Pratyusha Duvvuri, formerly an agent-observability lead at Galileo and now with Cisco, demonstrates a system her team calls “tokenomics” on AI Engineer.

One bill, three questions

  • Duvvuri says her team spoke with about 100 companies. “Tokenomics” meant different things to engineers, managers and finance staff.
  • Engineers want to choose models and track spend against a task or budget. Managers compare adoption, results and costs between teams. Finance needs ownership, budgets and a defensible account of spending.
  • A cheaper model reportedly scored 96% on one internal eval and a stronger model 99%. Whether the extra quality is worth the price depends on the task; these are her team’s measurements, not a general model ranking.

Reconcile the data before explaining it

  • Provider dashboards cover their own usage, while a company may buy from several providers under negotiated, changing rates.
  • The system needs provider organization IDs and API access, plus a workforce hierarchy and a separate finance hierarchy. A manager’s reporting chain may differ from the cost-center map finance uses.
  • Usage is refreshed on a reported daily cadence. Provider adjustments, staff moves and billing reconciliations mean previously displayed totals may change.
  • Finance views roll costs into departments; leader views show spend, usage and adoption by reporting chain or job family. Duvvuri describes role-based restrictions on who may inspect others’ data.

What the demonstration proves—and doesn’t

  • The presenter shows a working interface, with actual customer figures blurred. She says the team reached general availability after a ten-week sprint.
  • A projection chart appears in the demo, but she also describes a forecasting capability as not yet public. Do not treat every shown feature as generally available.
  • Cost attribution depends on accurate identity mapping, permission boundaries and later reconciliation. A usage chart alone cannot establish the business return on AI spend.

“Finance needs spend mapped to ownership.” — Pratyusha Duvvuri