The most consequential AI purchase most enterprises made this year was one nobody set out to make. It arrived on a renewal they were going to sign anyway. No use case, no vendor evaluation, no negotiation. Just a line on a contract that looked like last year’s.
Your existing software vendors, the ones already running your productivity suite, your CRM, your service desk, are embedding AI into products you already own. It shows up as a new module in a renewal, a price increase justified by capabilities you did not ask for, or a feature switched on by default after a product announcement. No new vendor. No purchase order. No procurement review. The exposure lands in your renewal terms and your data-processing agreement, and most organizations find it after they have already signed.
The mistake this piece is about: treating an AI-laden renewal as the routine renewal it looks like. The AI did not come through the front door where procurement is watching. It came through the walls.
The AI Spend Lifecycle. This piece is part of an eight-part framework for bringing AI under the same procurement discipline as every other major category: buy it right, manage it as a category, fund it against the plan. The eight: (1) the four types of AI you buy, (2) the embedded AI in your renewals, (3) negotiating the token deal, (4) what belongs in every AI contract, (5) AI as your largest unaddressed spend category, (6) matching models to tasks, (7) why AI savings never reach EBITDA, and (8) governing AI spend against the operating plan. You are reading Part 2.
The year embedded AI stopped being optional
For a while, embedded AI was a feature you could take or leave. That changed in 2026, and the change is worth seeing in detail because it is the clearest signal of where every major vendor is heading.
On July 1, 2026, Microsoft reset its Microsoft 365 commercial pricing worldwide, raising list prices across its suites by roughly 5 to 43 percent depending on tier. A mid-tier enterprise license rose about 13 percent, a frontline tier about 33 percent. At the same time, it folded Copilot Chat directly into most tiers for the first time, so generative AI stopped being a separate purchase decision and became part of the base package. The opt-out did not get harder to find. It disappeared.
The sharpest detail sits one layer down. In the top enterprise tier, Microsoft now allocates a fixed block of Security Copilot compute to every tenant, whether that tenant uses it or not. You are provisioned for AI you never requested, and it is priced into the license you renew. Salesforce made a parallel move, pushing cumulative list-price increases past 20 percent over three years, with Agentforce and consumption-based AI billing layered on top of the base platform as a third axis of cost.
The pattern, stated plainly: baseline AI has quietly become a utility cost baked into the software you already run, rather than a discrete decision you get to make. The vendors are not hiding this. They are counting on the renewal to look routine enough that nobody negotiates it.
Three doors, none of them procurement's front door
Embedded AI enters through three paths, and none of them trips the control that catches a normal purchase.
-
The terms-of-service update. A revised agreement grants the vendor expanded rights over your data, sometimes including the right to use your content to train or improve their models. It arrives as a click-through, and it is accepted by someone who is not procurement and not legal.
-
The bundled renewal. AI capability is folded into a higher tier or a price increase, so the renewal you were always going to sign now carries an AI charge you never evaluated. Buyers describe a double uplift: the AI line forces you up a license tier, and then the AI charge lands on top of the larger base.
-
The default-on feature. A capability is announced and switched on across your environment, no approval required. It is live before anyone has decided whether the organization wants it, let alone whether the data handling meets your requirements.
In all three, the question procurement should be asking is the same. What rights did we just grant this vendor over our data, and did we agree to that on purpose?
From the operator's chair. An enterprise SaaS vendor we already ran offered a new SKU, an AI analytics layer that sat on top of the reporting structure we owned. Clearly capable, clearly expensive, and with no ROI case attached. We bought it. Not because anyone was fooled, but because a P&L owner wanted to be AI-forward, which is the right instinct. The failure was not the decision to invest. It was that we bought the uplift without building the outcome layer to justify it. Nobody forced the questions that should have come first: how much CSAT could this actually move, which use cases justified the lift, and how would we measure it. We never set the baseline, so to this day the honest answer to whether it worked is that we cannot say. That is the trap. The instinct to move on AI is correct, and instinct is not an ROI case. Someone has to make the good impulse answerable to an outcome, especially when the impulse is right.
The return-on-investment question changes shape
When you go out and buy new AI, the return question is straightforward: what will this replace, reduce, or improve, and can finance validate the number. Embedded AI breaks that clean setup, because you are not evaluating a fresh investment. You are evaluating whether the incremental value of an AI capability justifies the incremental cost being added to your renewal, and whether the data rights you grant in exchange are acceptable.
Three questions decide it. First, what am I being asked to pay more for, and what specific AI capability is bundled into that increase? If the vendor cannot separate the value of the AI feature from the base product, the increase is not justified. Second, is the AI actually replacing a workflow, or just assisting one that still takes the same headcount and effort? Only replacement produces a return finance will validate. Third, what data rights am I granting, and what are they worth? Proprietary spend data, contract terms, and negotiation history have real commercial value, and handing a vendor the right to train on them is a transfer of value that appears on no invoice.
The framing that keeps you honest: a feature that assists a human is not the same as a feature that eliminates a process. The vendor will describe both as transformation. Only one of them shows up in the P&L.
What to check before you sign the renewal
Embedded AI turns your renewal into an AI negotiation whether you treat it as one or not. Three provisions deserve a hard look every time, and none of them is exotic. They are the same rights-and-accountability questions procurement already knows how to ask, pointed at a new place.
-
Model training rights. The terms must explicitly prohibit the vendor from using your data to train, fine-tune, or improve their models without your prior written approval. If that language is not in your data-processing agreement, add it at renewal.
-
Data residency. Confirm the AI processing does not route your data outside your required geography. Embedded AI features sometimes run on infrastructure separate from the base product, with different residency behavior than the product you originally cleared.
-
Opt-out rights. You should be able to disable AI processing for specific data types or workflows without losing the base product or paying a penalty. When the feature is default-on, this right is the only thing standing between a policy decision and a fait accompli.
And treat the double uplift as two negotiations, not one. Separate the base-product renewal from the AI charge, make the vendor quantify the AI value on its own, and decide on each independently.
Why this is the door that matters most
Of the four types of AI an enterprise buys, embedded AI is the only one that enters without a purchase decision. That is what makes it the most likely to slip past you. A token deal, an AI application, an agentic system, each at least triggers a review. Embedded AI arrives inside a renewal that looks like last year's renewal, and the exposure is not something you chose. It is something that got switched on while you were looking somewhere else.
There is a broader dynamic underneath, and one analyst put it more cleanly than I could. When a vendor's commercial strategy moves faster than the customer's governance maturity, the customer loses leverage. That is the whole embedded-AI problem in a sentence. The vendors have decided AI is now part of the base product. If your renewal process has not caught up, every renewal quietly transfers a little more leverage to the other side of the table.
The question for your next renewal: what AI got added to this contract since last year, what data rights came with it, and did anyone actually decide to accept them? If you cannot answer, the renewal is making the decision for you.
The discipline is not new. It is the same rights, residency, and accountability review procurement runs everywhere else, applied to the one purchase that tries hardest to slip past it. Read what changed in the terms, not just what changed in the price. The AI is in there either way.