AI and automation expert on how leaders use AI agents to get ahead | Pascal Bornet

Bornet & Wood (WorkLab) — how leaders use AI agents to get ahead

▶ Watch on YouTube · Microsoft · 40:15

AI agents can become essential team members at any organization if they’re deployed thoughtfully in business workflows. In this episode of Microsoft’s WorkLab podcast, AI and automation expert Pascal Bornet joins host Molly Wood to share practical frameworks for helping business leaders design and implement agentic AI, run high impact pilots, and build the human skills (creativity, critical thinking, and social authenticity) that set Frontier Firms apart.

TL;DR

A 40-minute WorkLab episode on the Microsoft channel, published 30 October 2025. Host Molly Wood interviews Pascal Bornet, who built intelligent-automation practices at McKinsey and EY and co-wrote Agentic Artificial Intelligence. Five claims:

1. Agents act; chatbots suggest. A chatbot plans your weekend and leaves you to book it; “information without action is very limited.” RPA scripts were “the old kind of agents”; LLMs have made them able to understand goals, plan and learn.
2. Four functions define an agent: sense, plan, act, reflect, which he calls the SPAR framework and compares to cooking pasta.
3. Frontier Firms share four traits: they reimagine the business rather than automate it as it is; they redesign work for human-agent collaboration; they build networks of agents, not point solutions; and they gain “compounding intelligence advantages,” which he says become “insurmountable learning gaps over competitors.”
4. “Humics”: genuine creativity, critical thinking and social authenticity are the abilities AI “will never be able to achieve as well as we do,” and the ones to train.
5. Trust is extended in steps. Managers become orchestrators who set goals and boundaries; trust in an agent grows like a parent letting a 16-year-old drive, first only near home.

Key claims

From apps to agents

“We won’t say there’s an app for that, we will say there is an agent for that.” Apps require humans to learn them; agents “will learn how to serve us” and “will use those apps on our behalf.” His client examples: Pets at Home, the UK’s largest pet-care company, reached “99% accuracy in transcribing veterinary consultations” with an ambient scribe, freeing the logistical work that takes “80% of the veterinarian’s time,” and runs agents for fraud detection (the same photo of a damaged parcel submitted for several refunds), insurance checks and staff guidance. Others he is building: invoice-to-pay, recruitment, learning paths, marketing campaigns, churn retention. He cites JPMorgan reducing fraud by 70% and McKinsey cutting client-onboarding time by 90%.

Why projects fail, and the Frontier Firm

Microsoft’s Work Trend Index found more than 80% of leaders plan to use agents to close capacity gaps. Failures he sees: agents built without proper process design, without the right security and guardrails, or confined to one use case. “The companies that succeed are only the ones that put the people in the center of that transformation”: inform people what it means for their role, then educate them. The Frontier Firm’s first trait is “improve before even automating”; its fourth, the compounding advantage, is that each agent interaction generates data that improves the agents, and early adopters also learn to work in hybrid teams, building “competitive moats, that become increasingly difficult to replicate.”

Humics

From his book Irreplaceable: because AI keeps taking skills once thought safe (creativity was the last), humans should be “as different as possible” from it, to avoid overlap. Genuine creativity (original ideas from lived experience, not recombination), critical thinking (judgment, “the kings of asking the right questions,” ethics) and social authenticity (trust-based relationships through empathy). Everyone is born with them, as children building a sandcastle show, and they can be developed without limit. “In a few years from now, AI will be a commodity,” so the differentiator will be combining human and agent capabilities well.

Managing hybrid teams

“Everyone is now becoming a manager in a way.” Humans and agents should have “completely different responsibilities,” and the manager’s work shifts “from task assignment… to goal setting and boundary definition.” Build trust through clear roles, transparency about each agent’s limits, and gradual delegation: a small task first, more after review, “baby steps.” Metrics: for humans, reward trying and failing with the tools and building human skills; for agents, alerts, user feedback, and audits of decision quality; for the team, qualitative measures such as customer and employee experience alongside cost and time.

Change readiness, and how to start

Three competencies: human-ready (humics), change-ready (filtering the important from the trivial in a stream of new models), and AI-ready (choosing what is worth trying). He recommends giving people “about 20% of their working time” to follow developments and experiment, and rewarding failed attempts; classroom training is outdated before it is released. To start: a clear vision from the top tied to business problems, a centre of excellence that may begin as one person with a mandate, and a pilot that is small but visible: “start small, but see big.”

Dynamic-capabilities reading

• digital-transforming/redesigning-internal-structures — hybrid human-agent teams, managers as orchestrators, and a centre of excellence.
• digital-seizing/rapid-prototyping — small, visible pilots and a fifth of working time set aside to try new tools.
• digital-sensing/digital-mindset-crafting — the three readiness competencies and the humics as the skills to cultivate.

How it connects

• ai-agents — sense, plan, act, reflect as a practitioner’s definition, and the move from apps to agents.
• agent-oversight-and-delegation — trust extended in steps, and goals and boundaries in place of task assignment.
• durable-skills — creativity, critical thinking and social authenticity.
• enterprise-ai-adoption — the four Frontier Firm traits and a centre of excellence started by one person.
• Sadun and George — the same show on the Frontier Firm.
• Egeriis (Danske Bank) — disagrees on the moat: a bank CEO sees “a few months” of head start, not insurmountable gaps.

What was actually ingested

The full episode (40:15, length_seconds: 2415) from the human-curated caption track. The episode is from October 2025, a year before ingest. No stills: a podcast.

Linked entities and concepts

• Microsoft — publisher (WorkLab); the Frontier Firm and Work Trend Index are Microsoft’s. Updated in this ingest.
• ai-agents, agent-oversight-and-delegation, durable-skills, enterprise-ai-adoption — updated in this ingest.
• dynamic-capabilities — the three cells tagged above.

Dangling (single-source mention, deferred): Pascal Bornet, Molly Wood (host, never author:), Pets at Home.

Scope and reliability

A consultant and author promoting his book on a vendor’s podcast. The client figures (99% transcription accuracy, 80% of a vet’s time, JPMorgan’s 70% and McKinsey’s 90%) are unsourced in the episode. The claim that AI will “never” match humans on the humics, and that early adopters build insurmountable moats, are positions rather than findings; the second is disputed elsewhere in the wiki. Its value is a practitioner’s vocabulary (SPAR, humics, the readiness competencies) and concrete advice on delegating to agents step by step.