When the Seller Becomes the Advisor

Between May and July of 2026, three things happened in the technology industry that, on their own, read as routine business news. Put side by side, I think they describe something more structural. At every layer of the stack, the party that builds and sells the technology is also becoming the party paid to advise on how to use it.

If you are making a decision about your AI, ERP, or CRM roadmap this year, that shift matters more than any single feature release.

Microsoft, the seller and the advisor

Microsoft's FY27 partner incentives took effect on July 1, 2026. The flat rebate that indirect resellers earned simply for holding Microsoft 365 and Dynamics 365 licenses went to zero. In its place is a tiered growth accelerator, reported at 7.5% rising to 12.5%, that pays specifically for moving customers into premium, AI-enabled tiers: Copilot, Microsoft 365 E5 and E7, and Agent 365.

The stated intent is reasonable. Reward partners who help customers grow into higher-value workloads, not just renew what they already have. Many partners already do that work well. What I would ask, from the buyer's side, is a narrower question. As the incentive to grow your spend gets stronger, is there value in a second, unincentivized read on whether the recommended tier actually fits your business? Alongside the partner relationship you already have, not instead of it.

Days after the incentive changes landed, Microsoft launched Frontier Company: a $2.5 billion commitment and about 6,000 of its own engineers, technical consultants, and industry specialists, embedded inside enterprise customers. The company that sells the product now also staffs some of the teams that help customers decide how much of it to buy.

Anthropic and OpenAI, hours apart

In May, Anthropic partnered with Blackstone, Goldman Sachs, and Hellman & Friedman to launch an AI services company, backed by $1.5 billion in committed capital. It is built to embed engineers inside mid-size businesses, in direct competition with traditional consultancies.

OpenAI announced a near-identical structure the same day: The Deployment Company, backed by TPG, Bain Capital, Brookfield, and other investors at a $10 billion valuation.

Neither is hiding the intent. The pitch in both cases is that enterprises need help turning frontier models into working systems, and the labs are now positioned to sell that help directly, next to the models themselves. Weeks later, AWS committed $1 billion of its own capital to a comparable forward deployed engineering unit. It is internal rather than a joint venture, but it is the same move.

The pattern underneath

None of this makes any of these moves wrong, or even surprising. Building distribution around a valuable product is an obvious strategy, and the technology may be genuinely good for the customers who adopt it. That is not my point. My point is where the advice is coming from.

Charlie Munger put the mechanism more plainly than any announcement above:

"Show me the incentive and I will show you the outcome."

I read his rule as reaching past business in general to the person advising you. An advisor's incentive shapes the outcome they steer you toward, whether that advisor is a reseller earning a growth accelerator, a vendor's own delivery arm, or a lab's forward deployed engineering team.

The incentive problem existed before 2026. What this year did was make the case for an independent read sharper. As every layer of the stack becomes more financially motivated to tell you what to buy next, the value of a read with nothing riding on the answer goes up, not down.

What I would do with this

If your Microsoft, ERP, or AI roadmap conversation is happening entirely inside one channel, I would not treat that as a problem to solve by itself. I would pair it with a second, independent read on whether the recommended tier, platform, or deployment model fits the business. Good partners welcome that. It is the same reason a second medical opinion is standard even when nobody doubts the first doctor.

It is inexpensive insurance against an expensive mistake. Just as often, it confirms the original recommendation was right.

The question I would put to any recommendation is not "is this a good product." It is "is this the right recommendation from someone who is paid nothing for making it."