A Vendor Who Wants to Lose Money on You
Salesforce's Chief Revenue Officer said something recently that every enterprise buyer should read twice. When asked about the company's new agentic AI pricing model, he said: "I would love to have a customer where I price an AELA at $5 million incremental, and the customer has deployed so much that the deal is not profitable for me. If the deal isn't profitable for me, it means that the customer is the happiest customer in the world. And then I have another 20 years to monetize that customer."
Read that again. Salesforce's top revenue executive is telling you, on the record, that they are willing to lose money on your contract today because they plan to make it back over the next two decades.
That is not generosity. That is a business strategy. And it is spreading across every major AI vendor in the market right now.
A new category of enterprise AI contract called the Agentic Enterprise License Agreement, or AELA, has emerged in 2026. Constellation Research predicts it will become the industry standard for AI procurement this year. If your organization is evaluating or has already signed one, understanding what is actually happening inside that contract is one of the most important things your leadership team can do right now.
What Changed and Why It Matters
For the past twenty years, enterprise software pricing was simple. More employees meant more seats. More seats meant more cost. Companies like Salesforce, Microsoft, and Workday built billion-dollar businesses on that model.
Agentic AI breaks it. An AI agent does not need a login. It can do the work of ten employees without ten licenses. The better AI agents get at automating tasks, the fewer human seats any vendor can charge for. Every major software company woke up to the same problem: their entire pricing model was designed for a world where humans are the users. What happens when the AI is the user?
The AELA is their answer. Instead of charging per human seat, vendors now offer a flat annual fee for unlimited access to their AI tools. It sounds like a better deal. In many ways, it is a better deal today. The trap is what happens at renewal.
The Trap Hidden in the Fine Print
Here is how the renewal works, and Gartner's director analyst for IT sourcing and procurement has confirmed this publicly.
When you sign an AELA, you get unlimited AI usage at a flat fee. The vendor tracks exactly how much you use throughout the contract period. When your contract comes up for renewal, they look at all that usage data and convert your "unlimited" agreement into a defined quantity contract. They set a ceiling based on how much you used, and they price the renewal above it.
Gartner warns that Salesforce is already proposing price increases of 6 to 15 percent above inflation at renewal. That uplift is calculated against a usage baseline you built while thinking the pricing was flat. The more you used the platform, the higher your renewal number will be.
One analyst described it plainly: unlimited agent licensing is basically the enterprise playbook. Eat margin upfront. Become the default workflow layer. Then monetize the dependency at renewal.
The reason this works is straightforward. Once your team has built AI workflows, automated processes, and daily operations around a specific vendor's platform, switching becomes genuinely disruptive. The vendor knows this. The pricing at renewal reflects it.
The Hidden Cost Nobody Is Budgeting For
The flat license fee is only part of what you are agreeing to. Most organizations signing AELAs are not modeling the data connection fees that come with them, and this is where the real cost surprise lives.
As AI agents connect across your systems, moving data between your CRM, your ERP, your data warehouse, and the AI layer, each connection generates a fee. These data tolls are separate from the license. Constellation Research has flagged them as the biggest emerging cost risk in agentic AI deployment, describing them as the new cloud egress fees. In some cases, data connection fees can exceed the platform license itself.
Most organizations do not know enough about their own data architecture to model these costs at the time of signing. If you do not know how many systems your agents will connect to, or how much data will flow between them, you cannot accurately estimate your total cost of ownership. You sign based on the license number, and the real cost reveals itself over the course of the contract.
This is not a niche problem. It is a structural feature of how agentic AI contracts are being written, and it affects every organization that is moving from pilot-stage AI into production-scale deployment.
Why This Is Specifically a Leadership Problem
The reason most organizations are walking into this without a clear framework is that AI procurement has not yet caught up to AI deployment.
Procurement teams are still using the same negotiation playbook they used for traditional SaaS contracts: focus on the per-seat price, negotiate volume discounts, and review at renewal time. That approach made sense when software was simple and pricing was transparent. It does not work for agentic licensing, where the headline fee is predictable but the total cost of ownership depends on data architecture decisions, usage patterns, and renewal terms that most organizations are not tracking during the contract period.
The vendor's negotiating team has spent months designing these contracts. Most enterprise procurement functions have not yet built the equivalent preparation on the buyer side. That asymmetry is why Gartner's recommendation is specific: before signing any AELA, negotiate explicit limits on renewal price increases, require audit rights so you can monitor your own usage throughout the agreement, and demand clarity on how the unlimited terms convert at renewal. These protections are negotiable at signing. They become significantly harder to negotiate once your operations depend on the platform.
What Prepared Organizations Are Doing
The organizations approaching AELAs from a position of strength are doing three things differently.
First, they understand their own data architecture before they sign. Knowing which systems your agents will connect to, how much data will move between them, and what each connection will cost is not a technical exercise. It is a financial modeling exercise that belongs in the hands of the CFO's office, not just the IT team.
Second, they model usage realistically rather than optimistically. Vendors build their AELA case on maximum deployment scenarios. An organization that signs based on best-case usage assumptions and then takes 18 months to actually deploy is paying for value it is not receiving, with no recourse. A realistic deployment timeline, grounded in honest assessment of your current data readiness and organizational capacity, produces a much more defensible contract.
Third, they treat the AELA negotiation as a multi-year financial commitment, because that is what it is. The vendor is explicitly planning to monetize your organization for 20 years. The initial contract terms should reflect the seriousness of that commitment.
Where KAIDATA Comes In
This is exactly the kind of problem that looks like a technology decision but is actually a strategy and infrastructure decision. And it is precisely where most organizations need independent perspective.
The vendors signing AELAs are not adversaries. They are partners who have made a business calculation that their long-term interests are best served by embedding deeply in your operations. That calculation can absolutely work in your favor if your organization enters the relationship with clear requirements, realistic deployment plans, and negotiated protections.
What makes that possible is having your own house in order first. Knowing your data architecture. Understanding your current AI readiness. Having a clear picture of which workflows you are actually ready to automate and which require foundational work before AI can operate in them reliably. Organizations that sign AELAs before answering those questions are handing the vendor significant structural advantage in the relationship.
KAIDATA helps organizations build that foundation before the contract is signed rather than discovering its absence after. Understanding where your AI readiness actually stands, what your data infrastructure can support, and what your organization needs from an AI vendor relationship before committing to one is the work that determines whether an AELA becomes a genuine competitive asset or an expensive dependency you are locked into before you understood what you were agreeing to.
The vendor is planning for the next 20 years of your relationship. Your preparation for that relationship should match.
