EPISODE 52

Jason Shafton and Kevin Henrikson unpack where AI is genuinely useful and where it starts to create more noise than leverage, using examples from AI email triage, long chat memory drift, and agentic workflows. Kevin explains how memory can become polluted when models start treating their own prior inferences as fact, including a prompt he used to compare what an AI thought was “ground truth” against what he had actually told it. From there, the conversation shifts into a practical framework for building AI systems and human teams the same way: define the job, provide the right tools and access, layer in review and guardrails, and judge success by whether time spent together compounds into more output. They close by connecting startup hiring, high-agency operators, and founder-led culture back to the same core test they use for AI: does this person or tool create leverage, or does it create drag?

CHAPTERS

00:00 – AI memory drift and false “ground truth”

01:24 – Testing AI email triage and the risks of over-filtering

03:13 – Good AI versus bad AI in real workflows

05:31 – Why controlled memory leads to more consistent AI outputs

08:29 – How to apply AI to workflows that currently rely on humans

11:12 – Building multi-agent content systems with clear roles and QA

13:40 – Hiring high-agency people for early-stage teams

16:01 – The “pick up the shovel” standard for startup operators

22:36 – The real test for both employees and AI: leverage or drag

26:16 – Founder Mode Top 5 Takeaways

LINKS

Connect with Kevin Henrikson

LinkedIn • X/Twitter

Stay Connected with Founder Mode

Subscribe to our newsletter

Connect with Kevin

LinkedIn • X/Twitter

Connect with Jason

LinkedIn • X/Twitter