Ovetta Sampson: AI done right — building responsible innovation

Ovetta Sampson is the founder of Right AI, a consultancy that helps enterprise organisations identify and reduce what she calls Human Engagement Risks in AI deployment. Before founding Right AI in 2025, she spent over a decade at Google — most recently as Director of AI and Compute Enablement — and has held senior roles at Capital One and IDEO. Business Insider named her one of the top 15 people in enterprise AI in 2023.

The talk was published on YouTube in late April 2026 and delivered as a keynote. It draws on Sampson’s experience watching AI systems fail not because of technical shortcomings, but because of decisions made early in the product design process.

Who it is for

The talk is most relevant to product managers and product leaders who are actively building AI-powered features and need a structured way to think about where products can cause harm. It is not a technical presentation — there is no code, model architecture, or benchmarking discussion — and it is not a compliance checklist. It operates at the level of product strategy and design methodology, covering the design decisions that determine whether an AI feature holds up when it reaches real users at scale.

Key takeaways

1. AI risk is primarily a human engagement problem. Sampson argues that the most significant risks in AI products are not model accuracy failures but failures in how humans interact with AI outputs — over-reliance, misunderstanding, or exclusion of certain user groups. Her Human Engagement Risks framework is a structured tool for identifying these risks during product design rather than after launch.

2. Business model fragility is a design risk. Sampson draws an analogy to MoviePass to illustrate how products built on AI interactions that are not economically sustainable at actual usage levels produce a category of design risk that does not appear in standard risk assessment. For PMs, this translates to a question worth asking early: does the product’s core AI interaction pattern work at the pricing and usage levels the business actually needs?

3. Responsible design requires systems thinking. Her critique of fast-moving product culture is specific: moving without mapping second-order effects produces brittle products that generate trust problems later. She frames this as a product quality argument rather than an ethical one — responsible design produces products that hold up over time.

4. PMs need to actively design for AI transparency. Sampson’s concrete point here is that users need to understand when they are interacting with AI and what the AI is doing. She presents this as an increasing product design constraint rather than an optional feature, given how user expectations about AI transparency are shifting.

5. Historical context changes prioritisation. The talk draws comparisons to earlier periods of rapid technological change to argue that AI’s current pace is not historically unique, but that the window for building good design practices is narrower than it appears from inside a fast-moving company.

Worth watching if

This talk is most useful at the stage when a product team is deciding how to introduce an AI feature to users — not during early technical prototyping. Sampson’s Human Engagement Risks framework is practically usable in a PM review, a pre-launch checklist, or a team conversation about what responsible AI design actually means in concrete product decisions.