The future of work has no org chart | Microsoft Katy George

George & Wood (WorkLab) — the future of work has no org chart

▶ Watch on YouTube · Microsoft · 28:18

Rigid org charts are becoming a barrier to AI-fueled business impact.

Microsoft’s CVP of Workforce Transformation Katy George joins WorkLab host Molly Wood to share how forward-thinking companies are breaking down silos, building teams around skills, and embracing constant change.

Discover why adaptability, curiosity, and a willingness to reinvent yourself are the new must-haves for leaders and employees—and how your organization can unleash human capability in the AI era.

TL;DR

A 28-minute episode of WorkLab, Microsoft’s podcast, published 27 July 2026. Host Molly Wood interviews Katy George, Microsoft’s corporate vice president of workforce transformation. Her team sits in the Office of Strategy and Transformation and acts as “a learning accelerator” across the company; before Microsoft she worked at McKinsey, much of it in manufacturing. Five claims:

1. AI is a business transformation, not a product launch. Microsoft first rolled Copilot out to its own sales force “as a typical product launch… And people tried it and then they stopped using it.” What worked was setting business outcome goals and redesigning the work with the people who do it.
2. Change the team and the system, not the individual. “Only if you change the whole system of work do you get different business performance.” Two programmes: Camp Air, an immersive bootcamp that whole teams attend together, and weekly role-based huddles taught practitioner to practitioner.
3. “Capability Add” over automation. The economists’ task model (“it’s always 30% on average”, drop it to the bottom line) “is not how it’s happening.” The larger value is new capabilities: “every internal audit now delivers proactive risk identification.”
4. Measure a chain, not usage. Adoption, then behaviour change, then business outcome: in sales, more time with customers, then win rates and pipeline, then revenue per rep. She reports “a really nice correlation between the reps who are using AI the most, and revenue per rep.”
5. Skills, not org charts. In a Frontier Firm, “hierarchy and job titles matter much less”; roles become “more T-shaped”, and leaders “have to become continuous experimentation officers.”

Key claims

The adoption gap is partly an identity problem

Wood asks why adoption lags a maturing technology. George’s answer is fear of losing the job, or of no longer being good at it. Microsoft’s software engineers “have grown up as master craftspeople in coding. That’s their professional identity… And now we’re telling them you no longer have to code. Now you are a product builder.” Builder may sound better than coder, “but it is a real mindset shift.” Her two working beliefs: change that drives sustainable results also makes the workplace better for people, and “the only way you can effectively redesign work… is by involving the people who actually do the work,” more so for knowledge work, where “so much of what we do… is tacit and it’s not written down anywhere.”

Teams, not individuals: Camp Air and role-based huddles

Wood notes that the show has often told listeners to go and learn the tools themselves, which puts “the burden of keeping up” on the employee. George describes two structured alternatives.

• Camp Air began in engineering and is spreading to other functions. Whole teams leave production work for an immersive bootcamp. Week one is mindset and immersion in all the tools relevant to the team’s work: “leave your discipline at the door,” whether product manager, engineer or designer. In week two they apply it to their own work, with coaching from other practitioners. In exit interviews, engineers who arrived with trepidation say “I never want to work the other way again.”
• Role-based programmes, used in the sales organisation. Everyone in the same role meets in a weekly huddle and compares “not on what experiments did you do with AI in general? But how have you used AI to get better at our job?” It is practitioner-taught, not run by an outside facilitator. She reports “significantly greater adoption, but also much more effective adoption”: behaviour change linked to performance.

A business transformation, not a product launch

George cites Microsoft’s collaboration with Harvard Business School on the Frontier Firm research, which “did a case example of our initial rollout of Copilot into our own Salesforce.” The product-launch rollout failed; what replaced it was clear business outcome goals and redesign of tech, process and people together, using Kaizens, Gemba walks, value-stream maps and day-in-the-life analysis. Her manufacturing background explains why knowledge work is harder: on a production line “you could see the widgets going down the line,” where they got delayed and where inventory built up. “In knowledge work, we don’t see any of that.”

From her team’s 100 case studies of AI transformation inside Microsoft, the consistent pattern is “top down, bottom up”: leaders cascade clear goals, name the processes that matter and fund both learning time and pro-code development, while “grassroots, citizen developer” builds are scaled to the whole organisation. Customers tell her the same about pilots: “a year ago we were just doing lots and lots of pilots… that’s a mistake. We’re now focusing on three end to end business processes that are much more boring.” One customer CEO: “I am not leading AI transformation… I am leading a business transformation and AI is helping me.”

Capability Add, and why ROI is hard to measure

She expected measurement to be straightforward because “we all thought of AI as primarily an automation tool.” The models economists use break jobs into tasks and estimate the automatable share, “always 30% on average”; you drop that to the bottom line and reorganise around the remaining 70%. Instead, AI automates “different pieces of people’s jobs, not whole jobs,” and adds more value through “Capability Add”: “it’s actually adding whole new capabilities in.”

Her example is Microsoft’s internal audit team. It found productivity savings, some used to avoid replacing people lost to attrition and some to extend audit coverage. “But what they would say is far more valuable than that, is that because of AI, every internal audit now delivers proactive risk identification… a new capability that AI has injected into the process.” Teams are asked where they should be doing “more decision analysis? More scenario work?”, more preventive quality thinking, more for the customer.

That makes ROI harder, because the value shows up as growth, innovation and customer and employee experience. “Most companies do not compete on labor cost efficiency.” Microsoft is working on standards for those metrics and maps a chain from adoption (“the easiest thing to measure”) through behaviour change to outcomes. In sales: are reps spending more time with customers and on the newest products, does that raise win rates and qualified pipeline, and does it reach revenue per rep? She quotes Satya Nadella’s question “what are your private evals?”, meaning what is the company’s own competitive advantage that AI should accelerate.

The Frontier Firm, roles and leaders

A Frontier Firm is “fully AI powered” in decisions and sometimes operations but “very much human centered and human led.” A colleague’s image: the organisation will feel “like a living organism and constantly learning”. She expects to “finally get out of the… 1900s… matrixed org chart” and to form teams around skills for projects.

For roles, “roles are becoming more T-shaped.” Expertise and judgment are “maybe even more important than before”: the view of a year ago, “that we will never need human expertise again,” “turns out to be false. To use AI really effectively, you need judgment, taste, expertise,” and people then integrate horizontally across the process, so teams become “much more dynamic and fluid.” Leaders “have to become continuous experimentation officers,” need more business understanding, and must cascade outcome goals, because “as we are all empowered… to build and create, we need to know what we’re building for.”

Who needs to be in the room: business leadership as an active sponsor (“This is not a tech project… not something that… your CIO leads on their own”), with power users as leaders; a cross-disciplinary team of technology, process improvement and people expertise, “It can’t be a handoff”; and the people who do the work.

Concrete advice

• For a leader moving from pilots to transformation: name the most ambitious goal and therefore the processes and metrics to focus on; build cross-generational teams that pair AI-native early-career people with people who know “what great looks like.” Microsoft engineering runs two-way apprenticeships, including a programme called “Praise”, “preceptorship for AI in software engineering.”
• For an individual contributor: experiment, and retry tools tried months ago; talk to people constantly about what works, in the same role and outside it; ask the AI itself for tips. Her own example: a colleague asks Copilot to audit how her calendar matches her priorities, a review George had always done by hand on planes.

Dynamic-capabilities reading

• digital-transforming/redesigning-internal-structures — skills-based teams instead of the matrixed org chart, T-shaped roles, cross-disciplinary transformation teams, and two-way apprenticeships across generations.
• digital-transforming/improving-digital-maturity — Camp Air and the role-based huddles are structured, team-level capability building, with exit interviews and adoption-to-outcome metrics as the check.
• digital-seizing/rapid-prototyping — leaders as “continuous experimentation officers”; grassroots citizen-developer builds scaled when they work; experiments framed by outcome goals.
• contextual/internal-barriers — the barriers named are fear of job loss, professional identity (coder versus builder), tacit knowledge that cannot be seen the way a production line can, and a first rollout run as a product launch.

How it connects

• enterprise-ai-adoption — product launch versus business transformation, team-level programmes, top-down goals with bottom-up builds, and the shift from many pilots to three end-to-end processes. The same show’s Sadun episode gives the research side of the same Frontier Firm programme, and the HBS AI Institute’s Frontier Firm page warns: “Do not assume that employees would self-discover uses without specific guidance.”
• micro-productivity-trap — the 30% automation model rejected, and adoption-to-behaviour-to-outcome as the measurement chain.
• automation-vs-augmentation — “Capability Add”: the process gains a capability it never had, rather than doing the old work faster.
• expert-generalist — T-shaped roles, with expertise and judgment worth more, not less.
• Raman & Wood (WorkLab, June 2026) — the same show a month earlier, on skills rather than titles; Raman leads LinkedIn’s economic opportunity work, also inside Microsoft.
• Dutt et al. — the same move from experimentation to transformation, from Bain’s client work.

What was actually ingested

The full episode (28:18, length_seconds: 1698) from the creator-uploaded caption track. The transcript panel did not render, so the captions were fetched with yt-dlp and merged into sentence-length segments. Where the captions write “I” for AI and “Katie” for Katy, quotes here use the intended words. No stills: a podcast recording.

Linked entities and concepts

• Microsoft — publisher (WorkLab) and George’s employer; subject of most of the examples. Updated in this ingest.
• HBS AI Institute — runs the Frontier Firm AI Initiative with Microsoft, which George cites. Updated in this ingest.
• enterprise-ai-adoption, micro-productivity-trap, automation-vs-augmentation, expert-generalist — updated in this ingest.
• dynamic-capabilities — the four cells tagged above.

Dangling (single-source mention, deferred): Katy George; Molly Wood (host of several WorkLab episodes, but author: is the channel); Harvard Business School.

Scope and reliability

An executive interview on the employer’s own channel. Every example is Microsoft’s own practice, and the reported effects are George’s account: no numbers are given for Camp Air’s or the huddles’ effect on adoption, and the 100 internal case studies are not published. The link between AI use and revenue per rep is a correlation; reps who sell well may also be the ones who adopt tools first, and the episode does not separate the two. Microsoft sells Copilot, so a better adoption story is also a sales argument. The value of the episode is the description of mechanisms (team-level programmes, role huddles, the measurement chain) and the “Capability Add” framing, which several sources on micro-productivity-trap reach by other routes.