{"id":25500,"date":"2026-09-24T08:30:00","date_gmt":"2026-09-24T15:30:00","guid":{"rendered":"https:\/\/cm-edgetun.pages.dev\/insidetrack\/blog\/?p=25500"},"modified":"2026-09-23T15:55:56","modified_gmt":"2026-09-23T22:55:56","slug":"prioritizing-ai-transformation-opportunities-in-microsoft-finance","status":"publish","type":"post","link":"https:\/\/cm-edgetun.pages.dev\/insidetrack\/blog\/prioritizing-ai-transformation-opportunities-in-microsoft-finance\/","title":{"rendered":"Prioritizing AI transformation opportunities in Microsoft Finance"},"content":{"rendered":"\n

When organizations begin adopting AI, one of the first important questions to be answered is where to start. The needs and options across different teams and functions can feel overwhelming.<\/p>\n\n\n\n

Our Microsoft Finance organization was no different. When our leaders were preparing to add AI into everyday operations, the opportunities for significant efficiency gains were clear, but achieving those came with significant challenges.  <\/p>\n\n\n\n

Within Microsoft finance operations, Treasury provides a useful example because of the scale, complexity, and control requirements involved. The organization manages the company\u2019s cash across a global financial environment. The group works with more than 100 banking partners, collects about $300 billion from our customers annually, and has about $100 billion in assets under management at any given time. Treasury also works with an outsourcing partner that employs roughly 1,000 people who support business process activities. <\/p>\n\n\n\n

At that scale, the stakes are high for AI adoption.<\/p>\n\n\n\n

Our professionals in Treasury have to balance factors such as speed, business value, regulatory expectations, data quality, and employee adoption when deciding which business tasks could benefit from AI transformation.<\/p>\n\n\n\n

This article explores broader lessons about how exactly our Finance team makes these vital decisions to help operationalize AI opportunities, aiming to produce the greatest AI impact while also protecting our data and preserving our partners\u2019 trust.<\/p>\n\n\n\n

\"A<\/figure>\n\n\n\n
\n

\u201cAlways start with what your process looks like first, not what technology you want to use. We always have a metric, business impact, or business ROI in mind when we launch any AI-related project.\u201d<\/p>\nKathy Brustad, director, Global Treasury and Financial Services<\/cite><\/blockquote>\n\n\n\n

Identifying repetitive and measurable tasks <\/strong><\/h2>\n\n\n\n

According to Kathy Brustad, a director in the Treasury division at Microsoft, the most useful starting point for determining the suitability of AI application is the nature of the work itself. It should include a clearly understood process, a measurable outcome, and a realistic view of risk. <\/p>\n\n\n\n

\u201cAlways start with what your process looks like first, not what technology you want to use,\u201d Brustad says. \u201cWe always have a metric, business impact, or business ROI in mind when we launch any AI-related project.\u201d<\/p>\n\n\n\n

Treasury\u2019s AI work focused on operations where manual effort slowed teams down, especially in areas like collections and risk management. Those processes featured proven business problems and repeatable steps, including case review, information gathering, invoice follow-up, exception routing, and customer communications. <\/p>\n\n\n\n

One Customer Zero<\/a> example, which involved streamlining our finance cash collection with AI<\/a>, followed that same pattern. Case managers were wasting time finding customer contacts, anticipating disputes, and routing exceptions. Information was scattered across different systems and handoffs. <\/p>\n\n\n\n

The team\u2019s response was not simply to add AI on top of a messy workflow. Instead, they consolidated tools and systems into an SAP and Microsoft Dynamics 365 environment, created one source of truth, then layered intelligence into the places where time was being wasted. <\/p>\n\n\n\n

Expanding organizational coverage areas with AI<\/h2>\n\n\n\n

While many AI initiatives begin with productivity improvements, Treasury soon discovered a second source of value: expanding organizational coverage.<\/p>\n\n\n\n

Treasury looked for valuable work that historically could not be covered because the organization lacked capacity, according to Brustad.<\/p>\n\n\n\n

Our Collections organization is one good example. Larger accounts often receive more personal attention. Smaller customers and their invoices may be subject to a standard dunning process (reminders when a payment is missed or fails), but the team did not always have the resources to follow up directly.<\/p>\n\n\n\n

Now, Treasury is evaluating and building AI agents that can identify those invoices and follow up. This solution is not yet in production, but it shows how the team could use AI as a way to extend their reach and increase efficiency.<\/p>\n\n\n\n

A similar pattern appears in the company\u2019s Business Operations work, which includes an AI toolkit for high-volume global operations. That team specifically went looking for work with repeatability and scale, especially manual steps, broken workflows, and disconnected systems. <\/p>\n\n\n\n

AI maturity and technology adoption don\u2019t scale at the same rate<\/h2>\n\n\n\n

The focus of Microsoft\u2019s Finance AI strategy has evolved over time. Early efforts focused on helping employees become familiar with AI assistants and other tools as part of their everyday work.<\/p>\n\n\n\n

As teams gained experience and governance practices matured, Finance teams began targeting more advanced human-led agent experiences. Today, some teams are evaluating more autonomous workflows in which agents can gather information, conduct research, and generate recommendations while humans retain final decision authorities.<\/p>\n\n\n\n

\n

\u201cAI adoption turns out to be the hardest part of this journey, versus just building the technology.\u201d<\/p>\nKathy Brustad, director, Global Treasury and Financial Services<\/cite><\/blockquote>\n\n\n\n

Our Treasury professionals now do AI-driven work across all phases: AI Assistants, human-led agents, and more autonomous agentic tools.<\/p>\n\n\n\n

As one example, with credit-check exception work, AI can gather data, evaluate the case, and send a detailed report to a reviewer. But the human still makes the final decision. <\/p>\n\n\n\n

While technology matured rapidly, another challenge emerged. Technical progress alone did not guarantee business impact.<\/p>\n\n\n\n

\u201cAI adoption turns out to be the hardest part of this journey, versus just building the technology,\u201d Brustad says. \u201cCreating the tools is actually not that difficult. But getting people to establish a new habit and change how they work is a lot more difficult.\u201d <\/p>\n\n\n\n

To support that, Brustad mentioned Finance invested in leadership support, volunteer communities, training programs, and safer experimentation that helped employees recognize AI tools as a way to reduce repetitive work, rather than as a threat to their expertise. <\/p>\n\n\n\n

The goal was not simply to deploy new tools, but to help employees build confidence while using AI in business-critical processes.<\/p>\n\n\n\n

Close the trust gap with verifiable receipts<\/h2>\n\n\n\n

In Finance, making a big error can erode trust quickly. This makes transparency central to how Treasury decides what to automate and how far an agent should go in carrying out important work.<\/p>\n\n\n\n

\n

\u201cThe biggest question when implementing AI or driving this AI transformation is, can I trust it?\u201d<\/p>\nKathy Brustad, director, Global Treasury and Financial Services<\/cite><\/blockquote>\n\n\n\n

Treasury teams focus on using the right data sources, showing where the numbers came from, and explaining key reasoning steps so people can understand the AI\u2019s outputs. That work matters in processes tied to controls, audit evidence, and regulatory requirements. <\/p>\n\n\n\n

\u201cThe biggest question when implementing AI or driving this AI transformation is, can I trust it?\u201d Brustad says. \u201cWe spend a lot of time making sure the numbers are right, making sure we use the right sources, and also being fully transparent with any output, showing where the number came from. We also clearly delineate the reasoning steps that the AI went through to come up with any answer.\u201d<\/p>\n\n\n\n

Brustad says some of the next AI opportunities in finance are more procedural, but equally important: recurring regulatory submissions, gathering evidence for Sarbanes-Oxley Act controls, and other manual reporting tasks that require repetition. The goal is to let agents help with manual, traceable steps while people\u00a0retain\u00a0oversight.\u00a0<\/p>\n\n\n\n

For finance leaders considering implementing AI solutions, the ideal first project is rarely the flashiest one. Rather, it\u2019s where the process is understood, the data is reliable, the risk can be managed, and the value can be measured.<\/p>\n\n\n\n

Our experience across Treasury, finance collections, and Business Operations all points to the same conclusion: Operationalizing AI requires organizations to expand value intentionally, help employees develop new ways of working, and make AI outputs transparent enough to verify and defend.<\/p>\n\n\n\n

\n

Key takeaways<\/h3>\n\n\n\n

Here are some key lessons from our adoption of AI across Microsoft\u2019s finance operations:<\/p>\n\n\n\n