Inverted Podcast #21: AI Regulations and Human-in-the-Loop

Build AI products that stay compliant before rewrites become inevitable. In Inverted Podcast #21, host Dario Salice talks with Dexter Awoyemi, an independent AI systems architect, about regulation, risk classification, and human oversight as product design inputs, not late-stage gatekeeping.

Dexter argues for “clarity before code”: start with intent, risk tier, and human-in-the-loop patterns. The conversation covers the EU AI Act, turning obligations into executable knowledge graphs, GDPR-style lessons about baking controls into the product, auditability and liability, and why early compliance work is not wasted even if rules keep evolving.

Listen on Spotify

About the guest

Dexter Awoyemi is an independent AI systems architect working at the intersection of AI, regulation, and product design. He joins Dario to unpack how teams can reason about obligations without drowning in hundreds of pages of legal text.

Key takeaways

Clarity before code: classify risk and define oversight before picking models and frameworks.
Human-in-the-loop is non-negotiable for accountable AI: it shapes UX, audit trails, and liability.
Map regulations like the EU AI Act into executable knowledge so product and eng can reason day to day.
Use AI for needle-in-haystack tasks carefully; do not outsource narrative judgment blindly.
Compliance coming late creates expensive rewrites; startups and enterprises fail this differently.
Like GDPR, AI rules are likely to set a global default: early controls are future-proofing.

Topics covered

AI governance and “clarity before code”
EU AI Act as an executable knowledge graph
How regulation changes day-to-day product building
Needle-in-haystack versus narrative uses of AI
Risk classification and human oversight
GDPR parallels: baking compliance into the product
Human-in-the-loop UX, auditability, liability
Where to start with risk tiers and future-proofing

Chapters

00:00:13 Introductions
00:00:44 AI Governance: “clarity before code”
00:01:13 EU AI Act to executable knowledge graph
00:02:45 How regulation changes day-to-day product building
00:03:34 Using AI wisely: needle-in-haystack vs narrative
00:05:13 Start with risk classification + human oversight
00:07:33 GDPR parallels: bake compliance controls into the product
00:10:39 Human-in-the-loop patterns: UX, auditability, liability
00:21:28 Why compliance comes late: startups vs enterprise committees
00:34:48 Where to start: EU AI Act risk tiers + future-proofing

FAQ

Who should watch this episode?

AI and platform product managers, founders in regulated markets, compliance partners who want earlier collaboration, and engineers designing review and escalation UX.

What is “clarity before code”?

Decide intent, risk class, and human oversight requirements before choosing models, vendors, or architectures so compliance and safety are designed in, not bolted on.

How does this connect to inverted products?

Inverted products succeed when harmful outcomes stay rare. Regulated AI is the same shape: human-in-the-loop and risk tiers are product controls that keep failure modes contained.

Related on Inverted PM

Connect with the hosts and guests on the video or on LinkedIn to continue the conversation.

🎧 Find the Inverted Podcast on LinkedIn, Spotify, YouTube, Amazon Music/Audible, iHeartRadio, or invertedpm.com.

Explore more

Browse all Inverted Podcast episodes, read the blog, or meet the hosts.