Rate Shop Radar: measure how much deposit is moving because of AI agents

I found no product that measures how much of a bank's deposit outflow AI agents caused, or that turns that into agent-adjusted beta and decay assumptions for ALCO. The pieces exist in separate places. Treasury analytics vendors (Curinos, Deposit Analysis) handle betas, decay and flow-of-funds. KlariVis and Flowtracker classify outbound transfers by destination using a bank's own transaction data. Incognia, BioCatch and Transmit detect agent sessions for fraud and authorization. A new entrant's opening is to join session-level agent signals to destination-classified outflows and sell the result to treasury and the CFO as a pricing and liquidity input rather than a fraud tool. The timing is good because the topic only went mainstream in August and September 2026 (Torsten Slok's agentic bank run note, Dallas Fed research on beta), but most commentary still treats agent-driven runs as a possible scenario, not something happening now, so measured agent outflow may be small today. The main risks are that Curinos or KlariVis add an agent segment cheaply, that fraud vendors sell their agent signals to treasury, and that agents moving money through Plaid, APIs or tokenized rails never open a browser session, so form-fill timing and tab patterns would miss much of the activity. A safer position is to sell rate-sensitive outflow measurement with agent attribution as one layer, and to partner with fraud-signal providers rather than rebuild detection.

• Curinos (Treasury Analyzer, flow-of-funds, Retail Deposit Optimizer)

  Incumbent deposit analytics vendor to regional banks. Offers deposit beta, decay, weighted-average life and duration studies, plus flow-of-funds analysis separating balance drawdowns from closures, and competitor rate data. No agent attribution found, but it owns the treasury and ALCO buyer relationship and could add an agent segment.

• KlariVis Transactional Intelligence

  Bank data analytics platform that already tags outbound transfers to named destinations in community banks' own data (its Coinbase 'Quiet Spread' study across 92 to 122 banks). Very close to the outflow classifier idea, applied to crypto rather than high-yield or agent destinations.

• Deposit Analysis

  Sells measured deposit betas and decay rates from Call Report data for about 4,300 US banks, prepared for ALM and ALCO use. Overlaps the ALCO pack but works from public data and does not attribute anything to agents.

• Flowtracker Analytics

  Markets analytics on hidden deposit attrition inside open accounts for banks and credit unions. Overlaps outflow classification and segmentation.

• Incognia AI Agent Detection

  Launched agent detection for financial institutions (signed-request verification, agent-origin classification, bot and automation detection). Covers the agent-session flag for fraud and authorization, not for deposit pricing.

• BioCatch agentic AI use cases

  Behavioral biometrics vendor that already sells session-level identification of AI agents in banking, framed as fraud and account takeover. Could repackage the same signals for treasury.

• Transmit Security

  Identity and fraud vendor researching how to tell agent sessions from human ones in banking. Says behavioral biometrics and device fingerprinting weaken with agents, which challenges the session-scoring approach.

Checked 2026-10-17