0:00–1:02 — Introduction

1:02–3:00 — Geoffrey Hinton’s Prediction 

  • Discusses Geoffrey Hinton’s famous claim that radiologists were “like Wile E. Coyote” already over the cliff — and that within 5 years, deep learning would outperform them.

  • Cameron stresses: Hinton is brilliant — but this prediction was mostly wrong.

3:00–4:05 — Why Hinton Was a Little Bit Right, but Mostly Wrong

  • Yes, AI may already outperform humans in image interpretation.

  • But radiology is far more than image analysis:

    • Preparing patients for stressful procedures

    • Being present during life-changing moments

    • Adjusting scans based on real-time observations

    • Using professional judgment

  • AI can handle ~30% of what a radiologist does — not 100%.

4:05–5:55 — Jevons Paradox & Radiology

  • Cameron introduces Jevons Paradox (1865):
    When technology makes something more efficient, usage often increases — not decreases.

  • Example: Watt’s steam engine was more efficient, yet coal consumption skyrocketed because cheaper, easier access increased demand.

  • Radiology mirrors this: easier, more accurate scans → more scans ordered → rising demand for radiologists.

5:55–6:18 — Where Automation Is Heading

Cameron introduces five key forces shaping automation — excluding AI because it’s already a given. These are the structural enablers that determine whether AI and automation succeed.

The Five Forces Shaping the Future of Automation

1) 6:18–8:02 — Data

  • Many organizations still can’t:

    • Access all of their data

    • Combine it

    • Interrogate it

    • Analyze it

  • Older organizations = worse data maturity.

  • Modern approaches like data fabric are changing this, allowing read/write access without forcing everything into a data lake.

  • Data is the gold of automation and AI — without mastering it, nothing else scales.

2) 8:02–10:12 — Process Design

Cameron’s area of passion.

  • Most processes were never truly designed — or were designed decades ago.

  • Processes evolve reactively:

    • “New regulation? Add a spreadsheet.”

    • “New product? Add a workaround.”

  • If you redesigned core processes today, they would look nothing like the current reality.

  • Without understanding the real process, you:

    • Miss opportunities for automation

    • Misapply AI

    • Fail to orchestrate workflows effectively

  • There is massive untapped potential hidden in poorly understood processes.

3) 10:12–11:26 — Technical Design

  • Explosion of tech providers: every moment, new software and AI companies appear.

  • Many tools overlap or attempt to solve the same problems.

  • IT often defaults to: “We’ll use Microsoft for that.”

    • Convenient, but not always strategic.

  • Organizations need rigorous thinking about when to use what.

4) 11:26–12:29 — Architecture

  • After choosing tools, the real challenge: making them all work together.

  • Cameron gives the example of a bank with an absurdly complex architecture diagram just to manage one process (changing a customer address).

  • The future requires:

    • Simplified architecture

    • The right tools for the right jobs

    • AI-supported management and orchestration

  • Complexity is the enemy of scalability.

5) 12:29–14:04 — Build vs Buy: The Return of Application Development

  • Cost of SaaS platforms (e.g., Salesforce) is skyrocketing.

  • Pricing is increasingly based on users + consumption, making total cost unpredictable.

  • Meanwhile, AI is drastically reducing the cost of developing custom software.

  • We will reach a tipping point where:

    • Building becomes cheaper than buying.

    • Low-code + AI-assisted development accelerates custom solutions.

  • For unique processes (e.g., bank bereavement journeys), building will make more sense.

14:04–15:48 — Wrap-Up

  • All five forces — data, process design, technical design, architecture, build-vs-buy — sit on top of AI.

  • To succeed with automation, organizations must:

    • Master their data

    • Truly understand their processes

    • Make smart technology decisions

    • Build sustainable architecture

    • Re-evaluate when to build vs when to buy

  • These areas will determine whether automation programs thrive or stall.