The agentic AI platforms banks are evaluating in 2026 fall into four groups: banking platforms adding an agent layer (Backbase, Oracle, nCino, Fiserv, Salesforce, ServiceNow and the core vendors), horizontal agent platforms (Microsoft, Google Cloud, IBM, Kore.ai), specialists that go deep on one workflow (Gradient Labs, Casap, Bretton AI, Taktile), and Indian players (Intellect, Perfios, Lyzr, Fluid AI). The right choice depends on which systems you already run, how many desks you want agents on, and how much control your risk team needs over each agent.

This list describes each vendor from its own public announcements or reputable press, linked in each entry, as of November 2026. It is a map for a first conversation, not a ranking. Tuvel is on it, described the same way.

How to read the list

Ask four questions of every vendor:

  • Who does it serve? Customers, staff, or both.
  • Does it act, or only assist? And if it acts, inside which limits.
  • How is it deployed? Inside the vendor’s own platform, as shared SaaS, or as a copy for your bank.
  • What governance does it describe? Approvals and audit are common. Per-agent levels, budgets and a kill switch are less so.

Banking platforms adding an agent layer

These suit banks already running the vendor’s platform, because the agents sit inside it.

Backbase (with Kasisto)

Backbase launched an AI-native Banking OS in April 2026, including Sentinel, which it describes as checking actions by customers, staff or AI agents against bank policy in real time (itbrief). It acquired Kasisto in June 2026 and offers an Agentic Banking suite to Banking OS customers (Backbase). Best fit: banks on, or moving to, Backbase.

Oracle Financial Services

An agentic platform for retail banking (February 2026) and corporate banking (April 2026), with pre-built agents for treasury, trade finance, credit and lending, and a human in the loop (Oracle). Best fit: FLEXCUBE banks.

nCino

Role-based “Digital Partners” (Executive, Analyst, Service, Processor and Client) and an Agentic Operating System, announced November 2025 (nCino). Best fit: lenders already on nCino.

Fiserv agentOS

Launched May 2026 with agents for commercial loan onboarding, daily operational analysis, deposit intelligence and AML triage, described with identity-bound execution, policy enforcement and audit (Amazon press). Best fit: Fiserv core banks.

Salesforce Agentforce for Financial Services

Advisor and banker agents for meeting prep and record updates, a banking service agent and a digital loan officer, with policy controls and an audit trail (cfotech). Best fit: banks that run on Salesforce.

ServiceNow

Financial Services Operations agents, disputes agents and an Agentic Contact Center for Banking, governed through its AI Control Tower (ServiceNow). Best fit: banks running operations on ServiceNow.

Core vendors: Temenos, Infosys Finacle, TCS BaNCS

Temenos offers an FCM AI Agent for sanctions alerts and copilots for its core (Temenos). Finacle exposes more than 200 “governed banking skills” callable as MCP tools (Forrester). TCS BaNCS AI Compass adds agents for account opening, credit underwriting and query resolution (itbrief). Worth knowing: skills a core exposes can be called by other agents too, under your own limits.

Horizontal agent platforms

These suit banks that want one enterprise-wide agent layer and will build the banking parts themselves.

Microsoft Agent 365

A registry and control plane for every agent in an enterprise, generally available since May 2026 (Futurum). Many banks will expect other agents to sit alongside it.

Google Cloud Gemini Enterprise for Financial Services

Launched August 2026 with a financial research agent for KYC, beneficial ownership and credit work, and a control plane with audit (Google Cloud).

IBM watsonx Orchestrate

A catalogue of more than 150 agents, as SaaS or on premises (IBM).

Kore.ai

Banking agents for customers and staff, with Artemis (May 2026) running in public, sovereign or private cloud, or on premises, and policy enforcement, human in the loop and step traces (Kore.ai).

Specialists that go deep on one workflow

These suit a bank with one painful queue it wants handled end to end.

  • Gradient Labs: customer operations, collections, disputes and KYB agents, where the bank designs the workflows and limits and approves each release (Gradient Labs).
  • Casap: disputes and first-party fraud agents that investigate and act under the bank’s policy, with full audit and human escalation (FF News).
  • Bretton AI (formerly Greenlite): agents for level 1 and level 2 AML and KYC investigations (RegTech Analyst).
  • Taktile: an agentic decision platform combining agents, rules and human oversight for credit, fraud, AML and claims (Taktile).
  • Saris: back-office lending agents with integrations to common US lending systems (BetaKit).

India

  • Intellect Design, Purple Fabric: more than 65 pre-trained “Digital Experts”, plus PF Credit (SME Channels).
  • Perfios: an agentic AI operating system shown in September 2026, with agentic underwriting, fraud and AML, and multi-agent collections (ThePrint).
  • Lyzr: a registry of more than 400 banking agents with staged autonomy, replayable logs and on-premises deployment (Lyzr).
  • Fluid AI: banking agents and voice, deployable on premises or air-gapped (Fluid AI).
  • Voice specialists: Gnani.ai and Sarvam AI build Indian-language voice agents, including for collections (Gnani, Sarvam).

Tuvel

Tuvel is an agentic banking platform with five products on one control plane: Tuvel Advisor, a customer money advisor; Tuvel Agent Teams, 99 staff agents on 14 desks from lending to compliance; Tuvel Rooms, where staff and agents work a case together; Tuvel Control Plane, with per-agent levels, budgets, approvals, model routing, a kill switch and audit; and Tuvel Launch. Customers choose the advisor’s level inside a ceiling the bank sets. Each bank gets its own private copy, under its own brand, alongside the core it already runs. Being deployed with a Fortune 500 US bank and one of India’s three largest banks. Best fit: banks, NBFCs, wealth managers and insurers that want agents for staff and customers under one set of controls.

How to choose

  1. Start from your systems. If one platform already runs most of your desks, see what its agents cover first.
  2. Count the desks. One queue suits a specialist. Several desks plus customers suit a platform with one control plane.
  3. Put governance in writing. Ask each vendor to show levels, limits, approvals, a kill switch and the record, working, in a demo. Our governance guide lists what to ask.
  4. Pilot one desk. Six to eight weeks, low limits, your approvers. See how a pilot works.