The Future of Work Has No Org Chart: Microsoft's Blueprint for Turning AI Pilots Into Real Transformation
The short answer
The best way to think about AI at work is not as a tool that automates tasks, but as a business transformation that never ends. In a recent episode of Microsoft's WorkLab podcast, hosted by Molly Wood, Corporate Vice President of Workforce Transformation Katy George argues that companies stuck in pilot purgatory are treating AI like a product launch when they should be redesigning whole systems of work alongside the people doing the work. The payoff is what Microsoft calls "capability add" — AI injecting brand-new capabilities into a process — and it eventually dissolves the org chart into fluid, skills-based teams.
Why AI adoption stalls: the identity problem
George says the adoption gap is less about technology and more about professional identity. Software engineers grew up as "master craftspeople in coding," and now they are being told they no longer have to code — they are product builders, end-to-end builders. That is a real mindset shift, and bringing people along has to be structured, not left to chance.
Microsoft runs Camp AI, a team-based boot camp where whole teams go together rather than individuals squeezing in solo training. In the first week everyone — product managers, engineers, designers — gets immersed in the tools with the instruction to "leave your discipline at the door." By week two they are applying the tools to their own work with practitioner-to-practitioner coaching. Engineers who walked in with trepidation walk out as evangelists: "I never want to work the other way again."
The "go learn these skills" exhortation puts the burden on the employee; real change is organizational. Only when you change the whole system of work do you get different business performance. Role-based programs like weekly huddles, where everyone in the same role compares how they used AI to get better at the job, drive significantly more effective adoption than generic training.
Treat AI like a business transformation, not a product launch
George's first piece of advice, shaped by her years at McKinsey, is to realize this is business transformation, not a product launch. Microsoft's initial Copilot rollout to its own sales organization was treated as a typical launch — here is a new product, we will teach you the product — and people tried it, then stopped using it. What finally worked was setting clear business outcome goals and redesigning work with the people doing it, using tools like Kaizens, Gemba walks, and value stream maps to bring tech, process, and people together.
Knowledge work makes this uniquely hard. In manufacturing you could watch the widgets move down the line and see exactly where they got delayed; in knowledge work, so much is tacit and unwritten that understanding the workflow itself becomes an art and a science.
Stop measuring AI like an automation tool
Most economic models of AI take job descriptions, break them into tasks, assume roughly 30% can be automated, and drop that to the bottom line. "That's not how it's happening," George says. AI does automate tasks — agentic coding is the dramatic example — but its bigger value is "capability add": adding whole new capabilities rather than doing the same process faster. Microsoft's internal audit team found productivity savings, but what they value most is that every internal audit now delivers proactive risk identification, something human beings never used to do. That is a new capability AI injected into the process, not a substitution.
The metrics that matter: adoption → behavior → outcomes
Measuring ROI gets harder when AI introduces new variables like growth, innovation, and customer experience. George's team maps adoption (the easiest thing to measure) to behavior change to business outcomes. In the sales organization, role-based change management increased usage; the question became whether reps were spending more time with customers and demoing the latest portfolio, and whether that moved win rates, qualified pipeline, and revenue per rep. Microsoft sees a nice correlation between the reps using AI the most and revenue per rep. Customers who moved past piloting now focus on a few end-to-end business processes that are "more boring" than the cool demos but make the biggest difference to business performance.
What a Frontier Firm actually looks like
In Microsoft's partnership with Harvard Business School, a "Frontier Firm" is fully AI-powered — decisions informed by AI, operations sometimes run by AI — but remains human-centered and human-led. These organizations feel like living organisms, constantly learning, with hierarchy and job titles mattering much less. George predicts we will finally get out of the 1900s matrixed org chart. Leaders become continuous experimentation officers, roles turn T-shaped (deep expertise plus horizontal integration across the whole process), and teams become dynamic and fluid. Satya Nadella's framing: identify your "private evals" — the secret sauce of your company — and accelerate exactly that with AI.
Who needs to be in the room
Business leadership has to sponsor and actively engage; this is not a tech project a CIO leads alone. One customer CEO put it memorably: "I am not leading AI transformation. I am leading a business transformation and AI is helping me." You also need a diverse cross-disciplinary mix — tech, process capability, and people understanding — working together rather than as a handoff. And the people doing the work must be there to redesign their own work.
What to do tomorrow morning
George's lightning-round advice for leaders: get clear on the most exciting, ambitious goal in your organization, and therefore which processes and success metrics to focus on. Then build cross-generational teams, pairing AI-native early-career people with those who understand what great looks like. Microsoft's engineering "PRAISE" program is a two-way apprenticeship model for AI in software engineering. For individuals: experiment and play — the tools change so fast that something you tried months ago may amaze you now — and talk to people obsessively about what is working. A colleague showed George that Copilot can analyze your calendar against your priorities, something she had been doing by hand in a notebook on planes for years. Her reaction: "Duh." Learning AI better starts with talking to more humans.
Verdict: the org chart is the last thing to keep
Across Microsoft's 100 internal case studies of AI transformation, the consistent pattern is a combination of top-down clarity and bottom-up grassroots innovation: clear goals cascaded from leadership, with citizen developers building things that scale to the whole organization. That combination, George says, is where the magic happens. The companies that treat AI as a permanent capability add, redesign work with the people doing it, and organize around skills rather than boxes will be the ones that make the org chart disappear.
Sources
Microsoft WorkLab: Katy George — Ditch the org chart, your team's future is fluid
Frequently Asked Questions
Why do AI pilots fail to turn into real transformations?
Microsoft's Katy George says most companies treat AI like a product launch instead of a business transformation, so people try the tools and then stop using them.
What is a "Frontier Firm"?
A Frontier Firm is a fully AI-powered organization — decisions informed by AI, operations sometimes run by AI — that stays human-centered and human-led, with hierarchy and job titles mattering much less.
What does Microsoft's Katy George do?
She is Corporate Vice President of Workforce Transformation at Microsoft, leading the company's research on how AI changes work and how organizations can transform successfully.
Is AI adoption mainly a technology problem?
No — George describes it as an identity and mindset problem: people fear losing their jobs, and professionals like engineers must shift from "master craftspeople in coding" to product builders.
What is "capability add" in AI?
Capability add means AI injecting whole new capabilities into a process — like internal audits delivering proactive risk identification — rather than just doing the same process faster.
How does Microsoft measure AI ROI?
Microsoft maps a chain from adoption to behavior change to business outcomes — for example, sales reps using AI more spend more time with customers, which correlates with higher revenue per rep.
What is Camp AI at Microsoft?
Camp AI is Microsoft's team-based immersive boot camp where whole teams learn AI tools together, leave their discipline at the door, and apply the tools to their own work with practitioner coaching.
Will AI replace jobs or change them?
George's view is that AI changes jobs constantly — "we are both doing our jobs and changing our jobs" — and that it automates tasks while adding new capabilities rather than replacing whole roles.
What is the first thing a leader should do with AI tomorrow?
Get clear on the most exciting, ambitious goal in the organization, and therefore which processes and success metrics to focus on.
Why do job titles matter less in AI-powered organizations?
In Frontier Firms, work is organized around skills and fluid teams rather than a 1900s matrixed org chart, because hierarchy matters less when anyone can build and create with AI.
What skills do AI-era leaders need?
Leaders must become continuous experimentation officers, comfortable working end to end, focused on where value comes from, and able to cascade clear outcome goals.
How should companies help employees learn AI?
Companies should create structured, team-based and role-based programs like Camp AI and weekly role huddles, rather than leaving individual employees to learn on their own.
What is a two-way apprenticeship in AI?
Microsoft's engineering "PRAISE" program pairs early-career, AI-native people with experienced folks who understand what great looks like, so both generations learn from each other.
Is AI ROI about saving money or adding value?
George argues the bigger prize is capability add and competitive advantage, not labor-cost efficiency, because most companies do not compete on headcount savings.
What should an individual employee do to get started with AI?
Experiment and play — tools change so fast that old attempts deserve retry — and talk obsessively with peers about what works, since learning AI starts with talking to more humans.
