Saurabh Agrawal, CTO: Software Engineering at DVT.
As organisations move generative AI from experimentation into everyday business use, the cost of running large language models is becoming harder to predict and control. Tokens are the fundamental unit that determines AI cost, performance and scalability, making LLM tokenomics an increasingly important discipline for enterprises. DVT’s upcoming Insights webinar will examine how organisations can better understand, manage and govern this fast-growing area of technology spend.
Taking place on 3 September 2026, The Economics of Tokens: From Unit of Measure to Enterprise Governance will be presented by Saurabh Agrawal, CTO: Software Engineering at DVT.
Each interaction with a large language model consumes tokens, the units used to process prompts and generate responses. Token consumption influences cost, latency and model performance, yet many organisations still lack a clear view of what drives usage, why prices vary between providers and how AI consumption should be governed across teams and projects.
“Token usage now plays a direct role in the commercial viability of enterprise AI,” says Agrawal. “Organisations need to know what they are consuming, what value they are getting in return and where better architectural decisions can reduce unnecessary cost.”
Beyond the price per token
Published model pricing tells only part of the story. The same prompt can result in different costs depending on the provider, model, context window, response length and application design.
During the webinar, Agrawal will explain how tokens are calculated, consumed and billed, and examine the factors that affect latency, performance and cost. He will also discuss how organisations can weigh model quality against affordability, speed, security and enterprise risk.
The session will also examine the architectural and governance decisions companies face as AI adoption accelerates. These include selecting the right models, implementing intelligent routing strategies and evaluating API-based services against self-hosted and open source alternatives.
Attendees will learn how techniques such as prompt optimisation, caching, context optimisation, context compression and output controls can reduce unnecessary token usage without weakening the quality of results.
The webinar will address the governance structures needed to monitor AI consumption across the business. This includes setting budgets, tracking usage by team or project and defining guidelines for model selection, deployment and ongoing optimisation.
Topics to be covered include:
- How LLM tokens are calculated, consumed and billed.
- The main drivers of token usage, latency and cost.
- Why similar prompts can cost differently across models and providers.
- How to balance cost, performance, quality and enterprise risk.
- Techniques for reducing consumption through caching, prompt optimisation, context optimisation, context compression and output controls.
- When to use premium, smaller or self-hosted open source models.
- Best practices for monitoring, budgeting and governing AI consumption across teams and projects.
The session is aimed at executives, architects, engineering leaders, developers and AI practitioners responsible for implementing or managing generative AI in the enterprise.
Join DVT Insights on Thursday, 3 September 2026 for The Economics of Tokens: From Unit of Measure to Enterprise Governance. Register for the webinar via this link.
