Show What Your Agents Are Worth With Tokenomics

As AI agents move from experiments into business workflows, a new question is reaching finance and engineering leaders: What are we getting back for the investment?

It’s a harder question than it sounds. A successful-looking response does not necessarily mean an agent did the job it was built to do. Token counts and model costs show consumption, but they do not tell us whether a workflow improved a business result, or if the work was done efficiently. And a quality score on its own can’t tell us whether that result mattered to the organization.

Building on the Tokenomics solution launched at .conf26, which attributes token spend across coding agents like Claude Code, Codex, and Cursor, we are launching Tokenomics by Splunk soon to be available on AWS Marketplace, the value layer for organizations running on AWS. It uses agent evaluations in Splunk Agent Observability to score agent runs against their intended jobs, then connects those scores to business-defined KPIs. The goal is to give finance and AI engineering owners evidence to help decide what agents to scale, improve, or stop.

Score Each Agent Run

A polished answer is not the same as a successful business outcome. Teams need to know whether an agent performed the task it was built for, and they need a way to compare that performance across agents, versions, and cohorts.

Splunk Agent Observability evaluations grade runs against the intended job. Teams can use Splunk evaluator models to assess every run and build custom ROI evaluations for their use cases. This visibility gives teams a consistent starting point for understanding agent quality and identifying meaningful differences in efficacy.

Connect Scores to Business KPIs

An evaluation score becomes more useful when teams can relate it to a business measure. Tokenomics by Splunk is designed to connect evaluation results with KPIs and baselines that customers define, such as cost per claim, cycle time, or net promoter score (NPS).

Trace-level results, tags, and metadata support KPI calculations, while business data provides context for the outcome. Teams can examine the scored run alongside its assigned cost, the KPI definition, baseline, and observed result. This context can create a more transparent basis for discussing how agent performance and business results relate.

Invest in What Pays Off

Once cost, agent quality, and business outcomes can be considered together, teams can compare the return on investment across agents and use cases. They can investigate where quality is at risk, identify which workflows merit further investment, and make evidence-informed decisions about what to scale, fix, or stop.

The approach is built around customer-defined measures, rather than a fixed KPI catalog or a single quality score. It is not a guaranteed ROI forecast or proof that an agent caused a business outcome. Instead, it gives teams evidence to review as they assess the relationship between agent runs and the business measures they already track.

Designed for Teams Building and Funding Agents

AWS partners building custom agents for clients can use Tokenomics to connect run-level evidence with outcomes that matter to the client. Enterprise finance leaders can examine AI investment in the context of their existing KPIs. AI platform teams and agent builders can compare usage and performance across teams and individual use cases.

The result is a common foundation for conversations that often happen in separate places: whether an agent did its job, what it cost, and whether its work is associated with a business outcome.

Tokenomics, soon to be available in AWS Marketplace, will help teams move beyond cost visibility to a clearer view of agent value — with evidence behind each number.

Availability and Next Steps

Tokenomics is in development. This month, the design partner program is offering a small group early access in private preview, with general availability in early 2027 on AWS Marketplace.

This small group will get hands-on work with Splunk engineers to create custom ROI evaluations, a direct line to the product team, and influence on the roadmap. This program is for teams that ideally have an agent running, interest in co-developing custom ROI evaluations, and capacity for weekly feedback sessions.

For more information about the private preview, sign up for the State of Observability: Where AI Meets Performance Event Series and join the December 15th Tech Talk about Splunk Agent Observability: Measuring AI Value with Splunk Tokenomics on AWS at 10am PST.

Check out our website and explore the documentation to learn more about Tokenomics.