Kasisto

KAIgentic is the agentic AI platform at the center of the product. It splits a banking workflow across specialized agents instead of asking one model to handle everything, so a dispute, an account-opening question and a wealth inquiry can each run through a narrower set of instructions. A console shows agents, workflows and knowledge hubs, and gives staff a place to inspect and intervene. Kasisto launched it in early access in August 2025 and later integrated it with Microsoft Agent 365.

KAI Answers is a generative search layer over the bank's policies, procedures, regulatory filings and product documents. It turns flat or tabular source material into a written answer and attaches a citation to the original passage. That citation is the useful part for a regulated business, because a reviewer can verify the claim instead of trusting the model. The same layer feeds answers to customer-facing assistants and to employees.

KAI-GPT is Kasisto's banking-specific language model, launched in May 2023. Banks can use it inside KAIgentic or choose a general model, and the banking model can run in the institution's own cloud or on premises. That option matters to banks that cannot send customer data to a shared SaaS endpoint. Model choice is a per-deployment decision, not a single fixed stack.

Kasisto ships pre-trained assistants for consumer and business banking rather than an empty development kit. They cover common intents such as balances, transactions, cards, loans and branch details, and they run in mobile apps, websites, SMS, chat apps and voice channels. KAI Express, the 2021 cloud package, was designed to deploy in 30 days with a 90-day exit clause if the bank was not satisfied.

Every agent runs through pre-processing and post-processing layers that check the request against policy, flag possible hallucinations and score confidence before an answer goes out. If confidence falls below the threshold, the conversation moves to a human with the history attached. This is the part that separates agentic AI for financial services from a generic support bot, and it is also the part a bank must test against its own rules.

KAIops is an operations layer that monitors system health, correlates telemetry and drafts root-cause analysis for IT incidents. It was announced in October 2025 and runs on top of KAIgentic. For a bank, the appeal is faster incident review without asking engineers to reconstruct a timeline by hand. It is a newer product than the customer-facing assistants, so production references are thinner.