Tech Connects

The episode contains some substantive ideas - particularly around workflow redesign, fraud risk stratification by geography/job level, and the data quality problem in HR - but is hampered by significant throat-clearing, repetition, and abstract framing. Guest often restates high-level principles (fit, personalization, intelligence) without drilling into mechanics or novel specifics that an HR operator wouldn't already intuit.

“work is fundamentally getting reset...all the workflows will get reset...we have right now to really rethink about not hundred percent, at least 40 to 50% how we are doing business for the last 30 years”

“the fraud is entirely different...the way you deploy agents are different, the way you deploy workflows are different”

While the guest articulates a sensible position that hiring varies by job level, geography, and industry (not a fresh insight), the framing around AI agents and identity fraud risk by region adds modest originality. However, the core moves - automating low-touch roles, personalizing high-touch ones, using data - are well-trodden. The e-commerce analogy and the assertion that standardization 'complicates the equation' are slightly contrarian but underdeveloped.

“the lower the comp, the higher the volume, the higher the automation...as you get higher in the knowledge working management, more senior levels, you just need a more personal appraisal”

“in E commerce you can learn a lot...when you're really working with the people, it's not just you like. They also have to like”

Mahi Bay Reddy is the CEO of Phenom, a real HR-tech operator with direct experience deploying AI agents at scale across multiple customer verticals (healthcare, retail, manufacturing, finance). He demonstrates legitimate domain expertise and has skin in the game. However, the episode reads partly as a product pitch, and guest does not bring customer-specific war stories or failure modes that would elevate caliber further.

“Phenom is an applied AI company specifically focused on HR domain...we are very simple. We want to make sure companies can hire faster and retain the people and grow them within the company”

“We have seen the same thing in Thermo Fisher...you can automate 80% of the use case. And the last 20% of the last mile recruiting efforts is where humans are required”

The transcript includes several named customer examples (LRC, Thermo Fisher, CircleCare, Kohl's, Phenom's acquisition of applied.ai) and some concrete metrics (80% automation in manufacturing, fraud climbing in specific verticals: tech, financial services, insurance, pharma). However, specificity is often undercut by vague language ('a lot of benefits,' 'we see a lot'), missing timelines, and lack of dollar figures or precise performance deltas. The Kohl's Australia example is more vivid but largely anecdotal.

“Thermo Fisher...how many people you're hiring in a manufacturing unit, you can automate 80% of the use case”

“fraud is climbing up because of the Gen AI infrastructure...more in tech space, more in financial space, financial services space”

Host asks reasonable foundational questions (apply process, interviewing, onboarding, identity verification) and does follow up on acquisitions and predictions. However, follow-ups are often soft and rarely push back on claims or ask for tighter proof. When guest gives abstract answers (e.g., on standardization), host accepts rather than probe. Host misses opportunities to challenge the 80% automation claim, ask for failure cases, or press on fraud risk quantification. The exchange reads more as a cordial product overview than a rigorous examination.

“Do you have a favorite interview question that you ask?”

“Right, Right, right.”