With Copilot, Microsoft has established itself as an early pioneer of AI applications. Would you say you hold an early-mover edge in AI agents?

We are already making sure to bring AI into every business scenario and the community at large. Today, Copilot has become, in some form or shape, the “UI for AI” across a variety of technology offerings and use cases.
When you bring in, for example, an AI Agent as a customer bot, the objective is to improve the customer experience. Agentic AI is the paradigm that will drive more business outcomes, apart from making sure repetitive tasks are more efficiently managed. Copilot is becoming the front and center for not just end-user productivity, but a variety of aspects. Today, we have Microsoft Security Copilot, which is helping make enterprises more secure. We have Copilot to drive more sales efficiency for the front office. Copilots for users like you and me do not just improve how we look at our emails, but also manage our day-to-day work, which is centered around productivity, around documents and spreadsheets, and the content we find online. So, it brings that all together and makes it more contextual

Is Copilot meant mainly for enterprises or also for everyday users? How can end users benefit from AI agents, and will they have as many use cases as GenAI?

AI needs to be democratized and be accessible at a consumer scale. Today, Copilot is available across the spectrum. In some form or shape, AI is moving from being just a digital agent or assistant to a digital teammate. AI has to be pervasive. For instance, during a meeting, a user might jot down notes in OneNote or Word. In the background, there may be systems holding knowledge relevant to them. With the agentic AI paradigm, when the user needs assistance, a digital teammate can step in and recall related content tied to their conversation. If you ask further questions, it also has memory and is context-aware. It’s a point in time where you feel there is somebody learning with you and helping you. In the broader landscape of adoption, Copilot apps have surpassed 100 million monthly active users across the commercial and consumer space. Similarly, the engagement of AI features across products has over 800 million monthly active users. So, the usage itself is the proof of the impact. In a business context, for example, LTI Mindtree is bringing AI Agents as part of their employee productivity landscape. Through AI Agents, they can automate talent mapping and their entire pre- sales process, and achieve a 30% gain in efficiency. Similarly, Persistent Systems is embedding AI Agents into its contract management system. The ContractAssist helps them reduce the negotiation time by 70% and email communication by 95%. Enterprises are now seeing efficiency and reducing the amount of time in getting things done, hence driving more business impact

Internally, how is Microsoft using Agentic AI as a customer zero before rolling it out to customers?

Within Microsoft, we are democratizing in some form knowledge and collaboration, bringing in Copilot, while keeping security first principles in mind, on a day-to-day basis. In fact, as stakeholders in what we’re doing with our customers, we’re also building those scenarios and use cases ourselves. We recently introduced Frontier Firms, an initiative that looks at AI — particularly agentic AI and AI agents — across the broader landscape, focusing on how enterprises will evolve by adopting AI-first and digital-first principles. There are three different phases — the first is where humans will have assistance. We’re already in that phase of moving from digital agents to digital teammates. The next phase is the human-agent team. Imagine you’re a four-member team of finance, marketing, operations, and sales coming together to launch an innovative customer campaign. For a finance team member, the agent can instantly check budget availability for a campaign, a process that earlier took emails and days. Similarly, a sales team member might turn to a digital teammate to understand how many customers could find a product useful based on telemetry, and what past grievances have been. Customer intelligence comes through another digital teammate. So four people now have six more digital teammates. The third phase will have human-led, agent-operated digital processes. In a typical procurement business process, systems are involved. There will be people placing a purchase order, and those servicing that order. While humans will still make the decisions, the entire assembly line of that particular process will be managed by agents

Have the POCs evolved to production-grade projects? Are customers seeing noticeable ROI?

Customers like LTI Mindtree and Persistent Systems are definitely seeing ROI. Microsoft also has a strategic partnership with Apollo Hospitals spanning co-innovation, joint product development and research on topics like disease progression or genomics. We are now taking it to large enterprises and businesses. Similarly, in the space of manufacturing or automotive with the Mahindra Group, we’re planning to develop AI projects that include agentic and multi-model scenarios for automated divisions. While agents have moved from promise to proof, the use cases will depend on the business requirements and growth aspirations. The most important thing when you start a technology project, whether you fund it or bring in people to improve an existing implementation, is knowing the end goal. Businesses usually invest either to improve their bottom line or top line. That means getting more customers and increasing market share, or driving internal efficiency to cut costs and do more with less. Depending on the goal, you then bring in technology implementations to turn that promise into action. The reality is that AI agents have already moved from labs to production. The volume of use cases depends on how you pivot on it