AI and Productivity at Work with Microsoft’s Matt Firestone
Matt Firestone, General Manager of Frontier Function at Microsoft, who leads the company's work on Microsoft 365 Copilot and agents, discusses Microsoft's latest Work Trend Index and its idea of the frontier firm. He explains why many organizations are not yet set up to capture the value of AI, and why he believes the answer lies more in culture and leadership than in technology. The hosts then debate whether companies outside Silicon Valley are ready to reorganize around AI.
Transcript
AI at work has been sold as a productivity revolution. Write the emails faster, analyze the document in seconds, build the deck better, let the computer be your creativity. But Microsoft's latest Work Trend Index points to a little bit of a more complicated story. Yes, workers are using AI, companies are buying the tools, and agents are beginning to take on more of the execution. But their report also suggests that many organizations are not yet built to capture the value.
So the question is no longer just whether AI can help people work faster. It's whether companies know how to redesign work around it. Matt Firestone leads Microsoft's work on Microsoft 365 Copilot and agents and has been closely involved in the company's latest research on what it calls the frontier firm. So today we ask him, is enterprise AI finally moving from experimentation to measurable impact? Or are companies still buying the promise before they know how to change the work?
We are talking the current state of AI, workplace productivity, and the rise of agents with Matt Firestone, General Manager of Frontier Function at Microsoft.
All right, before we get to Matt today, would love to get your thoughts on this because you're coming at this from a CEO perspective, and I think we might have differing opinions.
That never happens, but Stephen, Microsoft's latest research, this index that we're going to talk to Matt about, he's talking about how the productivity is really shifting from the individual to the organization that the company must have in place in order for their AI strategies to be productive. I see you shaking your head.
No, I'm nodding. I am all over this, I think. Couldn't agree more. I mean, I think I want to be a frontier company. Don't you want to be a frontier company?
Right, a frontier firm?
Yeah, we all want to be frontier firms. And what he's talking about is the importance of the firm getting their processes right so that they can build the firm around AI. And people don't like the word process. It sounds really boring, right? But most companies are not ready to do that. And those are the companies that are going to fall behind. But he also talks about value, about how you get value out of, not just your employees, but what's core to your business. And that's what this is all about, building the company around AI, getting the processes right, and unleashing that value. And I would say 80% of firms have got no idea.
I agree with that. Yeah.
You do agree?
I do.
Write this moment down.
Yeah, I mean, I do agree with that because anecdotally I just really hear from a lot of people that they know how to use AI and they in their personal lives use Claude, ChatGPT, Claude Code, et cetera, but they feel totally choked by their firms because of confidentiality, NDAs, risks, they can't actually freely use these models in the way that they want to.
And what I'm really interested in is like the concept of a product versus a substrate, where a product kind of comes in the package, and you are forced to use it in the way it's been given to you. So you can't actually use it in a malleable way. You can't hammer it into those processes and those new processes, a product attaches to existing processes, whereas a substrate, it's like very raw. It's almost like electricity. So, whereas a product is like a lamp. But electricity, you can use it in whatever way you want to and a much, much larger scale than like a hundred lamps, so-
Right, but a CEO can buy a lamp, get it out of the box, and switch it on, but a CEO's not going to buy electricity and turn it into lamps, is it? I think what you're missing is kind of how companies exist today. And the way that companies exist today has not changed that much since the Industrial Revolution, and the amount of change that companies are going to have to embrace to get to this, and the pace that they're going to have to adopt, I see no sign of that happening. What's going to be quite interesting to ask Matt is how companies like Microsoft sitting in Silicon Valley are actually going to be able to help, you know, companies all over the world, because if they can't, then this isn't going to change.
Do you think that companies are going to have to spend more to in order to do that? And that's the reason why?
I don't think it's the money. I think it's entirely reconfiguring, spending money's not difficult, reconfiguring a business is really, really hard. And the skills to be able to reconfigure the business, but even more important than that, people have to realize that that's what they have to do. Right now, people do not realize that's what they have to do. And believe me, that is what they have to do.
But when you're saying people, you're referring to the CEOs or the C-suite executives. The onus is not on me as an employee to make sure that the company that I work for has the infrastructure in place so that I can then utilize these AI products.
No, it is the senior leadership within the company, yeah. And the senior leadership within the companies can find it really hard.
But there's just also so much AI there, where it's like, okay, this is an AI initiative, this is an AI-driven thing. What does that mean? Just that hopefully your stock price will go up because you said you've got an AI thing.
You got an AI strategy-
It doesn't mean anything.
But actually the problem with this is it's quite simple and really difficult. First of all-
Two things can be true at once.
That is so useful.
All right, clearly it's going to be a good conversation with Matt, but we do want to know what you're seeing in your offices. Is AI at work genuinely changing how your organization operates, or is it still mostly experimentation at employees' leisure? If you are using AI tools, where are you seeing the real value? Are you even? Tell us in the comments, like and subscribe. Here now is our conversation with Microsoft's Matt Firestone.
Matt, thanks so much for your time today. Such a pleasure to have you with us. We'll get to the Work Trend Index and everything that we've gleaned from that in just a moment. But first, I kind of want you to explain for us sitting here and also everybody listening and watching what you mean by frontier firm, frontier professionals. Who are those people?
You looked at Silicon Valley for inspiration of AI native building from first principles and they're doing everything with AI. The joke is always, the answer to AI is always adding more AI on top of that. We meet with customers every single day and they ask questions, how do we transform ourselves? How do we raise our own aspirations and ambition to be more like Silicon Valley companies? How do we pull more AI into our companies?
And we created this idea of a frontier firm. It's a mental model. The beautiful thing about the frontier in the English language is one never actually reaches there. You get closer to it and the benchmark and the yardstick always moves. So it was kind of an organizing cry, a rallying cry, a mental model of if you're a company that's five years old, 10 years old, a hundred years old, 500 years old, how do you actually transform and how do you aspire to be more like an AI native company and how do you reach the frontier? And that's actually where the organizing concept came from.
And so when it comes to a frontier professional, you're talking about somebody who is utilizing AI X percentage of the time that they're conducting their work?
Yeah, so two things, so one, the very simple answer is think about the smartest person, the most interesting person, the builder amongst you that comes in on a Monday and they built a prototype, they have a new agent they want to deploy, they have a new workflow, they have a new prompt they want to share. It's that mindset of just building and creating and just wanting to do more. That's what we mean by the frontier professional.
The actual rigorous quantitative data behind that is that based in the WTI report, our Work Trend Index report, we found that 16% of people self-identify as this builder, this maker, this creator, this doer. And a big part of the report is actually speaking to who these frontier professionals are and then giving tips to leaders and organizations of how to unlock their potential and help them diffuse that ability and that knowledge through the organization.
On that note about the report, and I know it talks about obviously reorganizing the company at kind of a systems level and it's kind of the responsibility of the leaders, but what does that actually look like? Because it's clearly not easy, otherwise everyone would be doing it. And I'm also interested in how it compares to, I think the last time we really saw this at such a large scale, which was the Industrial Revolution, where you had factories completely reorganize their workflows to extract value and increase their productivity. So what does that look like in the age of AI?
We use a really fancy fun term in the report called the Transformation Paradox. And it's this idea of you have these frontier professionals and they're doing great things with AI and they come in and they just, they get friction, they get resistance. The organization just can't intake that. So how do you actually change those types of system?
You mentioned the Industrial Revolution. I always think about previous waves of technology as inspiration. I think back to the mobile days of, we had all these debates of how many mobile apps do you actually need? How do you actually go from being a web developer to an app developer? What are those types of skills? I think what's different, to answer your question very directly, is that the pace of change now is just so much more significant.
But I think that first principles thinking of organizational change and behavior, simple things like managers building in the open, encouraging spaces for building, hackathons. One of the cool things about non-deterministic systems is two people can set out to define the same outcome. They can initiate the same prompts and get two different responses. So that iteration and that pairing and that creativity I think is just one of those very simple behaviors we hope that people read the report and can unlock.
Matt, one of the things that I was really interested in in the report is the point that you were making and the point you made just now about how companies are ready or aren't ready to engage with the AI and how companies' processes are either built round or aren't built round, and it seems to me that Silicon Valley might be one thing, but, you know, the rest of the world doesn't look like Silicon Valley. And how are you going to get that message across to kind of just regular companies, regular firms?
The WTI is really a report that speaks to leaders and there was tremendous pressure on leaders in previous years of, you need an AI strategy, you have to answer to your board, you need to have something that cascades down the organization to every functional area, to every room in the house. And what's different about the report this year based on our data is that it's a simple message to leaders of, actually the people around you already know the answers. They're already building in their personal time, they're using it on their phones, they're using it in their personal capacity, they're bringing in those learnings into the workplace.
And it's actually a much more direct message of, your job actually as a leader is to re-architect work, to actually just create this space in the program for them to do the hard work. And it's not about getting out of the way, but it's about actually listening to your teams and using some of those more kind of empathetic skills about culture building.
But most leaders, and I can speak for an audience of one here, aren't trained in how to do that, you know, not trained about how to get out of the way, not trained about how to re-architect space. So how are people going to learn those new skills in a way?
I mean I would argue that, we mentioned the Industrial Revolution before, I mentioned mobile before. I think leaders have paradigms and frameworks and best practices about organizational change. You don't actually necessarily need to be an expert in AI. You need to be an expert in what are the types of behaviors and patterns and best practices. I think what is different, and where a degree of humility is required, is that the speed of change is just more significant than pathways. But I don't think those fundamental instincts of how to build the right organization are necessarily different in this era.
Before we go to break, can I ask you quickly, do you think that the reason that so many of these institutions and companies are not implementing that new type of infrastructure so that they can have a successful AI strategy, do you think that they're not doing so because either they don't think that their company will truly benefit from an AI strategy or they don't even know where to begin?
You know, we had this data in the report about individual workers actually feel that if they use AI for a solution, they're not actually rewarded by their manager and organization. And it's incredible when you think about that. People put individual pressure on themselves to do things with AI, but if they have a successful outcome, they're not rewarded for it by an organization. And like that comes down to, it's not about a technological barrier, it's about a cultural change and a mind shift.
And I keep coming back to what was so different about this report for us was we grounded it in telemetry. We had trillions of aggregate and anonymized signals across the M365 ecosystem. We studied what people were using Copilot for in the workplace. And it was such a surprising realization for us that so much of this came down to organizational psychology and best practices and about mindset shift. And that's where this message of, I think optimism comes from, is that you actually know how to do this as a leader, and it's about limiting the pressure of yourselves and actually just unlocking individual agency and helping people raise ambition.
And we're going to talk about the employees coming up on the other side of the break. We're going to ask Matt Firestone whether companies are ready for AI agents, and what happens to human judgment, accountability, and career development when AI starts taking on more of the work?
All right, back with Matt Firestone from Microsoft, and we've been talking about the company side of implementing a successful AI strategy, but we want to talk about the employees because one of the conferring narratives surrounding AI is AI is going to take my job. And that seems to be true when it comes to some of those entry level positions, more of that administrative work, drafting emails, creating summaries, building, you know, presentation decks. So I guess my question for you is, if AI does take over more of that work, what happens to all of these, you know, college graduates that are graduating right now when it comes to those entry-level positions? And also how do they work their way up the career ladder if they don't even have a place to start?
Think about the first job you had at a university or a college. It's a lot of rote tasks, it's a lot of drafting very simple straightforward emails, very basic templates in PowerPoint, you're doing the same thing over and over again. There's actually not much space for higher level intellectual work. What was really surprising about the hard data telemetry from the report is 49% of the usage of M365 Copilot was higher-order cognitive tasks. So things like deep data analysis, asking multi-shot questions, more sophisticated types of long-running research.
And when you think about this, it actually allows more people to raise the ambition about what they're doing and take on new orders and new types of jobs. So back to the question of new college grads, this is the type of tool and technology that actually raises the level and the scope and the breadth and the depth of what one can do in those first couple years out of university, which I think is a lot more fun than some of those kind of early level jobs that we may have had right out of university.
I think that's so interesting you say that, Matt, because I think we are finding even within our own company that we've been using AI to actually do the hardest task, which like at scale takes a really, really, really long time. And actually it's kind of left the easier, not so crazy task to be done by us. But I know you said previously about friction for employees like the frontier professional in companies.
And I want to come onto the difference between like a product and a substrate, where I've anecdotally heard a lot from people that a product, where they are being delivered AI kind of in a package where they can't be malleable with it, is actually where they're finding the most friction in their organizations, where they can't give it to the data they need to be able to give it, they can't give it the contextual knowledge they need to be able to extract the most value out of it. So what would you advise leaders to do to help AI become less of a product but more of a substrate, which their frontier professionals can build new workflows and processes around?
We had a really interesting finding in the WTI about the four types of usage. I guess on the X axis, it's how much is the human involved? And on the Y axis is how much is the AI or agent involved? And it's almost like the frontier professionals, it's like a Harry Potter sorting hat. They look at each individual task and they decide, is this a quick response? Is this just, I ask a question, I get the response back? And then we see some of that early exploration is actually about taking on more ambitious tasks that require access to more types of data or more types of tooling or more types of systems.
When you get into that upper right hand quadrant, that's about these kind of long-running tasks. You see it today in software development where you define early ambitious outcome and maybe that agent is running for four or five hours to define and deliver on a really ambitious outcome. And I just share that as a very direct example of, as you start thinking about what are the types of outcomes you can take on, it's easier to start advocating for more access for more types of data, for just larger projects. And this is what I mean by raising individual ambition.
I can give a quick example. I'm a student of history, of tech history, and I look at what software developers are doing, and you know, in a 90 day kind of sprint now, it's not just about future delivery, but it's entire products or entire platforms, it's very outcomes-based thinking. And I think about what is outcomes-based thinking in the context of knowledge work, it's not just a rote email or a rote deck, but it's actually trying to solve a really clear business problem, defining that outcome, and then actually using an AI to not just take the outcome of that system and use that as an answer, but to refine and to iterate and to ask more questions and be more ambitious and then pull in more data to your point. And that's how you bring in some of this kind of outcomes-based thinking into knowledge work. And we see a lot of early evidence that's starting to happen with our frontier professional cohort.
One of the things that I was really interested in the report is when you talk about value and you talk about unlocking value in in humans and within the organization itself. And I think one of the things that you say, it's very important for an organization to know what value it brings. And so I wonder if you could touch on that. How does a company on unlock value, both in its employees and in of itself?
You know, one of the things we talk a lot in the report is about the corresponding piece with agency is also about taste and judgment and applying aesthetics in the Silicon Valley terminology. And what this actually means is that a business has a very clear outcome, and you can work towards that by deploying this technology. And what's really wonderful about these types of systems is that the sensibility of, did I get the right type of outcome? How do I broaden the reach from there? How can I raise that level? These are the types of things that humans bring uniquely to this process. And I think about it very simply of like a job is just a collection of tasks, and the ability to actually architect that in really creative ways I think is what we're seeing early in the report.
Before we get to our rapid fire questions, I want to ask you about something and whether or not it gave you pause that you reported in your index, which is a significant percentage of the frontier professionals that you surveyed said that they refuse to outsource their thinking, knowing that their long-term career success is going to be based on continuing to build human skills. So they're aware that they need to keep their cognitive skills sharp. As somebody from Microsoft who I'm assuming you want to get your product into the hands of as many people as possible and make sure they're using it as much as possible, does that concern you at all that these frontier professionals even have a line where they're saying, "I probably need to pull back a little bit on how much I rely on AI?"
I'd say two things. I'd say one, there was this idea a year ago that using AI would just result in cognitive atrophy. That you're just kind of offloading your thinking to the systems. I go back to that 49% stat, we're seeing the actual opposite in our report, is that it's not just about rote tasks, but people are taking on higher-order cognitive functions. And I think that's just a very sharp data point for us that says what the meme of the last couple years was is actually not what's bearing out in the data today.
And then the second thing is, I go back to that idea of the Harry Potter sorting hat of the four types of behaviors that we're seeing. The fact that people now can take on more complex outcomes, they can use AI and agents to do higher-order running tasks, means that they can do more ambitious things. There's an IDC report that says by 2029, there's going to be a billion agents deployed in the workplace, a billion agents. It's such a crazy number. And it goes back to these early debates in mobile about how many apps does one actually need on the phone back then, was it 10, is it a hundred? And now there's thousands of apps in people's phones, millions of apps in marketplaces.
And I share all that of, we saw a 15X year on year growth in agents across our ecosystem. And then enterprises, which you'd imagine would be in some ways laggers or slower adopters, the number actually rises to 18X year on year increase. So back to the question of the idea of, if we're all becoming agent bosses and we're all becoming these agentic orchestrators and we're taking on more outcomes and we're deploying more agents, this is higher level cognitive thinking. This is people elevating their skillset to be leaders and managers earlier on in their careers. This is what it means to have really ambitious outcomes and then use these agents, these fleets of agents, to take on a higher order of work.
I totally agree-
Opposite of brain mush.
I feel like my job is literally asking an agent to work harder on a daily basis. But yeah, and just one final question from me, and we've spoken about what the report said and examples of data points from other companies, but what about Microsoft itself? How are you seeing this play out within Microsoft? Where are you seeing bigger gains? Where are you seeing, you know, you guys innovating? And where are you seeing the most agents?
Yeah, I can give a formal example and then just a casual example. I think the formal example is we have a customer zero program, in the Silicon Valley parlance that's eating our own dog food. We deploy agents, we do process transformation, we have the benefit of talking to customers around the world and we always look internally to best practices to share.
I think about my own team, what we do on a basis, carving out even just something simple, like what is the hackathon equivalent of knowledge work? It's carving out a couple hours on a Thursday afternoon to an outcome that traditionally we would've addressed with rote process. How can we solve it with an agent? How can we build something against it? It's a really fun toy example, but we used Copilot Cowork recently to figure out what our travel was and we gamified it and we made leaderboards and we got to see actually who was where and who was what. And you could stack rank and actually see like who put the most amount of miles on the road and you could draw an interactive map. And then think about how powerful that is.
Another fun toy example is, as a leader, I built a diffusion score to see, am I being equitable with my time? Am I spending time with this team versus that team? I used calendar data, I calculated the value of a one-on-one versus a team meeting to give me insights. And then I was actually able to automate that and then get reminders every Monday morning of, "Hey, like you should spend more time with this team. They haven't seen you recently." And it's just really a fun example.
All right, quick rapid fire questions for you. Mine is, where do you think people will draw the line when it comes to bringing AI into their personal lives?
Great question. I think people accept AI anywhere as long as they can credibly find themselves in the output. The moment something feels ghost written and not just like assisted, then the trust collapses.
Totally, and my question is, what is something that's universally accepted in your specialist field that you actually disagree with?
It's kind of the thesis of the WTI, which is, I think most pilots are designed to fail safely. They just prove that technology works, but they don't actually measure and test that the organization itself can change.
My question, Matt, which I don't think is going to be difficult for you, is what are you optimistic about?
I actually think it's more fun to be an optimist than a pessimist. But I'll say, I'll say very directly, I think we're going to see organizational designs that we don't even have names for yet. I think the current shape of organizations today solves for individual bandwidth, but they don't necessarily solve for ambition and potential.
Interesting. All right, lovely note to end on, all right. It is much easier to be an optimist. Matt, thank you so much for your time today. It's been a great conversation.
Thank you very much, Matt, thank you, thank you.
Thank you, Matt.
As a CEO, your thoughts on how Matt's essentially saying in order for an AI strategy to be successful, the company's going to have to reorg.
Yeah, right, so I loved Matt, and I really buy into everything that he's said and I completely agree with that. You know, the companies are going to have to reorganize themselves, reorganize themselves around AI, understand their processes, all that's going to change. But he made that sound easy.
He did. And not expensive.
All the CEO's got to do is get out of the way and, you know, and let a thousand flowers bloom. So I completely agree with the prognosis that everything process is going to have to be changed, everything's going to have to be changed, the company's going to have to change, but I think it's going to be really difficult for most companies that aren't in Silicon Valley to do that. So I think there's a big communication gap here.
I totally agree. I just don't think tech, like, communities realize how not engaged, because we're all on tech Twitter, tech X, and everyone's saying this new thing, if you've not seen it, you're so behind. And then suddenly you come out into the real world and you're like, "Have you heard of Claude?" And someone is like, "Who?" And honestly that is the reality amongst, especially leadership, because why would they be? You know, things have worked fine as they are. If it ain't broke, don't fix it. Like it's very much that mentality. So expecting people to just kind of wake up and be like, "Yeah, I guess-"
I'll turn my company upside down.
And you know, your question was a great question about sort of individual employees, you know, and I'm not sure how many are thinking, "Yeah, it's going to be great because my job's going to get a whole lot more interesting because I'm going to be able to deploy these agents." I just don't think that's what people are thinking.
No, and honestly, I think I am amongst a minority where I'm like, "Is your company not doing this? Is your CEO not on top of this?" And people are like, "No, like you are on your own there." And I'm like, "Wow, that's quite backwards. I'm really sorry for you."
You've got a frontier professional here.
And people are like, "You're so lucky that you don't have to, you know, really push your way through it." As in it is easy, but that is definitely not-
This is not the norm.
And even in like the most successful big supposed frontier corporate company, like, it's just not happening.
So what does worry me about this though is that the price of failure is very high here, is that some organizations will do this, some organizations won't, and the cost to those organizations that won't is going to be huge. And I don't think as a society we're ready for that.
And it's going to be even bigger because you just blink and then it's over, it's absolute wreckage, it's not like.
It's the point that Matt was making about the pace of change here is different to anything that's been before, right?
Yeah, yeah, totally.
And again, we do want to know what your workplace looks like. Has your company successfully implemented AI strategies that are supported by the proper framework, or are they woefully under-prepared on how to execute AI in your industry? Share some successes and some shortcomings in the comments below. And then be sure to subscribe so that you never miss an episode. We'll see you next time on "Agents of Tech."