Running AI in Production: Context, Workflow, Cost and Control
Four sessions at Innovate Americas focused on one question: how do operators run AI in production and keep it trusted, governed and affordable?
Operators have proved AI works in pilots. The harder question now is how to run it in production while keeping it trusted, governed and affordable. This Appledore Research event report draws on four sessions chaired by Patrick Kelly at TM Forum Innovate Americas 2026 in Dallas, with speakers from AT&T, EchoStar, Lumen, Interbank and Databricks.
We examine four themes: grounding agents in trusted enterprise context, redesigning workflows so AI drives decisions, managing token economics, and enforcing runtime governance over autonomous agents. We also give our view on where these threads converge, and why cost and control are becoming two sides of the same architectural question.
The report is for CSP AI, data and operations leaders moving agents from sandbox to production, and for vendors deciding where to position in the emerging control layer. Readers get a grounded view of what frontline operators are actually doing, and what it means for their own AI roadmaps.
Perspectives on token optimization, guardrails, governance and more from:
- Nevash Pillay, Head of Global Telecom Industry, Databricks;
- Idilio Moncivais, Head of Financial Intelligence and Data Architecture
- Lumen Christian Livia, CTO, Interbank
- Mark Austin, VP, Data Science, AT&T
- Singh Sardar, Director of AI, EchoStar
This report is available as part of an Appledore subscription.
To access free and subscription content please register or if you have an account, login.
For options about purchasing this single report, or to discuss subscription options, contact our team at sales@appledorerg.com
