AI Impact

Block's banking engineering team is 400% more productive per person than a year ago. They got there by applying a five-level AI adoption ladder at the workflow level, inside a heavily regulated vertical where compliance checks sit at every layer. Adoption spread organically through role modeling and AI champions on each team.

Bhupesh Bansal, Head of Engineering of Block Banking at Block, sat down with Ameya at SFELC Annual 2026 to break down how his org moved to flat, problem-owning squads of five, replaced siloed roles with "all builders," and turned engineers into end-to-end problem owners.

****All opinions expressed are those of the interviewees and do not necessarily reflect the views of their respective companies.****

Topics discussed:

- Flattening org layers into problem-owning teams of five

- Replacing siloed roles with cross-platform "all builders"

- Applying the AI adoption ladder at workflow level

- Driving organic AI culture through role modeling

- Human-in-the-loop quality checks for banking compliance

- Shifting hiring toward AI-native development interviews

- Engineers owning problems end-to-end including customer discovery

Listen to more episodes: 

Apple 

Spotify 

YouTube

When most of your code is written by AI, the job shifts from writing to reviewing, and reviewing is where the work piles up. Rahul Sanghavi, Director of Engineering at Twilio, describes one engineer who generated 60 to 70 PRs in a few days, more than anyone could review.

Recorded with Denis Scott at ELC Annual 2026, the conversation covers how his team uses tests as the gate for shipping safely, how AI now runs inside their on-call rotation, and how they are capping per-engineer spend as AI costs climb.

Topics discussed:

• Shifting from writing code to reviewing code

• Code review as the new engineering bottleneck

• Using tests as the gate for shipping AI-written code

• Writing tests deep enough to catch edge cases

• Running AI inside on-call to debug and mitigate incidents

• Setting per-engineer AI spend limits as costs climb

• Planning 2027 AI budgets after a free-hand Q1

• Weighing in-house LLMs to control rising AI spend

  
  

  
  

  
  

****All opinions expressed are those of the interviewees and do not necessarily reflect the views of their respective companies.****

Listen to more episodes: 

Apple 

Spotify 

YouTube

The engineering leaders who stay relevant over the next 24 months will refuse to become “leaders of leaders”. Abhishek Agrawal, Senior Director of Engineering at Oracle, still writes code and kills weak prototypes himself, even while running three layers of leadership.

Recorded at ELC Annual 2026, Abhishek joins Ameya Kanitkar to break down leading a 250-person organization inside Oracle Health and AI at startup speed, why team structures stay fluid, how AI closes the learning curve, and why staying green beats going ripe.

Topics discussed:

• Why engineering leaders must stay hands-on, not just managers

• Killing weak prototypes before they reach production

• Running a 250-person org like a startup

• Structuring teams around ownership, not headcount

• Using AI to close the learning curve fast

• Keeping team structures fluid as tooling evolves

• Why continuous learning separates leaders who last

  
  

****All opinions expressed are those of the interviewees and do not necessarily reflect the views of their respective companies.****

Listen to more episodes: 

Apple 

Spotify 

YouTube

AI agents handle the work. Employees walk the dog. Sounds like progress, but without a workforce ready to absorb AI, every enterprise dollar invested is lost.

Volker Schrank, Founder of Employee Xperience Labs, walks Russ through his “HR for AI” framework, why leading from the top is the adoption strategy that makes the real difference, and where companies go wrong by automating processes that only create value through human involvement.

Topics discussed:

• Distinguishing “AI for HR” from “HR for AI”

• Integration and usage as AI’s true limiting factors

• The “walk the dog problem” of AI productivity gains

• Leading from the top to drive enterprise AI adoption

• Moderna’s experiment merging the CHRO and CIO roles

• Automating HR shared services with AI agent workforces

• AI-powered manager coaching at scale

• Knowing when to stop automating human processes

  
  

Listen to more episodes: 

Apple 

Spotify 

YouTube

SurveyMonkey put AI on top of its curated data and got 14% accuracy. Nine months of context engineering later, that number hit 98%. Meenal Iyer, SVP, Data & Enterprise AI, walks Ameya through the semantic layer strategy that closed the gap.

Meenal talks through her GET framework, Growth, Effectiveness, Trust, and why she runs enterprise AI as a top-down mandate. SurveyMonkey gives each team its own AI budget with clear return expectations and uses a 30/60/90 day value framework to decide what stays and what gets cut. Next on her roadmap: capturing tribal knowledge so agents can work with real institutional context.

Topics discussed:

• Building a repeatable data playbook across industries

• The GET framework for growth, effectiveness, and trust

• Context engineering a semantic layer from 14% to 98%

• Capturing tribal knowledge for agent workflows

• Supervised flight hours as a model for agent autonomy

• Running enterprise AI as a top-down mandate

• Amending job frameworks to require AI proficiency

• The 30/60/90 day value framework for AI investments

Listen to more episodes: 

Apple 

Spotify 

YouTube

AI is not clearing work off the calendar at VF Corporation. It reveals more of it. Capacity gained gets redirected straight into the backlog leaders never had time to touch. Workloads did not get lighter. Ambition got bigger.

Andrew Weaver, Senior Director of Workforce Intelligence and AI Transformation, walked Russ Fradin through intake meetings, enterprise guardrails, and why an answer that is right 80 percent of the time sometimes belongs in a plain automated workflow.

Topics discussed:

• Why AI reveals more work instead of removing it

• Building HR products rather than AI solutions

• Intake meetings that separate AI from plain automation

• The 80 percent output problem in enterprise workflows

• Enterprise guardrails and the digital technology partnership

• Learning from shadow usage instead of policing it

• Smart experimentation without duplicating work across functions

• Community of practice for scaling super user wins

Listen to more episodes: 

Apple 

Spotify 

YouTube

Adoption without accountability is just expensive experimentation. Smriti Kirubanandan, Managing Director of Enterprise AI Transformation at Innovaccer, reframes the question every enterprise leader spending on AI should be asking.

Smriti tells Russ how her team cut a 32% claims denial rate to 11% in 14 months, recovering $23 million in one year. She breaks down why data quality must come before model performance, why every role must function as an AI steward, and where physical AI is headed next.

Topics discussed:

• Why AI adoption and accountability must be simultaneous

• Cutting claims denial rates from 32% to 11% with AI

• Measuring data quality before measuring model performance

• Building circular AI strategy across every organizational level

• Using small language models to control cost and accuracy

• Treating continuous data monitoring as a lifelong practice

• Why physical AI is the next major wave of change

Listen to more episodes: 

Apple 

Spotify 

YouTube

Getting tens of thousands of employees to adopt AI comes down to trust, skills built for each role, and change that moves top down and bottom up at the same time.

Matthew Smart, Director, AI Strategy & Business Transformation at Cardinal Health, joins Russ to break down agent hype, why token pricing broke ROI math, pitching value over cost savings, and treating data readiness as the real work under every AI use case.

Topics discussed:

• Driving AI adoption without scare tactics or job threats

• The four levels of agents most leaders conflate

• Sequencing quick wins before moonshot AI bets

• Why token based pricing broke traditional ROI calculations

• Pitching AI to CEOs on value over cost savings

• Building trust through role specific prompt libraries

• Treating data readiness as the real AI foundation

• Governance across legal, privacy, compliance, and architecture

  
  

Listen to more episodes: 

Apple 

Spotify 

YouTube

Fanatics runs 1,000 websites on a single platform, scaling to peak orders per minute when a team wins a championship. Sashanka Vishnuvajhala, SVP of Engineering, tells Ameya how AI reshaped the way his teams organize, ship, and review code at Fanatics.

Three-person triads now tackle complex customer-facing projects in rapid iterations. Intent specs replaced traditional JIRA tickets, and the increased throughput exposed bottlenecks in CI/CD and code reviews that had been hidden for years.

Topics discussed:

• Running 1,000 websites on a single elastic platform

• Three-person triads for complex customer-facing projects

• Intent specs replacing traditional JIRA tickets

• Code reviews as the new throughput bottleneck

• Building a model routing judgment layer

• Hiring for aptitude over specific technology experience

• AI-native young talent approaching problems differently

• Why increased speed exposed every leaky pipe

Listen to more episodes: 

Apple 

Spotify 

YouTube

Pinching pennies on AI tokens is how AI transformation fails. Michael Laudon, Chief Product and Technology Officer at Quest Software, argues the real question is whether that spend produces measurable output, and no single metric answers it.

Michael joins Russ to unpack how Quest runs its AI transformation, from a dedicated AI center of excellence to a governance model tied to business objectives and a deliberate break with the past that immerses eager teams in AI-native work end to end.

Topics discussed:

• Measuring AI value through combined metrics beyond token counts

• Running an AI center of excellence for enablement

• Tying AI transformation to business objectives and governance

• Breaking with the past through AI-native project teams

• Spreading AI proficiency with a land-and-expand pollination model

• Preparing data and semantic layers so models perform

• Splitting model strategy between bleeding-edge and commodity tasks

• Blurring roles as developers, PMs, and QA converge

  
  

Listen to more episodes: 

Apple 

Spotify 

YouTube