What's Up with Tech?

Interested in being a guest? Email us at admin@evankirstel.com
Enterprise AI is everywhere, yet most teams still struggle to move beyond pilots. We sit down with Jason Olkowski, Chief Strategy and Transformation Officer at Creatio to unpack what the data actually says about who is succeeding with enterprise AI adoption and why. One surprising insight: the job-cut narrative is overplayed. The strongest programs use AI to redeploy capacity, automate the right work, and help people do more high-value thinking, selling, and serving.

We dig into what separates the top adoption leaders from everyone else, including how they embed AI across the business, automate a larger share of critical workflows, and scale AI agents across departments instead of leaving AI as a personal assistant. Jason explains why governance is not a compliance checkbox, but a practical system for building, deploying, orchestrating, monitoring, and auditing agentic automation safely in production.

You’ll also hear concrete examples of AI agents inside CRM, like morning account briefings, voice-based updates on the go, and automated post-meeting follow-ups that push clean intelligence back into the system of record. We wrap with pragmatic advice for CIOs under board pressure: pick a governed, LLM-agnostic platform, start with repeatable workflows for fast wins, then reinvest the freed capacity into tougher friction points that slow the business down.

If you’re trying to turn enterprise AI strategy into measurable impact, subscribe, share this with a colleague who owns AI governance, and leave a review with your biggest “where do we start?” question.
Learn more about Microsoft's ISOC announcement: https://fandf.co/4iQ9b5n
Support the show
More at https://linktr.ee/EvanKirstel

Interested in being a guest? Email us at admin@evankirstel.com
AI gets a lot messier the moment it leaves the chat window and touches real sensors, machines, and buildings. We sit down with Anthony Canterbury, Head of Engineering and Architecture, JIG-SAW US to dig into AIoT, the intersection of AI and industrial IoT, and why reliability and safety matter more than hype when you are dealing with edge computing, automation, and real operational risk.

We talk through what has actually changed over the last couple decades of machine intelligence, from expert systems and early neural nets to today’s transformer models. Then we get practical: device onboarding, decoding weird formats, parsing, and data normalization. Anthony explains why pairing LLMs with deterministic algorithms can remove a huge amount of manual integration work, and why in-house benchmarking beats trusting public leaderboards when your data is specialized and messy.

From there we tackle the uncomfortable part: hallucinations. If you simply bolt a chatbot onto an existing platform, you can see extreme error rates on IoT workflows. The path forward is architectural, not cosmetic: constrain the model, validate outputs, specialize where needed, and keep a human in the loop so alerting systems evolve without becoming a false-alarm factory. We also dig into treating cloud APIs like “devices,” plus when small, custom edge models win on security, cost, and accuracy for sensitive environments.

If you care about industrial AI, IoT security, edge AI, and building systems that work in the real world, this one is for you. Subscribe for more, share this with a CTO or engineer who is building with AI, and leave a review with your biggest question about AIoT reliability.
Learn more about Microsoft's ISOC announcement: https://fandf.co/4iQ9b5n
Support the show
More at https://linktr.ee/EvanKirstel

Interested in being a guest? Email us at admin@evankirstel.com
Only 35% of executives say AI consistently delivers the business outcomes they expect, and that single stat explains why “AI everywhere” suddenly feels risky. We sit down with Sushant Tripathi Vice President & North America Lead – AI & Digital Transformation from TCS to talk about what changes when enterprises move beyond experimentation and put autonomous agents into real business operations, where the cost of a bad decision is measured in customer harm, regulatory exposure, and millions in downtime. 

We dig into why classic IT reliability metrics like uptime and latency are no longer enough. AI reliability now includes service-level experience, trust, and the ability to prove what happened and why it happened. Sushant breaks down how enterprises should think about trustworthiness as an engineering responsibility, not just a compliance checkbox, and why “more model capacity” does not automatically create dependable outcomes without strong data, integrations, evaluation, and observability. 

We also get practical about controls: why many companies can detect AI failures but far fewer can stop them, what “kill switch” thinking looks like, and how an AI maturity journey changes the right level of autonomy. Finally, we talk governance and ROI: who inside the organization must own the policies that let agents act, and how to turn individual productivity gains into measurable enterprise KPIs like revenue, faster resolution, and better customer experience. 

If you’re building enterprise AI, agentic AI, or AI governance programs, this is the reliability playbook you’ll want. Subscribe, share this with a leader who owns risk or operations, and leave a review with your take: what is the first guardrail every autonomous agent should have?
Learn more about Microsoft's ISOC announcement: https://fandf.co/4iQ9b5n
Support the show
More at https://linktr.ee/EvanKirstel

Interested in being a guest? Email us at admin@evankirstel.com
AI agents are starting to run real customer workflows, and the hardest part isn’t getting a demo to work, it’s knowing the agent is behaving correctly at scale. We talk with Mukul Karnik Vice President of AWS Observability at Amazon Web Services about what “observability for AI agents” actually means when the system can be up, fast, and still wrong. Think guardrails, accuracy, and policy compliance alongside the classic metrics like latency and availability.

We get specific about failure modes that don’t look like normal software bugs: an agent that confidently approves a hazmat return because a user asked the question in an odd way, or an agent that answers something it should refuse. Mukul explains how evals act like tests for agent behavior, how tracing and telemetry help you see tool calls and dependencies end to end, and why a unified view matters when the agent relies on apps, databases, and infrastructure to get the job done. We also dig into what CloudWatch Omni is aiming for with an AI-first experience, including deeper investigations that can be driven by a DevOps agent for root cause analysis.

We also cover how teams can start earlier in the workflow with a free local IDE plugin to visualize agent traces, then bring the same evals and monitoring into production. On top of that, we talk multi-cloud realities, supporting different agent frameworks, and why the line between observability and security keeps getting blurrier as more signals live in the same logs and telemetry. If you’re building agentic systems, this is a practical look at monitoring, debugging, and building trust in autonomous behavior.

Subscribe for more conversations on AI infrastructure and production engineering, share this with a builder on your team, and leave a review with the toughest agent reliability problem you’re trying to solve.
Support the show
More at https://linktr.ee/EvanKirstel

Interested in being a guest? Email us at admin@evankirstel.com
Robots are leaving the lab, and the real world does not hold still. We sit down with Dr. Amarjot Singh Founder and CEO of Skylark Labs to unpack what “physical AI” actually means when a machine has to operate outdoors, in hazardous environments, and keep learning after deployment, without a safety net of perfect data or constant cloud access.

We walk through mission-critical examples where edge AI and continual learning are not optional. Think military tarmacs where a 1 cm piece of debris can cripple a jet engine, and where new debris types and surface imperfections can trigger costly false positives. We also get into why on-device learning changes the game: lower latency, better privacy, and resilience when connectivity drops. Amarjot shares how Skylark shrinks models onto edge hardware so robots, robotic dogs, and other platforms can adapt in the field instead of waiting for a retraining cycle back at HQ.

Then we go deep on drones and UAV autonomy. The conversation moves beyond “spot the crack” into drones that can take action with robotic arms, from cleaning solar panels in dusty deserts to applying quick fixes on industrial equipment. That upgrade introduces hard control problems: once a drone touches the world, it has to handle new forces, wind, and unexpected obstacles like scaffolding, all while staying stable and safe. We also cover what still holds the industry back, including data bottlenecks and rugged humanoid hardware needs like waterproofing, safer batteries, and stronger hands.

If you care about robotics, edge computing, autonomous drones, and mission-critical humanoid robots, subscribe, share this with a friend, and leave a review so more people can find the show.
Support the show
More at https://linktr.ee/EvanKirstel