Skip to content

For fraud, financial crime, lending and wealth operations leaders

Work the alert queue with every step logged

Ask out loud for a fraud alert, a KYC case, a loan file or an advisor's out-of-balance households. Tiersel pulls the records, shows every system it checked and the evidence it found, and waits for an analyst to approve the freeze, the onboarding, the decision or the proposal.

Live demo on synthetic data · Built on Jev by TypeSafe AI

An animated example types requests as if they were spoken and shows the card Tiersel builds for each one.

For the Head of Operations, Chief Risk Officer or Head of Fraud and Financial Crime at a bank, credit union, fintech or wealth firm

Your analysts spend the day gathering, not deciding.

A fraud alert means opening the card platform, the device log, the customer profile and the transaction history before anyone can judge it. A KYC case means matching documents to a registry, clearing screening hits one field at a time and tracing owners through holding companies. A loan file means re-keying income and checking ratios against policy. Most alerts turn out to be nothing, but each one still takes the same trip through five systems, and the audit trail is whatever the analyst remembered to write down.

“Only one or two transaction-monitoring alerts per hundred is typically acted upon.”
Adrian Murphy, Kate Robu and Matthew Steinert, McKinsey & Company, McKinsey & Company, The investigator-centered approach to financial crime, 2020
$15.9Blost to fraud by US consumers in 2025, from 3 million fraud reports made to the FTC.Federal Trade Commission, testimony to the Joint Economic Committee, 2026$33.41Bin payment card fraud losses worldwide in 2024. The US had 26% of card volume and 42% of the losses.The Nilson Report, press release, 2026$61Ba year is what financial crime compliance costs financial institutions in the US and Canada.LexisNexis Risk Solutions, True Cost of Financial Crime Compliance Study, US and Canada, 202490%+of transaction-monitoring alerts turn out to be false positives at most banks.McKinsey & Company, The investigator-centered approach to financial crime, 2020

4 agent workflows

Built for the work your team already does.

Each workflow starts from a sentence. The agent calls your systems, shows its evidence, and waits for a person before anything consequential happens. Every card below is the real, working interface.

Workflow 01 · Fraud analyst

Fraud alert review

Fraud alert reviewFinancial services

  1. 01 · Trigger

    Alert lands in the queue

    The fraud model flags an authorization, or an analyst asks for the next alert on a card or account.

  2. 02 · Tool call

    Pull the authorization and history

    Reads the flagged transaction, the model score and 90 days of card activity from the card platform and core banking.

  3. 03 · Tool call

    Check device, location and links

    Looks up the device fingerprint, the IP location against the last card-present purchase, and other customers that share the device or payee.

  4. 04 · Reasoning

    Weigh the signals

    Shows each signal with its weight, such as impossible travel, a new device, rapid retries or an amount far above the customer's pattern, and recommends freeze, step-up or release under the card fraud policy.

  5. 05 · Human approval

    Analyst decides

    The analyst picks freeze, step-up verification or release and approves it. Linked cards and payees are included only if the analyst switches them on.

    Nothing happens until a person approves

  6. 06 · Outcome

    A decision with its evidence

    The card is frozen or released, the customer is notified, and the case record holds every signal, source and approver.

What changes

Each alert reaches a decision from one screen, and the case file records why without the analyst writing it up.

Card processing and fraud scoringCore bankingDevice intelligenceCase management

Workflow 02 · KYC or onboarding analyst

KYC and KYB onboarding review

KYC reviewFinancial services

  1. 01 · Trigger

    New account application

    A business or individual applies, or an analyst opens an onboarding case by name.

  2. 02 · Tool call

    Verify documents

    Checks formation documents against the state registry, the EIN, government IDs with liveness, and the age of proof of address.

  3. 03 · Tool call

    Screen people and entities

    Screens the entity, its owners and control persons against sanctions, PEP and adverse media sources.

  4. 04 · Reasoning

    Disposition hits and rate risk

    Compares each hit field by field and explains the call, such as a false positive because the date of birth differs. Traces ownership through holding companies to find owners at or above 25% and sets a risk rating.

  5. 05 · Human approval

    Analyst approves or requests info

    The analyst approves onboarding or chooses which missing items to request. Nothing reaches the customer until the analyst approves.

    Nothing happens until a person approves

  6. 06 · Outcome

    An onboarding file ready for review

    The account opens with a documented risk rating, or the request goes out with exactly what is missing.

What changes

Screening hits arrive with a field-by-field disposition and the ownership tree is already traced, so the analyst reviews a finished file instead of assembling one.

Account openingIdentity verificationSanctions and PEP screeningBusiness registriesCase management

Workflow 03 · Wealth advisor or portfolio manager

Client book rebalancing

Client book reviewFinancial services

  1. 01 · Trigger

    Advisor asks about the book

    An advisor asks who is out of balance, who has losses to harvest, or who has a required minimum distribution coming.

  2. 02 · Tool call

    Load positions and models

    Reads custodian positions for every household and compares each one to its model allocation and tolerance band.

  3. 03 · Tool call

    Check tax lots and RMDs

    Finds lots with harvestable losses, checks the 30-day wash sale window, and lists RMDs due this year.

  4. 04 · Reasoning

    Draft the trades

    Proposes trades that bring each household back to its model, with turnover, estimated tax and losses harvested, inside the household's investment policy.

  5. 05 · Human approval

    Advisor sends the proposal

    The advisor reviews the trades and approves sending the proposal to the client. Nothing trades until the client signs.

    Nothing happens until a person approves

  6. 06 · Outcome

    Proposals out the same day

    Out-of-tolerance households get a proposal with its tax cost stated, instead of waiting for the next quarterly review.

What changes

Drift, cash drag, loss harvesting and RMDs show up in one list, and each proposal goes to the client only after the advisor approves it.

Portfolio managementCustodian feedsTax lot accountingClient portal

Workflow 04 · Underwriter or credit officer

Loan underwriting

Loan underwritingFinancial services

  1. 01 · Trigger

    File ready for underwriting

    A loan application reaches the underwriting queue, or an underwriter asks for a file by borrower name.

  2. 02 · Tool call

    Read the application and credit

    Pulls the application, the tri-merge credit report and the property or collateral value.

  3. 03 · Tool call

    Verify income

    Checks each income source against payroll data, IRS transcripts, leases or bank statements, and leaves unverified income out of the ratios.

  4. 04 · Reasoning

    Run policy and draft conditions

    Calculates DTI and LTV, checks each credit policy rule, documents any exception with its compensating factors, and drafts the conditions list.

  5. 05 · Human approval

    Underwriter decides

    The underwriter edits the conditions and approves with conditions, or refers the file to a senior underwriter.

    Nothing happens until a person approves

  6. 06 · Outcome

    Conditional approval issued

    The processor receives the approval and the conditions, and the file records the ratios, checks and rationale behind the decision.

What changes

The underwriter starts from calculated ratios, verified income and a drafted conditions list, and every exception carries its reason.

Loan originationCredit bureausIncome and employment verificationValuationCredit policy

How it plugs in

An agent on your systems, one component per workflow.

Requests come in by voice, text or system event. The router picks the workflow, the agent calls your systems, and the answer lands as a live card where your team works. Nothing is written back until a person approves it.

In

VoicePush to talk, any browser
Typed requestSame intent, same card
System eventWebhook, queue or schedule

Tiersel

Intent routerPicks 1 of 4 workflows as you speak
Fraud alert reviewFraud analyst
KYC and KYB onboarding reviewKYC or onboarding analyst
Client book rebalancingWealth advisor or portfolio manager
Loan underwritingUnderwriter or credit officer

Your systems

Core bankingFiserv, FIS
Card processing and fraud scoringFICO Falcon, Featurespace
AML, screening and case managementNICE Actimize, Dow Jones Risk & Compliance
Identity and account openingAlloy, Socure
Loan origination and creditICE Encompass, nCino

Out

Live cardIn your app, sidebar or kiosk
Human approvalAnalyst decides
Write-back + auditEvery call and decision logged

Workflow blueprint · Fraud analyst

Fraud alert review

Starts when
Alert lands in the queue. The fraud model flags an authorization, or an analyst asks for the next alert on a card or account.
Reads
Pull the authorization and history. Reads the flagged transaction, the model score and 90 days of card activity from the card platform and core banking.
Check device, location and links. Looks up the device fingerprint, the IP location against the last card-present purchase, and other customers that share the device or payee.
Decides
Weigh the signals. Shows each signal with its weight, such as impossible travel, a new device, rapid retries or an amount far above the customer's pattern, and recommends freeze, step-up or release under the card fraud policy.
Waits for
Analyst decides. The analyst picks freeze, step-up verification or release and approves it. Linked cards and payees are included only if the analyst switches them on.
Writes back
A decision with its evidence. The card is frozen or released, the customer is notified, and the case record holds every signal, source and approver.
Card processing and fraud scoringCore bankingDevice intelligenceCase management

This workflow is one component

export const fraud = defineSkill({
  id: "fraud", category: "finance",
  // when this workflow, and not a neighbour, is meant
  criterion: "A bank or fintech analyst wants to review a flagged transaction or fraud alert on a customer's card or account and decide whether to freeze, verify or release it; not a person checking their own spending, and not home security.",
  examples: ["review the fraud alert on account 4471", "show me the flagged transaction on card ending 2290"],
  match: (text) => score, // intent score, runs on every word
  parse: (text, ctx) => record, // entities + tool calls into your systems
  reply: (record) => "…",
  Card: FraudAlertReviewCard, // the live UI, with its approval step
});

The id, criterion, examples and reply are read live from the skill that renders the card above. Adding a workflow for your team means adding one of these: its intent, the calls it makes, what it says, and the interface your people approve in.

Fits your stack, answers to your people

Reads your systems. Acts only with approval.

What the agent connects to

Core banking

FiservFISJack Henry

Card processing and fraud scoring

FICO FalconFeaturespacecard processor feeds

AML, screening and case management

NICE ActimizeDow Jones Risk & ComplianceLexisNexis Bridger

Identity and account opening

AlloySocurestate business registries

Loan origination and credit

ICE EncompassnCinocredit bureau reports

Wealth and portfolio management

OrionEnvestnetcustodian feeds

Product names are examples of systems in this category. Each deployment is connected to your systems through their APIs as part of the build.

Approval before any consequential action

Freezing a card, approving onboarding, issuing a loan decision and sending a client proposal all wait for a named person to approve. The agent prepares the action and stops.

Every step is logged

Each card shows the systems the agent read, what it found, how long each call took and which evidence supports the recommendation. Approvals and rejections record who decided and when.

Least-privilege access, data stays with you

When deployed, the agent runs in your environment with scoped service accounts. Customer and account data stays in your systems, and write access covers only the actions a person approves.

Synthetic data in this demo

Every customer, company, account number and figure on the cards is invented. No real customer records are used.

The cost of doing it by hand

Put your own numbers in.

Move the sliders to match your operation. Nothing is sent anywhere unless you choose to email it.

Each year

4,800

staff hours back, about 2.3 full-time people

$187,200

of staff time, at your hourly cost

Email me a pilot plan for these numbers

Assumes the agent takes over the data gathering and write-up, not the decision. The $39 default is the BLS May 2025 median wage for compliance officers ($38.81 an hour) before benefits and overhead, so the estimate is conservative. The volume and minutes are starting assumptions; change any number to match your team.

Hours = alerts and cases a month × 12 × minutes each × share, ÷ 60. A full-time person is 2,080 hours a year. Default hourly cost from U.S. Bureau of Labor Statistics, Occupational Outlook Handbook, 2025.

Questions buyers ask

Straight answers.

  • When deployed, the agent runs in your environment and reads your systems through scoped service accounts. Customer records stay there. This demo uses only synthetic data.

  • No. Freezes, onboarding approvals, loan decisions and client proposals all wait for a person to approve them, and every approval and rejection is logged with who decided and when.

  • No. Your fraud model and transaction monitoring still generate the alerts. The agent works the alert: it gathers the evidence, explains the signals and prepares the action for an analyst.

  • Each recommendation shows the evidence and the policy rule behind it, so an analyst can check it before acting. A pilot runs the agent alongside your analysts on past cases and measures how often its recommendation matches their decision.

  • The agent does not make the decision, and each card keeps the tool calls, sources and approver for review. Your model risk and compliance teams decide how it is validated before it touches production data.

  • The demo on this page is free and needs no signup. For a pilot on your own data, book a call and we will scope one workflow with read-only access first.

See it on your own workflow.

Watch the financial services workflows run end to end, or speak to the agent yourself. Then tell us which one to build on your data.

Scope a pilot with hello@sasid.ai
  1. 01

    Pick one workflow

    A short call to choose the workflow that costs you the most time, and who approves its actions.

  2. 02

    See it on your data

    We build that workflow against your sample or de-identified data so your team judges the real thing.

  3. 03

    Run it with people in charge

    Your staff use it on live work. The agent drafts and prepares; a person approves every consequential step.

  4. 04

    Decide from results

    You keep what measurably saves time and drop what does not.

Other operations

The same agent, other teams.