WorkedLisa Su, AMD, 2014 onwards

For most of the decade before Lisa Su became CEO of AMD in October 2014, the company's quarterly plans for consumer GPU competition with NVIDIA had been, in sequence, the same plan slightly modified. NVIDIA held 80–90% of the discrete GPU market.

The CUDA software ecosystem had made the moat structural, not just technical — developers wrote their software against CUDA because their customers ran CUDA, and customers ran CUDA because developers wrote against it. Each year AMD's plan was to close the performance gap in the next generation of consumer silicon. Each year the market-share number barely moved.

Su's response, after taking the CEO seat, was not a better consumer GPU plan. It was a multi-year reallocation of the firm's resource commitment toward a different category of customer entirely: enterprise and AI infrastructure, where NVIDIA's software moat was weaker, the customers were fewer and larger, and AMD's underlying silicon-design talent could compete on architecture rather than on ecosystem.

The EPYC server CPUs, the Instinct accelerators, and the open ROCm software stack were all consequences of that decision. The trade-off was visible: the consumer Radeon roadmap became notably sparse, and AMD took reputational hits from gaming-focused tech press for years before the enterprise bet paid off.

When last quarter's plan didn't land, the default reaction is to write a better version of the same plan. Su's reset required accepting that the previous quarter's plan, and several before it, had not been marginal failures — they had been the wrong plan, and the plan-revision process was reliably producing a slightly different wrong plan.

By 2025 the enterprise pivot had brought AMD record revenues and a 61% year-on-year net income increase. The lesson isn't that consumer GPUs were a mistake. It's that writing a better plan against a customer you cannot actually win is writing a better version of the problem.

Sources: AMD investor materials and earnings calls, 2014–2026. CES 2026 keynote (rev.com/transcripts/amd-at-ces-2026). AMD newsroom press release, 5 January 2026. ResearchGate, "Lisa Su, AMD, and the Illusion of Consumer Alignment in the AI Era" (2026). Lisa Su profile coverage in Fortune, Forbes, and IEEE Spectrum, 2018–2025.

WorkedIndra Nooyi, PepsiCo and Performance with Purpose, 2006 onwards

When Indra Nooyi became CEO of PepsiCo in October 2006, the prior year's plan had landed — by the numbers. Revenue was up, margins were holding, the quarterly narrative was defensible. What it had not done was resolve the structural question that had been growing underneath the portfolio for the decade prior: PepsiCo's core categories, carbonated soft drinks and salty snacks, were on trajectories that well-run quarterly execution would not fix.

Obesity rates were rising. Regulatory pressure was gathering. The analyst questions in the earnings calls were politely asking whether PepsiCo was, structurally, the wrong shape for the ten-year environment the firm was about to run into.

Nooyi's response, which she formalised as "Performance with Purpose" and set out at length in her 2021 memoir My Life in Full, was not a better quarterly plan. It was an acknowledgement that the quarterly plan was the wrong unit of answer to the question being asked.

The new framing committed PepsiCo to three simultaneous long-run agendas — shifting the portfolio toward healthier products, reducing the environmental footprint of operations, investing in the firm's people — none of which would show as legible gains in any single quarter's numbers, and some of which would show, visibly, as near-term cost.

When you're writing next quarter's plan and last quarter's didn't land, the honest question is sometimes whether the quarter is still the right unit to be planning in. Nooyi's move wasn't to write a better three-month plan.

It was to notice that the three-month plan was the form the wrong question was taking, and to commit the firm to a plan that the three-month reporting cadence could not, by design, fully measure. The quarterly plan still existed. It stopped being the top of the stack.

Sources: Indra Nooyi, My Life in Full: Work, Family, and Our Future (Portfolio, 2021). PepsiCo annual reports and proxy statements, 2006–2018, including "Performance with Purpose" framing across the period. Academic treatments in Strategic Management Journal and Journal of Business Strategy, 2010–2019. Indra Nooyi, Harvard Business Review interviews and articles on strategic leadership, 2007–2021.

SourcesIndra Nooyi, My Life in Full BuyAmazonBookshop

WorkedSherina Ebrahim, McKinsey and the "agile organization" framework, May 2019

In May 2019, Sherina Ebrahim co-authored a McKinsey Quarterly article with Daphne Brosseau, Christopher Handscomb, and Shail Thaker titled "The journey to an agile organization" — one of the canonical firm-level pieces of writing on why traditional annual planning cycles were losing their grip on digital-era enterprises, and on what a serviceable alternative looked like.

Ebrahim's role in the authorship was not as the named executive sponsor of a pattern documented by her team; she was one of the four bylined authors. The distinction matters because the paper's thesis rests on a specific structural claim the co-authors had been observing across client engagements for several years before the article appeared.

The claim was this. The reason last quarter's plan kept not landing was not that planning had become harder, or that the firms McKinsey was advising had become worse at execution. It was that command-and-control planning had a prerequisite — a reasonably stable environment between plan and execution — that had quietly stopped holding for digital businesses.

The paper codified the alternative as a "network of teams" operating model, where the unit of planning shifts from the annual divisional plan to the rapidly-reconfiguring team, and where the firm's architecture is designed to recompose around the work rather than to require the work to fit the architecture.

When last quarter's plan didn't land, there is a version of the diagnosis where you need a better plan, and there is another, more structural version where you need a different planning cadence altogether. Ebrahim and her co-authors were writing about the second.

The article is useful not because it is prescriptive — it isn't — but because it names the shape of the problem cleanly: if the environment is changing faster than your planning cycle, the planning cycle is the thing you can change.

*Sources: Daphne Brosseau, Sherina Ebrahim, Christopher Handscomb, and Shail Thaker, "The journey to an agile organization," McKinsey Quarterly, 10 May 2019 (mckinsey.com/business-functions/organization/our-insights/the-journey-to-an-agile-organization).

Handscomb, Ebrahim, and colleagues, "How to build an agile organization that will thrive long-term," McKinsey, September 2021. McKinsey Organization Practice working papers on agile operating models, 2018–2024.*

WorkedGameLayers, 2007–2010

The team built an ambitious passive multiplayer game that overlaid the entire web — a genuinely original idea that innovated on gameplay, theme, and platform simultaneously. The problem was that innovating on three axes at once leaves you nothing to anchor against when any of them fails.

They burned sixty thousand dollars a month without finding traction and pivoted to Facebook games as a survival move. That didn't work either. The founders used their remaining capital to shut down gracefully — laid off the team properly, paid what they owed, closed the studio with their relationships intact.

SourcesGameLayers

WorkedMarc Hedlund, Wesabe, 2010

Wesabe was an early personal finance startup that lost to Mint and eventually shut down. Hedlund wrote the post-mortem himself, naming two specific strategic errors: refusing to use an automated data aggregator (which meant users had to do manual work Mint's users didn't), and prioritising tools designed to change financial behaviour over tools designed to give users an immediate useful thing.

Mint did the opposite of both and won. The post-mortem is honest in a way that's rare — Hedlund doesn't blame the market, the timing, or the investors. He names the calls he got wrong.

SourcesWesabe post-mortem

Shouldn’t have been triedGroupSpaces, early 2010s

GroupSpaces had a real success on its hands. The product — a tool for managing the membership and activities of small groups — had found genuine product-market fit serving university student groups in the UK and US. Adoption was strong. Engagement was high. The metrics were honest, not vanity.

The founders, encouraged by the unmistakable signal, raised venture capital and put together a plan for the following quarter that would expand the product into much larger markets: enterprise teams, professional associations, sports leagues. The plan didn't land. The features that worked for student societies didn't work for enterprise customers, who wanted things like permissions hierarchies, audit trails, and integration with existing identity systems.

Each new market required a different version of the product to work, and the team didn't have the bandwidth to build any of them while also maintaining the original. In the post-mortem, the founder named the specific error. The next quarter's plan had been an extrapolation, not a continuation.

They'd assumed that proving the product worked in one market meant they were entitled to the next. The plan had been built on confidence rather than evidence, and the confidence had been earned in a market that had nothing in common with the markets they were now trying to enter.

SourcesGroupSpaces post-mortem