Banking and finance

Your bank’s own AI.

Forecasting, operations triage and KYC review, on a model the bank owns and runs inside its own walls.

StatusWhere our first model was built

A long trading floor at night, two rows of dark screens under one line of light
Eigna TLM ForecastRuns local
What cash demand should the branch network expect next week, by region?

Demand rises in the north and on the coast ahead of month end 1. The band for the coast is wider than usual, so a person should check it 2.

  1. 1Branch cash history, 2019 to 2026
  2. 2Treasury forecast, approved 20 October

Illustrative

Tasks to start with. A Test Lab takes one.
3
The Formula it starts withWith OLM and RLM
TLM
To roll back to the last release, by design
60 s
Eigna Local. Data never leaves.
Local
An analyst at her desk in the evening, two screens in front of her

The work that cannot leave

Client data, transactions and internal policies sit under strict rules. In a 2026 IDC survey of large enterprises, banks showed the highest concern about data leaks of any industry.

  • Client and account data
  • Transactions
  • Credit and KYC files
  • Internal policies and models

Where it runs, for banking

Eigna Local

Data never leaves.

On the bank’s own servers or an Eigna Box on site.

Training runs on a GPU we rent in a region you approve: an encrypted snapshot goes there for each run and is deleted after it. Where nothing may leave your network, it runs on your own GPU.

The first expert

Three tasks to start with.

Approved reviews, decisions and corrections from the bank’s own teams. A correction is not training data until someone approves it.

Illustrative questions and answers
01

Forecasting

Time series models trained on the bank’s own history, tested against its own baselines before any release.

Eigna TLM ForecastRuns local
What cash demand should the branch network expect next week, by region?

Demand rises in the north and on the coast ahead of month end 1. The band for the coast is wider than usual, so a person should check it 2.

  1. 1Branch cash history, 2019 to 2026
  2. 2Treasury forecast, approved 20 October

Learns from · Forecasts the treasury team signed off, and the bank’s own history.

02

Operations triage

Sorts requests and exceptions to the right team, and says “no call” when it is unsure.

Eigna OLM TriageRuns local
Where does this payment exception go, and is it urgent?

Trade finance operations. The beneficiary name does not match the account 1, and today is the cut-off for this currency 2. Urgent.

  1. 1Exception playbook, section 2
  2. 2Currency cut-off table

Learns from · Exceptions your operations team routed and closed.

03

KYC review

Reads identity and company documents, finds what is missing and flags what needs a person.

Eigna RLM ReviewRuns local
Is this company file complete for onboarding?

Two items are missing: proof of address for one director 1 and the ownership chart of the parent company 2. No call on the source of funds: it needs a person.

  1. 1KYC policy, section 4
  2. 2Onboarding checklist, corporate clients

Learns from · Reviews your onboarding analysts approved.

The proof we will measure

A metric, a baseline and a sealed test, agreed before we start.

We measureForecast error, and reviews approved without rework
AgainstYour current model and today’s process
OnHeld-out weeks in a sealed test set

The target is set with you before the Test Lab starts. If the expert does not beat your baseline, you keep the report.

17.8%

Our first training run cut forecast error 17.8% against its base model. It is not yet ahead of a simple last value baseline, and we publish that too.

Rules in plain words

What the rules ask, and what every release gives you.

  1. 01

    Know who changed what

    Model risk guidance such as the Federal Reserve’s SR 11-7 asks for validation, monitoring and change control. Every release leaves that record.

  2. 02

    Dated duties

    The EU AI Act’s high-risk duties apply from 2 December 2027 to the uses listed in its Annex III.

  3. 03

    One accountable owner

    A named owner, a signed Model Deed and rollback in 60 seconds. Designed to support your compliance work, not to replace it.

Banking and finance

One accountable owner.

A named owner, a signed Model Deed, a release record and rollback in 60 seconds. Designed to support your compliance work.

The owner’s Key standing upright in its reader on dark stone

Banking: questions and answers

Where is the model trained?

By default on a GPU we rent in a region you approve: an encrypted snapshot goes there for each run and is deleted after it. Where nothing may leave your network, training runs on your own GPU.

Can it sit inside our model risk process?

Yes. Every candidate is tested on a sealed set it never saw, against the version you use today, and the release record shows what changed, the test result and who approved it.

Which task should we start with?

One task with a clear owner, data you already hold and a number to beat. Forecasting, operations triage and KYC review are the usual first three.

Test Lab

Request a Test Lab for banking.

One task, your data, a sealed setup, about four weeks. If the expert beats your baseline, the test set becomes your first release gate. If not, you keep the report.

  • RouteOwn sales and partners in every market. We start with one team, one task and a Test Lab.
  • Watch forSecurity reviews take time. A Test Lab on one task keeps the first step small.
  • ReplyWithin two working days, from a founder.
Which task would you bring?

We use your details only to answer you. No newsletter unless you ask for it.