AI inference is rewriting sales playbooks: How technical athletes win

Apr 30, 2026 · The Revenue Builders Podcast
🎧 PodShort 27 min squeezed to 2 AI SprinklerAS Sales Tech New
Episode artwork
Alex Varrel
EVP of Worldwide Sales at Cerebras Systems
The Revenue Builders Podcast
27 min squeezed to 2
Full episode from The Revenue Builders Podcast
Quotable Moments

The whole bottleneck with AI inference is memory bandwidth. Because you have to move all the weights and all that data again just to generate the next word. And inference is all about memory bandwidth and the way that they architected the world's largest chip, it solves for that.

Alex VarrelEVP of Worldwide Sales at Cerebras Systems
08:48

The great thing is though, I think humans are going to crave sellers' interpersonal and intrapersonal capabilities and communication skills and just being face-to-face with them. The email is dead, it's just been taken over by AI slop.

Alex VarrelEVP of Worldwide Sales at Cerebras Systems
14:40

The AE is got to be an extreme orchestrator that listens and understands where the value is and understands how to command the premium of the world's fastest tokens.

Alex VarrelEVP of Worldwide Sales at Cerebras Systems
21:30
Key Insights
  • The world has shifted from AI training to AI inference, where the focus is now on using trained models for various applications, leading to a massive increase in requests and complexity.
  • The primary bottleneck in AI inference is memory bandwidth, which Cerebras Systems addresses with its wafer-scale chip that co-locates memory and compute for extremely fast data movement.
  • A significant percentage of traditional software sellers are at risk in the evolving AI landscape if they don't adapt to new technologies and augment their productivity with AI tools.
  • Sellers will need to embrace AI to automate non-productive administrative tasks, allowing them to spend more time directly selling to customers.
  • The real differentiator in the AI-driven market is the speed and scale of the compute engine, as models and applications are converging in their capabilities.
  • The ability to translate technical capabilities into business value will be crucial for sales professionals, requiring strong business acumen and the ability to orchestrate complex buying committees.
  • Companies are now having to place bets on which AI-native companies will have durable, growing consumption of tokens and create new categories, rather than solely focusing on traditional large enterprises.
  • Sales leaders should encourage their teams to get hands-on with AI tools, understand the AI landscape, and build personal learning paths to future-proof their careers.
Metrics Mentioned
  • 140 gigabytes of data (Moved from compute to memory to generate just one token (one word or part of a word) on a 70 billion parameter model.)
  • 25-30% of time selling (Traditional sales professionals spend only 25-30% of their time actually selling, with the rest spent on non-productive activities.)
  • 500 milliseconds or less (Required latency for conversational AI to be effective and provide a good user experience.)
  • 200 milliseconds (Cerebras Systems enables AI responsiveness in 200 milliseconds, significantly faster than typical SLAs.)

RevBots.ai View:

  • AI Sprinkler teams must upgrade sellers to technical athletes who speak both AI and business.
  • SaaS Hoarders risk obsolescence if they don't automate admin work with AI copilots.
  • ARM-stage orgs will compete on inference speed: 200ms response times become table stakes.
  • Tab Hoppers should start building AI literacy now before the skills gap becomes unbridgeable.
🎧Full Episode:The Revenue Builders Podcast →