Customer Stories
Asurion's finance team needed to recover touchless invoice processing after exception-heavy scenarios slowed adoption.
Working with Microsoft, the IT team embedded a Copilot Studio agent into the invoice capture process so common exceptions could be resolved before invoices were posted.
Touchless processing rose from below 20% to more than 70%, creating a pattern for broader finance automation.
Asurion provides tech care services through protection, repair, installation, replacement, and expert support for the devices people rely on every day. This includes phones, smart home devices, major appliances and operates globally through a network of warehouse, technicians and more than 600 uBreakiFix stores.
The finance modernization journey began with a practical target: make accounts payable automation work better at scale. Asurion was already using Microsoft Dynamics 365 Finance, invoice capture, Microsoft Power Automate, Microsoft Copilot Studio, Microsoft Power Platform, Microsoft AI Builder, and Microsoft Dataverse as part of its finance automation environment. Don Baldwin, Senior Director of Software Engineering at Asurion, frames the business direction clearly: “Our finance roadmap this year is all about automation, being an AI-enabled, automation-first company, finance processes.”
“The real value of AI is creating capacity for people to focus on higher-value work. By embedding Copilot Studio into critical finance processes, we're improving automation outcomes and laying the foundation for broader transformation across finance.”
Tony Delligatti, VP Fianance, Asurion
Asurion's accounts payable (AP) team needed more than document capture. Several years earlier, the company had implemented an OCR approach that stored invoice images and gave AP analysts a way to view them, but much of the processing still happened manually. As the team expanded its use of invoice capture, it expected the application to match invoices to purchase orders and prepare them for payment. The goal was roughly 80% touchless processing globally.
The reality was more complex. Some vendors submitted external item numbers that did not map cleanly to Asurion's internal item structure. A single external OEM item could correspond to different internal SKUs depending on vendor context. Some invoices included hundreds of line items, requiring AP analysts to correct errors in the capture tool and then repeat some updates in the finance system. The challenge affected a substantial share of Asurion’s invoice volume. The company processed approximately 30,000 direct purchase order invoices annually, and roughly 90% had at least one exception. It also handled approximately 25,000 indirect invoices each year; more than 75% of which had header-level issues. Baldwin summarizes the challenge this way: “My finance team with invoice capture was facing a challenge of actually getting more than a 20% success rate for touchless processing.”
When adoption dropped below 20%, the finance team evaluated third-party OCR alternatives. After reviewing vendor proposals and reassessing the existing Microsoft approach, the team reached a practical conclusion: the same mapping, derivation, and exception challenges would need to be solved no matter which product Asurion selected. Rather than abandon its investment, Asurion asked Microsoft whether AI and agents could help address the root issues.
Asurion and Microsoft moved quickly. Within about a week, the teams had working proof points that addressed four of the biggest issues preventing wider adoption of invoice capture. The solution centers on an Invoice Capture Derivation agent built with Copilot Studio and embedded directly into the invoice capture process. When invoice data enters an error state, Power Automate routes the relevant details to the agent, which applies a defined set of instructions to resolve the issue.
Baldwin explains the approach and its impact: "We created a Copilot Studio agent and embedded it into a Power Automate flow. Embedding the agent let us resolve invoice errors before they ever hit finance.” Invoice capture identifies error-state data, Power Automate extracts the relevant information, and the agent uses a defined set of instructions to resolve the exception. A follow-up flow then reprocesses the data, so a complete invoice can move into Dynamics 365 Finance for downstream processing.
The agent is designed for deterministic finance scenarios. For example, when a vendor submits item numbers that don’t directly match Asurion's purchase-order data, the agent can trim or adjust the item number, compare it to the specific purchase order, and identify the correct item. It can also modify purchase-order numbers when vendors append extra information that prevents a match. Asurion is also exploring MCP server capabilities in a QA environment, while continuing to validate security safeguards and large-file handling.
“We created a Copilot Studio agent and embedded it into a Power Automate flow. Embedding the agent let us resolve invoice errors before they ever hit finance.”
Don Baldwin, Senior Director of Software Engineering, Asurion
Touchless processing rose to 70%+
For AP analysts, the change was visible immediately. Exceptions that once required line-by-line correction no longer demanded the same level of manual intervention. Scenarios that previously required manual rework could now move through an AI-assisted derivation process, reducing duplicated effort and giving finance leaders a clearer path back to automation.
The results followed quickly. As Baldwin puts it: "As the agent went into production, we saw an immediate increase in touchless invoicing." Asurion moved from less than 20% touchless processing for two vendors to more than 55% within the first week for North and South American vendors. With further refinement, the team reached more than 70% touchless processing across North and South America, with a path toward 85% or more as MCP server work advances from QA into production.
Beyond the metrics, the project preserved Asurion's existing invoice capture investment, avoided a near-term shift to third-party OCR tools, and demonstrated that the Microsoft stack could handle exception-heavy finance scenarios. Just as important, it rebuilt confidence among finance leaders, who began exploring additional opportunities for automation.
Tony Delligatti, Vice President of Finance at Asurion, emphasizes the broader value: “The real value of AI is creating capacity for people to focus on higher-value work. By embedding Copilot Studio into critical finance processes, we're improving automation outcomes and laying the foundation for broader transformation across finance.”
“As the agent went into production, we saw an immediate increase in touchless invoicing.”
Don Baldwin, Senior Director of Software Engineering, Asurion
Invoice capture is now the foundation for a broader finance automation strategy. Baldwin explains, "Invoice capture is just the beginning. We have a lot of other opportunities in front of us." Next focus areas include journal entry automation, journal entry review, and month-end account reconciliation. In one test scenario, an agent extracts data from the finance system and supporting documentation, applies defined instructions, and returns AI-generated insights so an accountant can decide whether to post entries or resolve discrepancies.
Asurion is also evaluating first-party finance capabilities for subledger-to-ledger reconciliation and future inventory-related functionality. The team is balancing ambition with governance: agents need the right permissions, read-only access where possible, and human review in the right places. Baldwin's advice to others exploring agents is to start with something attainable, build trust through a focused use case, and then expand once business partners see clear value.
The near-term roadmap includes improving large invoice PDF processing, targeting remaining exception-heavy vendors, addressing VAT tax to charge-code conversion and unit conversion scenarios, and extending the pattern to additional regions such as APAC and Europe. In parallel, cash application to settle payments to open AR invoices using AI insights for a human in the loop and journal entry automation and review processes along with balance sheet reconciliations are also being rolled out. The company plans to continue to expand and roll out the functionality globally. Baldwin describes the momentum this way: "The business team is getting really, really excited about what they're seeing with invoice capture. And they're excited to see what we can continue to build to automate those processes."
“The business team is getting really, really excited about what they're seeing with invoice capture. And they're excited to see what we can continue to build to automate those processes.”
Don Baldwin, Senior Director of Software Engineering, Asurion
Follow Microsoft