Question 1

What is agentic AI in finance?

Accepted Answer

Agentic AI is software that pursues a goal using tools and bounded actions, rather than answering a single prompt. In finance that means agents that monitor the close checklist, chase late submissions, check intercompany mismatches as they appear, and re-run forecasts when a driver moves, escalating to a human when they are stuck.

Question 2

How is agentic AI different from generative AI?

Accepted Answer

Generative AI drafts: you ask, it answers, a human decides. An agent initiates work: it is given a goal, decides which step comes next, acts within limits you set, and escalates exceptions. The difference is who starts the work and how many steps the software takes on its own.

Question 3

What agentic AI use cases work in finance today?

Accepted Answer

Close monitoring and submission chasing, intercompany mismatch detection, reconciliation matching, and forecast re-runs with drafted variance commentary. Agents that post journals, approve payments, or sign off reconciliations unsupervised remain demo-ware: bounded action with human sign-off is the honest frontier in 2026.

Question 4

What does a finance function need before adopting agentic AI?

Accepted Answer

One governed data model with clean hierarchies and matched intercompany, and a control framework every agent action lands inside. An agent inherits the data it runs on; give it mappings that do not tie and it chases the wrong exceptions faster. A fixed-fee AI readiness assessment maps the gaps and the order of adoption.