FINO by Nagent · For bank CIOs and CTOs
An AI money advisor for your customers, under your bank’s controls.
FINO is an AI money advisor a bank offers its customers under its own name: one face, as an avatar or as text, in front of specialist agents for budgets, investments and credit. Every money action is checked against the level the customer sets and the rules the bank sets, and an audit agent records every turn.
- One face in front of three specialists and an audit agent
- Five autonomy levels, checked when an action runs
- Commitments always wait for the customer’s confirmation
- Every turn, consent and hand-off on the record
- Every control on this page labelled with its status
What is FINO?
FINO by Nagent is an AI money advisor that a bank puts in front of its customers under its own brand. One face, avatar or text, routes each question to a Finance Manager, Wealth Advisor or Credit Advisor. Money actions are checked against the customer’s autonomy level when they run, and every turn is recorded for audit.
Last reviewed
what FINO does today
One advisor in your bank's name, under rules you set. Each part says how far it has got, so nothing here reads as further along than it is.
One face, four specialists
An avatar the customer talks to, or text only. Behind it a Finance Manager, a Wealth Advisor, a Credit Advisor, and an Audit & Compliance agent that never chats and records every turn.
An autonomy level each customer sets
Five levels, from the first, which only informs, to the fifth, which can turn on autopay, start SIPs and set up mandates within the customer’s limit. Checked when FINO acts, not only in what it says. Credit applications, balance transfers, one-time investments and redemptions ask first at every level.
Rules the bank controls
The highest level each customer segment may use, which money actions may run unattended and from which level, and what each agent may propose. Loosening a rule needs an emailed code; tightening takes effect at once. Every change is versioned.
Agents your bank builds
A bank builds its own agents in its private copy of Nagent and connects its systems through an OpenAPI description or an MCP server. Its agents propose; FINO shows the card and the customer taps. They see only the customer details they were granted.
Checks before an agent goes live
Agents move from sandbox to staging to live, with nine automated checks before each step. Only a bank admin puts an agent live, and the Activity tab keeps the record.
Your brand, your markets, your languages
Offered under the bank’s own name and look, for India and the US, with spoken replies in English, Hindi and other Indian languages.
how it works
One face, specialist agents, and a gate at the moment of action. The shape an architect will ask about: where a request goes, which tools each agent can reach, what is recorded and where the autonomy check sits.
- The customer talks to one advisor. As an avatar with a voice or as text. The avatar is optional, and the advisor is the same either way.
- The advisor routes the request. Each question goes to the specialist whose job it is: the Finance Manager, the Wealth Advisor or the Credit Advisor.
- Each specialist works through its own tool server. Budgets and bills, investments and credit each sit behind their own MCP (Model Context Protocol) server.
- Audit & Compliance records every turn. A fourth agent never chats. It records every turn, every consent and every hand-off between agents.
- Actions pass the autonomy gate when they run. Before a money action runs, the gate checks it against the customer’s level and the bank’s rules at that moment. Commitments wait for the customer’s confirmation.
security controls
Security your team can check. Concrete controls, each with an honest status: what runs in the live demo, what is built and tested on staging, what is built but off in the public demo, what is in build and what is on the roadmap.
In the live demo
Read-only access, with consent
FINO reads linked accounts only after the customer consents, and it only reads them.
A sample consent journey
Linking runs through a sample consent journey modelled on Account Aggregator: mobile number and one-time code, the consent details, approve, and revoke.
Five autonomy levels, checked at the moment of action
Each customer sets a level from one to five. FINO checks it when an action runs, not only in what it says.
Commitments always wait for a confirmation
A credit application, a balance transfer, a one-time investment or a redemption waits for the customer’s confirmation at every level. A spoken yes alone never confirms.
Every turn on the record
The Audit & Compliance agent records every turn, every consent and every hand-off between agents.
Text only, if you prefer
The avatar is optional. FINO works in text-only mode.
Built, tested on staging
Loosening a rule needs a code
In the bank’s console, raising a ceiling or an auto-pay limit waits for a one-time code emailed to the admin. Tightening takes effect at once. Every change is versioned and can be rolled back.
Your agents see only what they were granted
Agents a bank builds see only the customer details they were granted, and they can only propose. The customer taps to act.
Checks before an agent goes live
Agents move from sandbox to staging to live, with nine automated checks before each step. Only a bank admin puts one live.
Each money action on the record
Each money action is recorded with the gate’s decision and the outcome.
Built, off in the public demo
Customer vault
Each customer gets a random ID. Names and personal details are swapped for tokens before any text reaches a language model, and swapped back only in the reply to that customer.
PINs and one-time codes stay away from the model
A PIN or a one-time code is never sent to a language model.
Every vault lookup recorded
Each lookup in the vault is recorded.
In build
Connector layer for bank systems
Outbound calls go only to hosts on an allowlist, and secrets are stored as tokens.
Money calls that cannot pay twice
Each money call carries an idempotency key, so a retry cannot pay twice.
A tamper-evident audit log
Each audit entry is chained to the one before it by hash, so a changed or missing entry shows.
Roadmap
Mutual TLS and signed requests
Calls to bank APIs over mutual TLS, with each request signed.
Signed inbound webhooks
Events from bank systems accepted only with a valid signature.
Maker-checker and your SIEM
Maker-checker approvals for changes, and export of the audit trail to the bank’s SIEM.
An independent adversarial evaluation
An outside evaluation of FINO against MITRE ATLAS, the public knowledge base of attacks on AI systems.
deployment options
Hosted by us, in a private copy, or in your cloud. Where FINO runs today and where it is going, with the status of each option.
Hosted pilot
Nagent runs FINO and the bank’s console. FINO’s services run in Google Cloud’s Mumbai region.
Dedicated private copy
A single-tenant copy of the Nagent platform under the partner’s own brand, with its own database, its own region and a fixed outbound IP address the bank can allowlist.
In your cloud project
FINO and the private copy run inside the bank’s or the partner’s own cloud project, with the language model running inside that project too.
Reach internal systems without opening inbound ports
An on-premises relay, so agents can reach internal apps and databases without a port opened to the internet.
integrations
Bring your own agents and systems. Your agents propose and FINO shows the card; the customer taps. They see only the customer details they were granted.
Your own agents, through OpenAPI or MCP
A bank adds its own agents and connects its systems through an OpenAPI description or an MCP server.
Templates for Salesforce, MuleSoft and core systems
Ready-made connections for the CRM, the integration layer and the core banking system.
the autonomy ladder
Five levels, set per customer, checked when FINO acts. At every level, commitments wait for the customer’s confirmation: a credit application, a balance transfer, a one-time investment or a redemption.
- level 1ObserverExplains and informs. Never moves money.
- level 2AssistedSuggests and drafts, explaining each step. Every action waits for a confirm.
- level 3PractitionerPrepares the action, ready to approve in one tap.
- level 4SeniorPays known bills within the customer’s limit, then tells them. Anything unusual still asks.
- level 5PrincipalAlso turns on autopay and starts monthly investments within the limit, then tells them.
how a pilot starts
Four plain steps to a first pilot. A small group first, on sample or sandbox data, with the audit trail open from day one.
- Pick a segment and the agents. Choose the customers the pilot serves and which specialist agents they get.
- Set your brand and the autonomy levels. Put FINO under your name and look, and set the highest level the segment may use.
- Start on sample or sandbox data. Run it on sample data or your sandbox before any live account is linked.
- Go live with a small group. Open it to a small group of customers and read the audit trail together.
from the FINO blog
Start with the whitepaper. The market evidence, the failure modes and ten questions to put to any AI advisor vendor, with where FINO stands on each.
AI money advisors for banks
What banks and fintechs are deploying, what fails, and how FINO by Nagent puts one advisor in front of specialist agents, under an autonomy level each customer sets. With a ten-question checklist to put to any vendor, including us.
Questions CIOs and CTOs ask
Does customer data reach the language model, and is it used for training?
FINO calls hosted language models through their APIs and has no model of its own to train on customer conversations. In the public demo the conversation reaches the model as it is. The customer vault, built and off in the public demo, swaps names and personal details for tokens before any text reaches a model, and PINs and one-time codes never reach one.
Where does FINO run?
In the hosted pilot, Nagent runs FINO and the bank’s console, and FINO’s services run in Google Cloud’s Mumbai region. A dedicated private copy under a partner’s own brand, with its own database, region and fixed outbound IP address, is running for a partner. Running inside your own cloud project, with the language model there too, is on the roadmap.
What can FINO do without asking the customer?
That depends on the level the customer sets and the rules the bank sets. Levels one to three never move money on their own. At level four FINO pays known bills within the customer’s limit, then tells them; at level five it can also turn on autopay and start monthly investments within the limit. A credit application, a balance transfer, a one-time investment or a redemption always waits for the customer’s confirmation.
What does the audit trail record?
In the live demo, the Audit & Compliance agent records every turn, every consent and every hand-off between agents. In the bank’s console, built and tested on staging, each money action is recorded with the autonomy gate’s decision and the outcome, and every rule change is versioned. A tamper-evident audit log chained by hash is in build.
Can we add our own agents?
Yes. A bank builds agents in its private copy of Nagent and connects its systems through an OpenAPI description or an MCP server; this is built and tested on staging. Your agents see only the customer details they were granted and can only propose: FINO shows the card and the customer taps. Templates for Salesforce, MuleSoft and core systems are on the roadmap.
What happens if the avatar or voice vendor fails?
The avatar is a layer over the conversation, not the advisor itself. FINO works in text-only mode, so the same advisor, with the same autonomy gate and the same record, answers in text.
What is live today, and what is on the roadmap?
Every control and option on this page carries its status: in the live demo, built and tested on staging, built but off in the public demo, in build, or roadmap. Nothing here is further along than its label says. FINO has no live bank deployment to quote yet.
How does a pilot start?
In four steps: pick a customer segment and the specialist agents, set your brand and the highest autonomy level, start on sample or sandbox data, then open it to a small group of customers and read the audit trail together.
How should we evaluate FINO?
Open the live demo with a work email and try each autonomy level. Read the whitepaper, which ends with ten questions to put to any AI advisor vendor, with where FINO stands on each. Then book a demo and bring your security and risk teams: we walk through the controls on this page against your own checklist.
Last reviewed
