In this episode, Zachary Hanif, VP of AI, ML, and Data at Twilio, joins Amir to talk about the engine behind B2B AI innovation. From selecting the right tools to navigating the shift from POCs to production, Zachary offers an insider's look at how enterprises can thoughtfully and effectively integrate AI.

We unpack:

The danger of "boiling the ocean" with AI

Why chatbots aren’t always the right starting point

What makes an AI POC actually valuable

And why UX in the age of AI needs systems thinking

💬 “If you come into it with a technology and not a firm understanding of the problem, you're going to solve a problem that isn’t there — and at best, you'll just end up with a great tech demo.” – Zachary Hanif

🔑 Key Takeaways

Start with the use case, not the tool: Jumping in with LLMs without a clear business problem leads to superficial results.

UX in AI is different: You’re not just designing for humans—you're designing for a human-model-human interaction loop.

POCs must build trust: Especially with generative AI, proof-of-concepts must feel reliable and human-like to succeed.

AI increases surface area: Models introduce new attack surfaces and complexities. Security, observability, and model risk management are critical.

Think systems, not screens: LLMs change how users interact with software. This demands broader thinking from designers and PMs.

⏱️ Timestamped Highlights

00:00 – Intro to Zachary Hanif and Twilio's AI mission

02:05 – Why most companies are AI tool users, not tool makers

04:25 – The “chatbot temptation” and why it might not be the best starting point

06:00 – UX lessons from Google’s early search box vs. today’s LLMs

08:30 – Why we’re still early in discovering transformative AI use cases

11:55 – How AI changes what a good POC looks like

14:59 – Should AI UX be its own discipline?

18:23 – How to know when a POC is ready for production

22:12 – Dealing with AI’s expanding “surface area” and model drift

25:56 – Why model risk management matters more than ever