Illustration by Alex Robbins
Think about stepping onto a steamship alongside the Hudson River in 1900, setting off on a transatlantic journey that may take a full week. You flip to a fellow passenger and say, “In only a few many years, this identical journey will take seven hours—and we’ll do it within the sky.” You’d probably be met with well mannered disbelief, possibly even a chuckle. The thought would sound extra like fantasy than foresight. And but, that’s precisely what occurred. What as soon as appeared inconceivable rapidly grew to become the brand new regular.
And that’s about to occur with AI and enterprise, however with as soon as unimaginable progress compressed into a number of years as a substitute of many years. Many organizations are within the preliminary part of AI transformation, with people tapping AI as an assistant. Some have already reached the second part, forming groups of people and brokers working collectively towards frequent objectives.
However part three, the Frontier Agency, the place people empower AI to turn into the operational engine of enterprise—and brokers turn into the first producers of labor—is more durable to examine. It’s “imagining the inconceivable” territory. And the primary barrier to getting there’s not the bounds of expertise; it’s the problem of imagining a completely new solution to work and construction organizations, then determining the right way to execute that transformation.
The three phases of AI transformation
As we mentioned final month within the 2025 Work Development Index Annual Report, we see organizations shifting by way of three phases on their solution to turning into fully new companies, powered by intelligence on demand. Data staff will not deal with all information work, with brokers executing increasingly of it. Human roles will refocus round imaginative and prescient, technique, relationship constructing, and managing brokers.
At Microsoft, we’re not simply speaking about this transformation—we’re residing it. Our gross sales group is actively navigating all three phases directly. By testing and scaling new instruments like Microsoft 365 Copilot, Gross sales Chat, and our autonomous Gross sales Agent, we’re studying what works, the place the friction factors are, and the right way to evolve each our expertise and our methods of working. These inner experiences are serving to us refine the trail ahead—not only for ourselves, however for each group on this journey.
“The core of what makes a great vendor hasn’t modified,” says Pam Maynard, our Chief AI Transformation Officer for Microsoft Business. “They deeply perceive the client’s context, priorities, and ache factors, after which map our options to these wants.” More and more, that may come from hybrid groups of people + brokers.
Section 1: Human with assistant
Each worker makes use of AI as a private assistant to work higher and quicker—writing, analyzing information, producing code. That is about productiveness on a private degree relatively than staff or organizational. At scale, the impression means organizations working in the identical manner, utilizing the identical processes however extra effectively. In part one, people nonetheless drive the work.
For our gross sales group, this step has targeted on high-volume duties: the executive work that sellers typically repeat dozens of instances a day, like updating CRM information or taking assembly notes. “Copilot tackles the executive trivialities so sellers can spend extra time connecting with clients,” Maynard says.
The outcomes? One gross sales staff of almost 700 folks has already reported 9.4% increased income per individual and 20% extra offers closed in comparison with their friends with decrease Copilot utilization1—concrete positive aspects that present how AI is driving actual enterprise outcomes.
Section 2: Human-agent groups
Brokers be part of the staff as digital colleagues, taking up particular duties at human course—triaging assist tickets, dealing with challenge administration, facilitating conferences, contributing to brainstorming classes. Brokers increase effectivity and productiveness on the staff degree, increase staff with new expertise, and free them as much as tackle new, extra useful duties.
At Microsoft, this shift is embodied in Gross sales Chat, an extension of Copilot that brings wealthy CRM and gross sales context instantly into the vendor’s workflow. As an alternative of toggling between dozens of instruments to arrange for a buyer assembly, sellers can now get prompt entry to detailed account and alternative data, together with licensing particulars, renewal timelines, and key decision-makers. It additionally surfaces high-level account summaries and income breakdowns, offers real-time teaching, and offers predictive insights to forecast deal outcomes and recommend subsequent steps. “Earlier than Gross sales Chat, sellers needed to swivel between 20 completely different instruments simply to get the total image,” Maynard says. “Now they will merely ask, ‘What do I have to know earlier than my assembly with this buyer?’ and get all the things in a single place. It factors them straight to the highest-value motion they will take.”
Section 3: Human-led, agent-operated
People outline technique and assign objectives to brokers. These brokers work principally with different brokers to carry out complicated operations, escalating points to people for steerage when obligatory. Brokers will design workflows and even consider each other. People will monitor their progress by way of dashboards that summarize agentic actions, intervening just for crucial choices. We’ll see the emergence of a brand new profession alternative for people within the AI age: agent supervisor, charged with constructing, delegating to, and supervising digital colleagues.
Progress by way of these phases gained’t essentially be linear—many firms will discover themselves in multiple on the identical time—and there will probably be a “jagged frontier” of AI transformation as you roll out part three approaches throughout tasks and groups. That’s a part of progress, in any case—folks crossed the Atlantic by way of steamship nicely past the daybreak of the aviation age.
The part 3 frontier
Letting go of a extra hands-on human position in day-to-day operations will really feel daunting—particularly in these early days. And being the primary mover will all the time really feel dangerous, however granting machines such unprecedented autonomy requires a brand new degree of belief in our instruments, and an understanding of what they will and can’t do nicely.
However with a little bit creativeness you’ll be able to transfer ahead boldly and with manageable threat by taking a step-by-step strategy: experimenting on low-stakes tasks, verifying that they’re working nicely, documenting your learnings, after which scaling them.
For instance, many firms rent gross sales growth representatives to do the mandatory however repetitive work of prospecting potential clients. An agent can tackle that entry-level job in a territory the place you weren’t planning to rent a human anyway. It’s the distinction between hiring a digital employee—and discovering worth in an in any other case uncovered market—or no employee in any respect.
Our gross sales staff makes use of our new Gross sales Agent to deal with precisely that kind of labor. It may analysis leads, arrange conferences, and e mail clients. A human vendor takes over when it’s time to shut the deal, similar to they’d with any early-in-career gross sales growth rep. (Ultimately, the agent will be capable of shut the deal itself.)
We’ve began in our small-to-medium enterprise phase, the place the size of alternative far exceeds what human sellers can cowl. “There may be completely zero probability we may successfully cowl that chance with people,” Maynard says. “It’s not about changing sellers—it’s about unlocking worth that was beforehand out of attain.” Many of those clients are already utilizing Microsoft merchandise however haven’t had a direct relationship with a vendor. Gross sales Agent modifications that, proactively figuring out wants, providing tailor-made options, and bettering the client expertise in a manner that’s each scalable and cost-effective. In simply the previous three months, the agent has reached out to 36,000 prospects to generate gross sales leads, changing 10.4% into gross sales alternatives.2
Section three remains to be early days for us. Even so, the shift is already giving rise to a brand new idea: the AI territory. “In a conventional territory, you assign a vendor. In an AI territory, you assign an agent that may function independently and ship worth,” Maynard says. “It’s a brand new enterprise mannequin for scale.”
A brand new perspective on AI and belief
It’s nicely established that we readily settle for the dangers that include human errors whereas holding machines to a a lot increased customary (self-driving vehicles, for instance). Within the context of enterprise, we fear AI may say or do the flawed factor, although we tolerate these identical flaws in folks every single day. This uneven threat tolerance deserves examination, particularly since AI methods can already do far more than we allow them to.
Take into account how OpenAI o3’s capability to cause has introduced breathtaking progress on key intelligence benchmarks. With AI creating so quickly, is it extra of a threat to take a leap or to maintain ready? Bear in mind: firms have numerous methods in place to mitigate threat attributable to people, and so they can and can develop them for AI as nicely.
What’s extra, no firm is touring the trail to turning into a Frontier Agency for its personal sake. Section three–model transformation gained’t simply impression organizational constructions; it should additionally drive progress. As you progress by way of this journey, the price of human labor will lower whereas digital labor prices rise, albeit marginally. Over time, new income streams will emerge, resulting in an general enhance in revenue margins.
Section three means shifting from directing folks to designing methods. Management, in the meantime, shifts past managing folks to orchestrating efficiency. To get began, take a look at a brand new strategy: select one course of in an space of low threat and robust potential, and take a look at making it agent-led.
Worrying in regards to the unknown is comprehensible. However the true threat right here is inaction. The groundbreaking Frontier Corporations that may reimagine information work and dominate their industries within the coming decade are already pulling forward.
