 2

    Week 2

    Systems of record

    Status APIs, document lists, the identity step you already have.

  3. 3

    Week 3

    Closed pilot

    Staffed users, full logs, a kill switch tested.

  4. 4

    Week 4+

    Expand intents

    Only after the first set has a clean audit week.

Enterprise-grade security

Controls your compliance team can live with

No advice

Product recommendation, suitability, returns — out of bounds. The model does not get a sales target.

Least privilege

It sees the fields the task needs. Not the whole core.

Retention

Logs live where your policy already says they live.

Human on the bind

Anything that changes a customer record waits for an officer.

Ideal for

Service desks that need speed without a rogue advisor

Deployment

Your cloud and region, on the core and CRM you already run

Commercials

A controlled build — compliance is in the scope, not an add-on

How we approach it

  1. 1

    We scope the AI tightly to informational and administrative tasks — never advice.

  2. 2

    Sensitive actions require identity checks and, where needed, human approval.

  3. 3

    Every interaction is logged with the retention and region controls your policy demands.

  4. 4

    We work with your compliance team on the boundaries before anything ships.

Scope this for your team

FAQ

Questions about financial services

Will this give investment or product advice?

No. That is a hard boundary. Informational and administrative tasks only, signed off before go-live.

Can we produce the conversation for audit?

Yes. Logging, retention and region are part of the build, not a nice-to-have.

What if a customer asks anyway?

The layer refuses, points at a human or a registered channel, and logs the turn.

Want faster service that compliance can still sign?

Bring the policy and the five questions that clog the desk. You leave with a written boundary and a plan for the first intents.

We’ll tell you if a platform already fits better.