ConnectHubFlowX.AI brings banking agents to Gemini Enterprise
Agents that check lending files and reconcile onboarding documents are now on Google Cloud Marketplace.
The launch
On August 25, Google Cloud launched Gemini Enterprise for Financial Services, a preview package for capital markets and corporate banking with 13 connectors and 50 ready-made skills. It is one of the first industry packages built on Gemini Enterprise, alongside one for legal work, and includes agents from partners such as D&B and S&P Global.
FlowX.AI's agents are among them. They are listed on Google Cloud Marketplace, sold as "AI Agent as a Service" and compatible with the Agent2Agent (A2A) protocol, which lets agents from different vendors work together.
Agent 1: is the loan file complete?
The first agent reviews an entire lending file against the institution's own document policy. It checks that the required documents are there, in the right quantities and recent enough, then cross-checks the information between documents to find real inconsistencies.
The result is a list for the case handler: what is missing, out of date, insufficient or contradictory, and why.
Agent 2: do the onboarding documents agree?
The second agent reads an institutional onboarding pack for capital markets. It extracts the required information into a structure the customer defines, gives every value with its evidence and page reference, and links mentions of the same person or company across documents.
Every result is labelled as agreed, single-source, conflicting or missing, so a reviewer can see straight away where the documents agree and where they don't.
Rules where it counts, people for judgement
The most interesting design choice is what is not left to generative AI. Checks such as whether a document is missing or whether a policy date has passed are handled by fixed rules. AI does the reading and cross-checking; the rules handle the yes-or-no questions that must never be wrong.
Each agent works out what it can prove and then hands the decision to a person, with the evidence attached. FlowX says governance, traceability and human control are built in from the start and recommends starting with a few focused, high-value cases before expanding.
Chief executive Ioan Iacob argued that the biggest opportunity is building AI into the processes that run a business, not adding another assistant on top. FlowX also targets insurance, logistics, telecoms and retail.
Why it matters for your business
This is the pattern that works in regulated work: fixed rules for checks that must never be wrong, AI for reading messy documents, and a person signing off on the decision. It is also how you get an AI process past an auditor, because every answer points back to a page.
You don't need to be a bank to use it. Any business that checks packs of documents, such as supplier onboarding, insurance claims, tenancy applications or customs paperwork, can apply the same split. We design document AI and OCR workflows this way: rules for the checklist, AI for extraction, and a review screen that shows each value next to its source.
What to do next
- Pick one document pack your team checks by hand
- Write its checklist as rules: which documents, how recent, which fields must match
- Use AI for extraction only, and show reviewers each value next to its source page
