Microsoft puts Copilot for Finance in public preview
Microsoft's new generative AI assistant, Copilot for Finance, looks to help financial professionals become more efficient -- but adoption will depend on data quality.
Microsoft's wave of generative AI assistants continues with the recent debut of Copilot for Finance, now available in public preview. It will be embedded in Excel, Outlook and Teams and provide capabilities to help finance professionals perform routine tasks and derive insights from data across systems.
Finance is an area that can benefit from a GenAI assistant like Copilot for Finance, according to industry analysts, but it is still in its early stages and organizations will need to be careful about how the data sources are used to generate trustworthy results.
Users can ask questions or select from prompts, and Copilot will access structured data from financial systems such as Dynamics 365 and SAP, as well as unstructured data from applications such as Outlook. It then presents recommendations on actions to take.
The idea is to have Copilot take on repetitive, manual tasks and free users to focus on strategic work in financial processes such as collections or forecasting, according to Georg Glantschnig, vice president for Dynamics 365 ERP applications at Microsoft.
Copilot for Finance currently focuses on three areas: variance analysis, which looks at the differences between financial forecasts and performance; collections for resolving delinquent accounts; and invoice matching and reconciliation.
Reducing manual effort
These three areas typically require manual effort to look through systems and applications to find the data, he said. Now, users can select from natural language prompts such as, "Help me understand forecast-to-actuals variance data," and Copilot will pull data directly from ERP systems and other financial applications.
"Copilot for Finance is designed to connect different financial systems with the productivity tools like Excel and Outlook -- because that's where the users are -- to help you to get your job done without having to jump around between different systems and using GenAI to help write emails, understand insights," Glantschnig said.
Copilot for Finance connects directly to Dynamics 365 and SAP, and can be connected to more than 1,200 other systems via Copilot Studio, a development application for extending Microsoft Copilots.
Copilot suggests actions that users can take and generates text messages with attached relevant files such as invoices, Glantschnig said, but users can check data sources before taking any actions.
Copilot for Finance joins other Microsoft Copilots that are designed for specific roles such as sales or span platforms, including Copilot for Microsoft 365 and Copilot for Dynamics 365.
Finance a prime use case for GenAI
Finance is a prime use case for GenAI, and Copilot for Finance is built around tasks such as budgeting and planning or forecasting, where it can help organize data and combine structured and unstructured data, said Kevin Permenter, research director at IDC.
Putting unstructured and structured data together -- like combining the numbers on an Excel spreadsheet with the notes on the sheet -- is one area where GenAI can take off.
Kevin PermenterResearch director, IDC
"Putting unstructured and structured data together -- like combining the numbers on an Excel spreadsheet with the notes on the sheet -- is one area where GenAI can take off," Permenter said.
Embedding Copilot for Finance in ubiquitous applications like Excel, Outlook and Teams will be a boon for adoption as people are always looking for ways to save time and applications that help them work more effectively, he said.
"You'll probably see it in its most full form inside Outlook, where you're reading emails, it's writing or summarizing emails for you," Permenter said. "Especially in something like a financial close, where there are a lot of emails and texts to get a handle on, there's a real advantage for Microsoft, and Copilot opens the door for them to press that advantage."
Users will have concerns about being able to trust the recommendations that Copilot for Finance provides, he said, but humans propagate errors as well, and Copilot will not make decisions itself.
"It will run scenarios and identify five or six that are most likely [to have an impact], and out of those will say which are the ones that are most aligned with your goals and narrow it down to two or three, but it's up to you to go through and check everything," Permenter said.
Still in early days
Areas such as finance, which still require manual effort and time to conduct routine tasks, are prime areas for GenAI tools like Copilot for Finance, said Liz Herbert, vice president and principal analyst at Forrester Research.
Copilot for Finance promises to connect information from applications like Outlook emails and Excel files with ERP systems like Dynamics 365 and SAP, but making it all work won't be easy, she said.
"Organizations will still have to do some work. It's not like it's magically going to know where to look and dig up all that stuff," Herbert said. "When you get to something that's complex and spans systems and collaboration tools like Excel and email, it will probably need some time to get more accurate because so much information is scattered about different places."
The accuracy of Copilot for Finance's recommendations will be lower if data sources are incomplete, but there will be "humans in the room" to check the work, she said. However, because financial data is often audited, it tends to be more accurate than other enterprise app data.
"Some of this will be easier in finance because that type of data has such a high and necessary degree of accuracy," Herbert said. "It's not like a salesperson's notes as to what they thought happened or if they met with a client. Those kinds of things have giant data quality issues."
The most interesting use case for Copilot for Finance is its reconciliation capability, because that addresses a problem of matching invoices that organizations face, said Moutusi Sau, an analyst at Gartner.
However, trust will be an issue, as there is some evidence that testing for Copilot on Teams has been spotty on tasks like summarizing action items on a call, Sau said.
"If [the Copilots] going in an action space, like giving you the action item, you have to be really careful," she said. "But when it comes to simply matching and backfilling information, that's probably a good application for Copilot for Finance."
Embedding Copilot for Finance in Microsoft's productivity applications will be important for adoption, given industries such as investment banking still do everything in Excel, she said.
"Because financial analysts deal with so much data on a regular basis, if you can build all these functionalities inside the spreadsheet itself, that's a smart move," Sau said. "They're trying come to the users where they are."
Jim O'Donnell is a senior news writer for TechTarget Editorial who covers ERP and other enterprise applications.
