Keynote: The Next Era of AI: Moving from “Can We Build It?” to “Can We Trust It?”
summary
Argues that enterprises face a growing trust gap as capability outpaces confidence, and that pilots do not equal production; recommends continuous end-to-end assurance using agent observability frameworks and a trust score per agent to measure drift. Gives a concrete drift example—125ml becoming 130 then 140—to show small shifts can erode margins, and lists a checklist: define business success, set acceptable risk with stakeholders, build governance, and monitor post‑go‑live.
AI-written summary · about
chapters (42)
- 0:00The Next Era of AI: From Can We Build It to Can We Trust It
- 1:06Guest Intro - Andrew Duncan, Chief Executive Officer at QualityAI
- 2:08Two Eras of AI: Intelligence, Then Trust
- 2:39Trust Is the Foundation for Confidence, and Confidence Enables Scale
- 3:10Struggling With the Second Era and the FOMO It Creates
- 3:41The Trust Gap Is Expanding, Not Contracting
- 4:11The Questions That Reveal the Cracks in "We've Got It"
- 4:41Adoption, Quality and Trust as Three Separate Checks
- 5:42A Successful Pilot Is Not a Solution Ready for Production
- 6:12Risk and Assurance: What Is Acceptable Risk for a Probabilistic System?
- 7:14What Evidence Do You Need to Believe the Risk Is Mitigated?
- 7:45This Is No Longer an IT Issue - The Stakeholder Group Widened
- 8:48Where and When Should a Human Be in the Loop?
- 9:18The Question That Gives Everyone Pause: Who Is Accountable?
- 10:19Does All This Governance Slow Everything Down?
- 11:19The Airline Analogy: Invest in Maintenance Before Adding Flights
- 12:50The Cloud Parallel - and Why AI Isn't There Yet
- 13:52We're Between the First and Second Era Because Confidence Is Missing
- 14:53The Most Common Mistakes Enterprises Make
- 15:23It Is Not a One-Off Gate You Pass Through
- 15:53AI Solutions Drift, Adapt and Make Decisions - Unlike Old Software
- 16:23You Cannot Test an AI Solution in Isolation
- 16:54Continuous Assurance in an End-to-End Context
- 17:25"Trust Me, It's Working" Is No Longer Sufficient Evidence
- 17:55Agent Observability Frameworks and a Trust Score Per Agent
- 19:26Building a Business Case When the System Will Drift
- 19:58The Wine-Pouring Agent: 125ml Becoming 130, Then 140
- 21:29Drift Is Dull in Business Applications and Dramatically Expensive
- 23:30Loop Engineering and Quality Embedded Across the Lifecycle
- 24:31Counterintuitively, AI Creates More Demand for Assurance, Not Less
- 25:31Why AI Solutions Don't Fit SDLC-Style Thinking
- 26:32Testing Has to Be Designed In From the Very Beginning
- 27:33AI Makes Assurance a Strategic Business Capability
- 28:03AI Assurance Belongs in Enterprise Governance, Not Just Technology
- 28:34The Case for a Chief Quality Officer in Every Organisation
- 30:07Reimagining the Business Process, Not Just Inserting AI
- 31:40The Closing Checklist for Taking Pilots to Production
- 32:11Define Success, Quality Expectations and Acceptable Risk First
- 33:14Test the Whole Ecosystem, Not Just the Model
- 33:45Confidence Before Go-Live Through Evidence, Not Assumptions
- 34:15Bringing Employees With You and Naming the Human Triggers
- 34:46Measure Success by Reliable AI Delivering Business Outcomes
related talks
10mKeynote: The Last Mile Problem: Why AI W...
38mKeynote: Thriving in the Age of AI
29mKeynote: Building Trust in a Data Driven Enterprise
19marchive – Shane Murray
29mBuilding Trust in AI Agents
15mSuccessfully Scaling AI and the Role of Trust
44mThe Future of Engineering Teams in an AI-Native World
33mUnlocking Agentic AI: [...]