I designed Microsoft's $5B EA channel architecture in 2001. The 2026 transition is missing what made it work

The Register's reporting on the Microsoft Enterprise Agreement (EA) commission collapse has been the most data-rigorous coverage in the trade press. And the trajectory it documented ($2.5 billion in LSP commissions in 2023, $1.67 billion in 2024, $583 million in 2025, zero in 2026) is the financial signature of a structural transition I have seen before, because I designed the original architecture that is now being retired. Between 1998 and 2001, I was the sole designer of the Enterprise Software Advisor (ESA) channel architecture at Microsoft. Working within Worldwide Licensing and Pricing, I built the direct-billing model that converted the EA channel from an indirect, margin-based reseller structure to a direct-billing, advisory-fee structure. The compensation model, fee schedule, three-tier segmentation covering 75,000 accounts across 24 countries, and the geographic rollout sequence were my design. The ESA designation remains named verbatim in Microsoft's FY2025 10-K, twenty-four years after it launched. I'm writing this because the comparison between 2001 and 2026 reveals something the coverage has not yet examined: the 2001 transition included a specific mechanism designed to preserve channel expertise and align partner incentives with the new model. The 2026 transition does not. That difference matters, and it has now drawn regulatory attention. What the channel Looked Like in 1998 When I joined Microsoft in June 1998, the enterprise licensing channel was collapsing from three directions at once. Large Account Reseller (LAR) margins had fallen from a sustainable 4% to 2.2%, with some partners operating in the red. Dell had developed what it internally called its "bulldozer" strategy: winning Enterprise Agreement deals at zero to negative 2% margins, using the EA as a loss leader to gain a foothold for auditing and displacing competitors' hardware across the customer's entire estate. No LAR could compete with a player pricing below cost. At the same time, Microsoft's own field sales teams had begun negotiating EA deals directly and handing them to the channel for execution. The advisory work that had earned partner margin had moved upstream. And the licensing had grown so complex that enterprise customers described the pre-EA compliance experience as "Kafkaesque," a characterization Barb Darrow's reporting in CRN helped establish in the trade press. Software Asset Management was cumbersome, customers blamed Microsoft's licensing complexity, and nobody wanted to pay for the expertise required to manage it. The channel that helped build Microsoft's enterprise business was becoming structurally unviable. My boss, Bill Henningsgaard, gave me the assignment directly: "The channel is going out of business. Give me some ideas." The architecture: How the ESA model actually worked Rather than studying other software companies' channel models, I looked at industries that had solved analogous problems: insurance and automotive. Both had built models where an intermediary earned fees for defined advisory activities in addition to or in place of margin on transactions, and competed on service quality rather than price. That cross-industry research became the conceptual foundation of the ESA model. The core design decisions were to convert billing from indirect to direct (Microsoft invoices the customer, not the partner), redefine partners as advisors rather than resellers, and compensate them through activity-based fees rather than margin on transactions. The economics were self-financing. Microsoft was already spending 17.7% of EA contract value as a volume discount flowing to LARs. The new advisory fee model redirected a portion of that existing expenditure from an undifferentiated discount to a performance-based fee. Partners who invested in licensing expertise, compliance management, and deployment support earned more. Partners seeking compensation without delivering meaningful advisory services earned less. The reallocation improved partner quality while remaining cost-neutral to Microsoft. When I presented this to roughly forty people including Bill Gates and Steve Ballmer in a mid-year review, Ballmer's reaction was immediate. He called it "a perpetual motion machine." The skepticism was fair. The model survived a second presentation with granular financial detail and made it through the most demanding executive review process in the company. The EA channel was not one segment. I defined three customer tiers with distinct economics: Microsoft-Led (1,150 global strategic accounts, $5B opportunity, 4% ESA fee), Channel-Assisted (14,000 corporate accounts, $3B, 9% fee), and Channel-Led (60,000 medium enterprise accounts, $3.5B, 15% fee). Each tier defined different levels of Microsoft engagement and partner autonomy depending on account complexity, and the tiered fees were structured across pre-sales, base, and true-up activities to align incentives with the work that drove our desired customer outcomes. The results Microsoft's unearned revenue, which represented committed multi-year licensing and is equivalent to what the industry now calls ARR, increased from $1.92 billion in June 2001 to $7.74 billion in June 2002: a $5.82 billion surge in twelve months. Microsoft's own 10-K filings attributed revenue growth across fiscal years 2002, 2003, and 2004 to the pre-transition enrollment period. The FY2003 filing stated that revenue growth in fiscal 2003 was driven primarily by multi-year licensing that occurred before the Licensing 6.0 transition date. Licensing 6.0 launched on October 1, 2001. Within eighteen months, 2,577 Enterprise Agreements were executed across the United States, Canada, and 22 Western European countries. The channel was not just preserved. It became more productive, because the new economics rewarded expertise rather than transaction volume. What is structurally different in 2026 Both transitions share the same logic. In each case, Microsoft is internalizing margin, moving to direct billing, and redefining how channel partners are compensated. The Register's reporting has documented the financial trajectory precisely. What has not yet been examined is why the 2001 transition preserved channel productivity while the 2026 version risks destroying it. In 2001, the ESA model was designed to function as a bridge. It gave partners a defined advisory role, an economic model tied to activities rather than transaction volume, and a clear reason to remain invested in the Microsoft enterprise business. The best-performing partners earned more under the advisory fees than they had earned under the margin discount. The channel was restructured. It was not abandoned. The 2026 transition has no equivalent mechanism. The commission pool has been eliminated without a replacement economic model for partners who built their businesses around enterprise licensing. The guidance from Microsoft is to pivot toward managed services and cloud consumption. That is reasonable long-term counsel. It is not a structural answer to the immediate question of how to replace billions in annual commission revenue that is disappearing within a 36-month window. The consequences are visible now. On May 14, the UK Competition and Markets Authority (CMA) opened its formal Strategic Market Status (SMS) investigation into Microsoft business software, with the company's licensing practices explicitly in scope. If the CMA designates Microsoft under the SMS regime, it can impose conduct requirements on how the company prices, bundles, and licenses its enterprise stack. Public comment closes June 4. In 1998, the US Department of Justice sued Microsoft for using Windows to box out rivals. I was a few months into my ESA work when the trial opened. The bundling-and-tying questions before the CMA today are the same questions, only aimed at the AI layer rather than the browser. May also produced a natural experiment among UK-listed Microsoft resellers. Bytes Technology Group reported on May 12 that operating profit had fallen, with its CEO attributing the decline to "well-published Microsoft incentive changes." Ten days later, Softcat raised FY26 guidance to mid-teens operating profit growth. Similar customer bases, opposite outcomes, ten days apart. That is the channel economics this piece describes, in real numbers. The specific risk now is that the enterprise licensing expertise currently embedded in the partner channel will dissipate before Microsoft's own direct sales organization can fully absorb it. Enterprise customers do not engage with Microsoft's licensing structure once every three years at renewal. They engage with it continuously, through compliance management, true-up calculations, product deployment decisions, and budget reviews. That ongoing advisory function has historically been performed by partners. It does not disappear because the billing model changes. Someone still has to perform it. The question before Microsoft is not whether it can operate a direct billing model. It demonstrated that capability in 2001. The question is whether it can absorb the advisory function the channel has performed for more than two decades without providing a transition mechanism for the expertise that function requires. In 2001, we built the bridge before we asked anyone to cross it. In 2026, the bridge is being removed while partners and customers are still relying on it.(R) Brendan T O'Connor is Founder & Principal of The Brushton Group, a strategy consulting practice delivering enterprise-grade commercial strategy for growth and scale. During his eighteen-year Microsoft tenure (1998 to 2016), he designed the Enterprise Agreement direct-billing architecture, including the indirect-to-direct conversion, the ESA compensation model, the fee schedule, and the geographic rollout, and co-developed the broader Licensing 6.0 program.