Enterprises will keep asking a familiar question in a new form: How much is our AI costing us? But as vendors push AI deeper into enterprise software, another question is becoming harder to answer: What exactly are buyers paying for now?
This is no longer just a budgeting issue. It is becoming a software management issue.
Vendors are embedding AI into products, bundling it into suites and building new coordination layers to make AI agents work across applications, workflows and human handoffs. The result is that AI is getting harder to price, isolate and manage once it is woven into the software itself.
Why AI is getting harder to price
Sitecore offers one version of that shift. CEO Eric Stine said the company "got rid of the seat-based model" and now prices based on content interactions rather than seats alone. He also said, "Our SaaS product is an AI product," and added, "Software is automation." That is a much more expansive view of what customers are buying.
From the vendor side, the logic is easy to follow. If software companies understand the domain, the workflows and the process logic, they will argue that they are the natural place to embed AI. Stine makes exactly that case, saying the companies that survive will be the ones with the domain expertise to embed AI into workflows at scale. He also argues that enterprises are not going to want to build edge-case AI processes on their own for functions like account-based marketing, employee relations, supply chain management or finance.
That might be true. But it also changes the buyer's problem.
When AI is embedded this tightly, it starts to feel less optional and less separable from the product itself. Cost clarity declines because buyers can no longer easily tell which parts of the bill are tied to AI, especially as pricing shifts toward interactions, retrieval and usage. Functional clarity drops, too. It becomes harder to separate what AI is actually doing from the product's broader automation claims.
Buyers are no longer just choosing software for features. They are increasingly buying into a vendor's theory of how work in that domain should be automated.
3 ways vendors are turning AI into software spend
Sitecore shows one path: AI woven so tightly into the product that it starts to blur into the software itself. Salesforce shows another, packaging agentic AI inside familiar suite subscriptions as a visible extra. Then there is orchestration -- the growing sense that companies may need some added layer of control to keep agents across systems from creating more confusion than value. Taken together, those approaches suggest that enterprises are not confronting a single clear AI pricing model but several overlapping ones.
Bundled AI still blurs the bill
Salesforce is taking a different route. In its SMB suites, the company added prebuilt Agentforce AI agents at no extra cost. Those agents can handle work such as updating CRM records, visualizing lead activity, drafting customer emails and generating customer record summaries. They are included across the Free, Starter and Pro subscription levels.
That model gives buyers a little more visibility than one in which AI is so tightly fused to the product that it nearly disappears. Agentforce is still presented as a distinct capability, not something hidden entirely inside the software. Even so, it pushes AI closer to becoming part of the standard software bill. Over time, that changes the expectation. Instead of treating agentic AI as a separate buying decision, customers might start to see it as part of enterprise software.
That also helps explain why vendors are keeping the first use cases relatively narrow instead of jumping right to full autonomy. Salesforce is using agentic AI for routine tasks first, the lower-risk work that can make the technology feel useful before customers are asked to trust it with more sensitive decisions. That is not only a product strategy. It is also a way to ease customers into the technology without creating as much risk upfront.
Narrow use cases are staging trust
Vendors are not just narrowing AI use cases. They are staging trust.
Building trust is essential for any technology to become commonplace in the enterprise. That is especially true with automation, and even more so when it takes the form of agentic AI, which can appear to think and act on its own.
Narrower use cases protect both enterprises and the reputations of the vendors supplying the technology. They create space to demonstrate value, absorb failure and make AI feel manageable before customers are asked to rely on it in more sensitive work.
What buyers lose when AI disappears into the product
As AI gets folded more deeply into software, buyers can lose sight of more than one thing at once. The cost becomes harder to pin down when pricing is tied to usage, interactions or bundled capabilities rather than a more familiar line item. The function can get blurrier, too. Once AI is woven into broader workflows and automation claims, it becomes harder to tell what the technology itself is actually doing. That is where a pricing issue becomes a software management issue.
Orchestration adds a new cost layer
But once AI becomes more common across enterprise software, another problem arises: how it is supposed to work together. This is where the pricing question starts to widen into a management question. Vendors can ship agents. The harder challenge is orchestrating how they work across applications, workflows and human handoffs without creating more risk.
Vendors are not just narrowing AI use cases. They are staging trust.
Orchestration keeps coming up because most enterprise environments are already complicated. Companies are dealing with multiple platforms, overlapping systems and workflow layers that were never designed to line up cleanly. At the same time, vendors are starting to treat agents as a normal part of the software package. That leaves buyers with a coordination problem they will still have to solve, no matter how smooth the demo looks.
The responses are starting to take different shapes. Salesforce is stressing context, control, observability and orchestration. Oracle is connecting agents to business goals and a staged move toward greater autonomy. Workday is centering the discussion on trusted HR and finance data, permissions and process logic. Zoom is pushing the idea deeper into the workflow layer, where conversations and meetings can become the inputs for completed work.
In practice, that means buyers could be taking on more than the cost of the agents themselves. They might also be taking on the cost of keeping those agents coordinated across systems, teams and workflows. Whether that shows up as added tooling, bundled controls or extra management work, it still adds up.
Either way, it looks less like another flashy AI feature and more like another software management burden enterprises must absorb.
AI pricing is becoming a software management problem because enterprises are not just deciding whether to buy AI. They are dealing with AI as embedded product logic, bundled suite capability and cross-stack coordination overhead all at once.
The problem is not only what AI costs on paper. It is how difficult AI is becoming to isolate, measure and manage once it is woven into the software itself.
And that brings buyers back to the same two questions, now with more urgency: How much is our AI costing us? And what, exactly, are we paying for now?
James Alan Miller is a veteran technology editor and writer who leads Informa TechTarget's Enterprise Software group. He oversees coverage of ERP & Supply Chain, HR Software, Customer Experience, Communications & Collaboration and End-User Computing topics.
