SAP is trying to move the AI pricing conversation away from tokens and toward business outcomes.
And honestly, that is probably the right instinct.
Most companies do not want to think about AI in terms of tokens, prompts, model calls, or technical consumption metrics. They want to understand something much more practical: what is this AI actually improving inside the business?
SAP is taking the debate around the economics of AI seriously. It is not the only enterprise vendor trying to move beyond raw token-based pricing, but its approach is interesting because the conversation is happening inside the business processes where SAP already has deep context: finance, procurement, HR, supply chain, consulting, and operations.
The reality is that CIOs, and of course CFOs, do not want to keep getting unpleasant surprises every month when the bill lands in their inbox. SAP seems pretty aware of that, and wants to find a way to connect consumption with business value instead of focusing merely on how many tokens your team burns every month.
The question is: are they getting it right?
In this article, we will try to add a bit of clarity around all these changes and something that is still relatively new for many companies: controlling the effects of what they pushed so aggressively over the last few years.
For a while, the mantra repeated in boardrooms was:
“Use AI, or we are out of business.”
Now leadership seems to be changing its mind:
“Use AI, but control what you are spending… Otherwise, we may end up out of business anyway.”
Why Is Tokenomics Becoming a Bigger Topic in SAP?
For years, SaaS pricing was relatively easy to understand. You paid for users, modules, licenses, subscriptions, or infrastructure capacity.
AI changes that logic because consumption is harder to predict. AI usage can appear in many small moments across the business. Joule may help someone complete a task. An agent may run in the background. A document may be processed automatically. A consultant may use AI to move faster during an implementation.
And that is where the debate starts.
More AI usage does not automatically mean more business value. It may mean adoption is growing. It may even mean people are becoming more comfortable experimenting with AI, which is necessary. But if consumption becomes the main signal of progress, companies can easily drift into tokenmaxxing: teams trying to climb internal usage leaderboards, employees using AI to show activity, or agents running on workflows that were never tied to a real business case.
That is the uncomfortable part of SAP tokenomics.
Tokenmaxxing is not always a sign of bad use. But it is not the best incentive either. It rewards activity before proving whether that activity improved anything measurable.
The question is not only how AI is priced. It is whether companies can connect consumption with business value before usage becomes another metric that looks good on a dashboard but says very little about ROI.
How Is SAP Structuring AI Consumption?
1. Base AI vs. Premium AI
SAP is separating its AI capabilities into two broad categories: base AI and premium AI.
Base AI includes capabilities SAP sees as expected parts of the product experience. These are features that are included in the existing subscription because they are becoming table stakes. A simple example would be an AI-supported task inside Concur, such as uploading and processing a receipt as part of the normal application experience.
Premium AI is different. These are the capabilities SAP believes create additional business value and therefore justify additional commercial treatment. This is where Joule Agents, Joule for Consultants, developer capabilities, finance use cases, supply chain use cases, HR use cases, and other advanced SAP Business AI capabilities enter the conversation.
For those premium capabilities, SAP uses AI Units.
2. What Are AI Units?
AI Units are SAP’s commercial model for consuming premium Business AI capabilities across its portfolio.
They are not exactly the same as tokens. A token is a technical unit related to how language models process text. An AI Unit is more like a commercial consumption unit that gives customers access to premium AI capabilities across different SAP applications and business functions.
That distinction matters because SAP does not want customers to feel they are simply paying for abstract AI usage. SAP wants to position AI Units as a way to access AI capabilities that should be tied to business outcomes.
So the framing is not supposed to be:
“How many tokens did we consume?”
The framing SAP wants is closer to:
“What value did this AI capability create for the business?”
Why Does SAP Want to Move Away from “Tokens”?
Because Tokens Are Not Business Outcomes
Philipp Herzig, SAP’s Chief AI Officer, made a very clear point in an interview with Stuart Lauchlan for Diginomica: “Our customers really don’t like tokens; they like business outcomes.”
That statement captures the whole debate.
Tokens can show consumption, but they do not show value. Herzig used a simple analogy to explain it: measuring tokens is like measuring a company’s performance by how much electricity it consumes. You may be using a lot of electricity because the lights are on all night, but that does not mean the company is performing better.
That is the point SAP is trying to make with AI Units.
The question is whether that consumption can be connected to something the business can actually measure: fewer billable hours, lower days sales outstanding, faster processes, or another operational metric that matters.
Herzig’s argument is that SAP’s commercial model should start with value and work backwards. In his words, the basic principle is “value-based and then working backwards.”
What Is SAP’s Value-Based Argument?
SAP’s argument is that the commercial model should start with a value hypothesis.
Take Joule for Consultants as an example. The value is not that a consultant uses AI more often. The value is that Joule may help reduce research time, avoid rework, accelerate implementation decisions, and lower billable consulting hours during SAP projects.
If that happens, the business outcome is relatively easy to understand. The customer spends less time and money on implementation work, while SAP captures part of the value created through its premium AI model.
But this is where CIOs and CFOs need to ask a simpler question: are we getting more value than what we are spending?
A company may start spending more on AI because it is trying to find value. That is not necessarily wrong. Some experimentation is needed. But if that experimentation happens without direction, ownership, or a clear connection to business processes, the cost can grow faster than the return.
Imagine a consulting team using Joule heavily during an SAP implementation. Consultants ask it to summarize workshops, rewrite notes, explain configuration options, draft status updates, and prepare internal documentation. The usage looks strong. The team may even climb the AI adoption dashboard.
But if none of that reduces billable hours, avoids rework, accelerates a project milestone, or becomes part of a repeatable delivery method, the business case is weak. The company consumed more AI, but it did not necessarily improve the economics of the implementation.
At some point, leadership may look at the AI bill and decide to slow everything down.
That is the real danger. Poorly directed consumption does not only hurt the use case that failed. It can close the tap for other AI projects that may have had a much stronger business case.
Where Is the Customer Concern?
Trust Is Still the Main Problem
The challenge is that many companies are still early in their SAP AI journey.
They may be interested in Joule, but they do not yet know which use cases will generate ROI, how much AI Unit consumption to expect, how usage will grow over time, or how to separate valuable AI adoption from noise.
That is why customers are cautious. They have already seen cloud consumption models become difficult to predict in other areas of technology. AI creates a similar concern.
How Does SAP for Me Fit into This?
SAP for Me is an important piece of this conversation because it gives customers a place to see AI Unit consumption across the SAP portfolio.
That matters because companies need visibility before they can build trust. If AI Units are consumed across different business areas, products, users, agents, and workflows, customers need a centralized way to understand what is happening.
SAP has also said that budgeting capabilities are expected to be added, which would allow customers to manage AI consumption across different categories or buckets.
That is important because SAP for Me could become part of the AI FinOps layer inside SAP environments.
Joule Is Moving Beyond the Copilot Model
This debate becomes more important because SAP is positioning Joule as more than a simple copilot.
Joule is not just an assistant that answers questions or helps users complete isolated tasks. SAP is moving toward a broader vision where Joule connects assistants, agents, business data, process context, and workflows across SAP and non-SAP systems.
That creates a bigger opportunity, but also a bigger cost-control problem. When agents start acting in the background, the economics become harder to follow.
An agent may trigger multiple actions, consult different sources, process documents, support decisions, or move a workflow forward without a user seeing every step. If it is well configured, that can create real value. If it is not, it can quietly increase consumption and hit the AI bill before anyone understands what changed.
AI FinOps Will Become Necessary in SAP Tokenomics
The concern around AI bills is already moving beyond the SAP ecosystem.
Companies are realizing that the same AI adoption they pushed so aggressively can create a new kind of cost problem. First comes the pressure to experiment. Then come the agents, pilots, dashboards, and internal usage goals. And only after the bill arrives does the conversation shift to governance, limits, and ROI.
There are already enough warning signs. Business Insider reported that a fintech company said one employee burned through more than $80,000 in AI coding tokens while building a shooter game. On the same note, Uber reportedly capping employee AI use after burning through its 2026 AI coding budget in just four months.
The same concern is showing up in workplace conversations. In one Reddit thread, an employee described a company of around 500 people switching off most AI tools after reportedly racking up a £300,000 token bill in a single month. Another commenter described the shift from “use AI 5 days a week” to “conserve your tokens and be mindful of the agents you’re using.”
SAP has tools that can help create more visibility into AI Unit consumption, especially through SAP for Me. But visibility alone does not solve the problem:
First, companies need to select the right use cases. Not every process deserves AI. The best candidates are usually processes with enough volume, friction, cost, or manual effort to justify the consumption of AI Units.
Second, the process or agent needs to be designed correctly. An agent may be too broad, consult too many sources or execute unnecessary steps. All of that can increase consumption without creating proportional value.
Third, the agent needs to fit the real SAP landscape. SAP rarely operates in isolation. A company may have S/4HANA, SuccessFactors, Ariba, Concur, BTP, Integration Suite, Datasphere, legacy systems, approval flows, master data, roles, and permissions. The value of Joule or a custom agent depends on whether it can operate with the right context without breaking governance.
Finally, companies need economic and operational control. This is where AI FinOps becomes necessary: who owns the consumption, what budget each use case has, when it should be reviewed, which KPIs matter, which alerts are useful, what should scale, and what should be stopped.
That is where a SAP partner can help.
As SAP partners, we can help companies organize their SAP AI strategy, identify the right use cases, configure agents with cost and governance in mind, and put the controls in place to manage AI consumption more carefully.
If you want to avoid unpleasant surprises in your SAP AI bill, book a discovery call with our team and tell us what you are trying to build.