How Ashling, UiPath, and Google helped a North American security services company apply thousands of unmatched payments the same business day they arrive.
A leading North American security services company is preparing to clear unmatched payments from cash suspense the same business day they arrive. Once live, the solution is expected to give its finance team more than 500 hours of manual research back every month.
The work will run on the company's first fully autonomous AI agent in production: a Google Gemini agent that investigates each payment, orchestrated by UiPath Maestro alongside UiPath robots and Google Cloud data services.
Every day, checks and money orders arrive from customers waiting to see them reflected on their accounts. Most carry what the bank needs to match each payment to an account. About 4,000 a month don't: a missing account or invoice number, or an address on the check that doesn't match the one on file.
Those payments land in cash suspense, and the finance team has to work out where each one belongs. The process was entirely manual. Each transaction took about 7 minutes: pull the check or money order image from the bank, search the system of record, confirm the match, and post. Across thousands of payments, that added up to hundreds of hours every month.
The company's goal is to apply every payment the same business day it arrives. Each day a payment sits in suspense is another day a customer waits for their balance to clear, and another chance that wait becomes a call to support.
The team had looked at this use case before. Its intelligent automation program began with a UiPath proof of concept in financial operations more than six years ago. Rules-based RPA alone couldn't handle the fuzzy logic and business exceptions that unmatched payments bring.
Agentic AI gave the company a way to reopen the use case. Ashling applied its reference architecture and the company's governance policies to design and build a process that gives each part of the work to the technology best suited for it: rules-based robots for high-volume tasks, and AI models only where judgment is needed.
|
Component |
Role |
What It Does at ADT |
|
UiPath robots |
The hands |
Run the high-volume, rules-based steps reliably, including retrieving payment images from systems with no API integration options |
|
Google Gemini Enterprise Agent Platform |
The brain |
Brings LLM reasoning and agentic decision-making to ambiguous cases that used to need a person, such as unclear remittance details or partial payments |
|
UiPath Maestro |
The conductor |
Orchestrates robots, AI agents, data services, and people in one end-to-end process, capturing every step along the way |
Maestro's visibility matters as much as its coordination. Cash application is a financial process held to SOX standards, so every transaction needs a complete audit trail. Maestro records each hand-off and decision as the work happens.
Each day, a batch of unidentified payments arrives from the bank. UiPath robots retrieve every check and money order image, along with any supporting documents.
The Gemini agent then takes over the investigation. It reads what an analyst used to hunt for: invoice numbers, account numbers, names, addresses, and money order serial numbers. It searches the company's systems of record, working through each identifier in order of accuracy.
Next, the agent confirms the match. It checks the payment amount against the customer's balance. When those don't line up, it reviews six months of payment history to see whether the amount fits the customer's pattern. It also sorts a single check split across multiple accounts.
Once the account is verified against multiple sources, the payment posts and clears suspense. When a payment can't be validated or falls outside business rules, it routes to the finance team for review, with the agent's full case notes attached.
Throughout, the agent documents its work: which searches it ran, what it found, and the reasoning behind every match. Analysts start each exception with the investigation already done.
Once live, customers should see their balances clear faster, the finance team will spend its time on the payments that need human expertise, and each of those will arrive with preliminary research in hand.
The company plans to add an outreach agent, built on Gemini Enterprise Agent Platform, for payments the first agent can't verify. It will contact customers to request missing information, raising throughput with fewer human touchpoints.
The bigger shift is on the roadmap. Cash application was one of several high-value use cases the company had shelved because they were too complex for RPA alone. With an orchestration layer that combines rules and reasoning, those use cases are back in play, and the same governance and reference architecture carry over to each one.
That is what Better Together looks like in practice: UiPath and Google, each doing the work it does best, coordinated in one flow the company can trust and audit.