Agentic AI Isn't a Tool. It's Your New Workforce.
This is a recording of the keynote presentation delivered by Simon Torrance, CEO of AI Risk, at the ‘Rise of the Agents’ event in London on 25th November 2025, co-hosted by AI Risk and Instech. 350 executives from the insurance industry attended. It was the first major physical event dedicated to this important topic for the insurance industry, worldwide. The principles apply to companies in all knowledge-intensive sectors.
For more on this topic, please see our new Agentic AI Enterprise Blueprint and Agentic AI Accelerator.
Below is a transcript of the presentation. It’s how I explain Agentic AI to Leadership Teams and Boards.
Agentic AI Isn't a Tool. It's Your New Workforce.
Why the next 24 months will separate winners from laggards
SLIDE 1: Introduction
Good afternoon everyone.
I'm Simon Torrance, CEO of AI Risk. We're a specialist advisory firm helping insurance companies build competitive advantage with agentic AI.
Over the next 25 minutes, I want to share what we're seeing at the frontier—and more importantly, what it means for your business over the next three years.
SLIDE 2: The 2028 Question
Let me start with a question I'd like you to hold in your mind throughout this talk.
In 2028—just three years from now, when we come back to this event—will your company or business unit have unbeatable agentic economics and superior commercial performance?
Or will you be explaining to your board why you didn't move quickly and effectively when you saw the signals?
This isn't hypothetical. The signals are here. The question is what you do about them.
SLIDE 3: Agentic AI = A New Type of Workforce
This is Jensen Huang, CEO of Nvidia—arguably the most important technology company in the world right now.
Last October he said: "By 2030, we will be a 50,000-employee company with 100 million AI assistants. AI will recruit other AIs to solve problems. We'll have a very large workforce—some digital and some biological."
I want you to sit with that ratio for a moment. 50,000 humans. 100 million AI assistants. That's 2,000 AI workers for every human employee.
Now, Nvidia is an extreme case. But the direction of travel is clear. Agentic AI isn't a tool. It's a new type of workforce.
SLIDE 4: AI Maturity Model
This is how we think about AI maturity. It's a framework we use with all our clients. (Details here).
On the left—Level 0 and Level 1—you have RPA, chatbots, assistants. This is table stakes. It's what most companies have today. It helps humans work a bit faster, but it doesn't fundamentally change anything.
The dotted line is the critical threshold.
On the right—Levels 2 through 5—you have true agentic AI. Specialist agents that complete tasks autonomously. Multi-agent workflows where agents hand off to each other. Agentic teams that coordinate dynamically. And eventually, ecosystem agents that work across company boundaries.
This is your new digital workforce—scalable, on demand.
Most companies are stuck at Level 1, running pilots, wondering why they're not seeing transformational results. The answer is simple: they haven't crossed the threshold into agentic AI.
SLIDE 5: Gen AI vs Agentic AI
Let me make this distinction concrete.
On the left: Gen AI. Copilot, ChatGPT, the tools most of you are using today. You ask questions. You get answers. Brilliant.
But then what happens? You do the work. You take that answer, and you still have to act on it, implement it, complete the task yourself.
On the right: Agentic AI. You assign tasks. The work gets completed. Not advice about the work—the actual work, done.
Notice the loop on the right—rapid learning. Every task makes the system smarter. That's your appreciating asset, which I'll come back to.
The shift from left to right is the shift from AI as a tool to AI as a workforce. And that changes everything about your economics.
SLIDE 6: What's Fundamentally New & Different
This slide is the heart of what I want you to take away today. Four dimensions of difference.
Results: Previous technology helped you do things faster. Agentic AI lets you do three times more for less. Not optimisation—transformation.
Pattern: We're used to Tool-to-Human-to-Output. The human is always the bottleneck. Agentic AI is Capital-directly-to-Output. You're buying productive capacity, not hoping for productivity.
Value: Previous technology created gains for everyone. When processors got faster, your competitors got faster too. No advantage. Agentic AI creates differentiating intelligence—your underwriting expertise, your claims patterns, your customer nuances—captured permanently.
Balance Sheet: And here's the strategic choice. Previous tech is a depreciating cost. You spend money, get temporary benefit, it fades. Your best people leave, their expertise walks out the door. Agentic AI—when built proprietary—is an appreciating strategic asset. It compounds. It never leaves.
Notice the asterisk. This only applies when you build proprietary. Rent from vendors, and you're back to "everyone gains." That distinction matters enormously.
SLIDE 7: Mid-Sized Company Example
Let me make this tangible. Imagine a mid-sized insurer—or a division of a larger company—with 250 people.
Customer service: 90. Sales: 40. Claims: 35. IT: 25. Other functions: 60.
Total headcount cost: about £16 million annually.
But here's the uncomfortable truth. You're paying for 2,000 hours per person, but only getting about 50% productive time. The rest is meetings, admin, data entry, moving information between systems. I call it the Admin Trap.
So your actual productive capacity is about 250,000 hours. That's what £16 million buys you today.
SLIDE 8: ‘FTEs’ vs ‘PCUs’
Now you have two options for growth.
Option one: Hire more humans. Buy time and hope for productivity. Pay for 2,000 hours, get 50% productive. Cost: £65 per productive hour. Available during working hours only. And critically—expertise is lost with turnover. Every time someone leaves, you're rebuilding.
This gives you linear scalability. To do twice as much, hire twice as many people.
Option two: Deploy Productive Capacity Units—AI agents. Pay for output directly. Get 100% productive time. Cost: around £24 per productive hour in year one, dropping to £12 by year three. Available 24/7/365. And learning is captured permanently.
This gives you exponential scalability.
The footnote is important: proprietary agents create competitive moats. That's the key. It's not just cheaper—it's strategically different.
SLIDE 9: Two Growth Flywheels
Look at the cycle times.
Traditional flywheel on the left: Identify hiring need—8 to 12 weeks. Recruit—variable. Onboard and train—3 to 6 months. Then deploy. Then generate profit. Total cycle: 18 to 24 months before you see returns.
Agentic flywheel on the right: Design solution—fast, it's new and different. Deploy AI capacity—1 to 2 weeks. Scale and optimise—non-trivial, I won't pretend it's easy. Expand capability. Generate profit. Total cycle: 1 to 3 months.
And notice what this unlocks: underserved segments. Markets that weren't economically viable to serve with human labour become viable with AI agents.
The speed difference isn't incremental. It's structural.
SLIDE 10: Operational Capacity Transformed
Back to our 250-person company. Here's what 2028 could look like.
Human headcount drops from 250 to 205. But look at where it drops and where it grows.
Customer service: 90 down to 45 humans, but now supported by 135 AI agents. Sales actually grows—40 to 55 humans, plus 165 agents. Strategic roles expand. You add a new function: AI Agent Managers. These become your most valuable experts.
The outcomes: Salaries down 25%. Total capacity up 3x. Available 24/7/365.
This isn't about replacing people with machines. It's about restructuring work so humans do what humans do best—strategy, judgment, relationships—while agents handle volume and routine complexity.
SLIDE 11: New Value Frontiers
And this opens entirely new value frontiers.
When you have near-infinite operational capacity, things that weren't economically viable become possible. Instant claims settlement. Closing protection gaps. Prevention ecosystems. Real-time dynamic pricing.
This isn't just doing today's business better. It's accessing new sources of value creation that human-only workforces could never reach.
SLIDE 12: European Protection Gaps
Just one example: European protection gaps. £350 billion of unmet demand across cyber, healthcare, retirement, natural catastrophe.
This isn't theoretical market size. It's real demand that's currently unserved because it's too expensive to underwrite and service with human-only operations.
Near-infinite operational capacity changes that equation entirely.
SLIDE 13: The Future's Already Here
Now, you might be thinking: "This sounds compelling, but is anyone actually doing it?"
Yes. This is a real deployment—a policy and claims contact centre running with agentic teams.
On the left: Human FTEs. Franck, Pierre, Charlene, Arnaud. Real people, doing valuable work.
On the right: An agentic team. AI representatives, some of them digital twins of the best human performers. AI experts in actuarial, CRM, process. A meeting facilitator agent. All coordinated by what we call a swarm algorithm.
The results: 70% reduction in FTE hiring needs. 25% higher NPS. 3x productivity gain. 30% improvement in profitability. And—this matters—high employee satisfaction. The humans are doing more interesting work.
This isn't a pilot. This is Level 4 agentic AI running in production. The future is already here—it's just not evenly distributed yet. (Full case study here).
SLIDE 14: Tale of Two Approaches
Let me paint you a picture of two firms over the next three years.
Firm A: Cautious and tactical. Traditional markets only. 9-to-5 service. Focused on efficiency and stuck in pilot purgatory. NPS steady. Growth still tied to headcount. Level 1 to 3 maturity. By month 36, they've achieved 1.5x capacity.
Firm B: Bold and strategic. Traditional plus new markets. 24/7 service. Focused on growth with rapid prototyping. Much higher NPS. Software economics. Level 3 to 5 maturity. By month 36, they've achieved 3x capacity.
Look at the divergence in that chart. By 36 months, Firm B is serving twice as many customers as Firm A—from the same starting point.
This is the critical insight: a 24-month implementation lead creates permanent structural advantage.
The gap doesn't close. It widens. Because while Firm A is trying to catch up, Firm B's systems are learning, improving, compounding.
This is why the decision you make in the next 6 to 12 months matters so much.
SLIDE 15: 5 Mistakes to Avoid
So what separates Firm A from Firm B? Let me give you five mistakes we see companies making.
One: Letting IT drive agentic AI. This isn't an IT project. It requires CEO leadership and cross-functional strategy—COO, CMO, CHRO, CIO all aligned.
Two: Automating legacy processes. If you just automate what you do today, you get incremental gains. Reinventing workflows around agentic capabilities is the key to EBIT impact.
Three: Relying on same-old vendors. Your traditional systems integrators and consultants are great at Level 0 and 1. Level 3 to 5 requires specialist expertise they don't have.
Four: Outsourcing core competitive capabilities to hyperscalers. This is where we see companies about to repeat the Guidewire mistake. "Build on Azure" sounds like you own it. You don't. Understand the strategic trade-offs of build versus buy. Beware vendor lock-in.
Five: Underestimating change management. Building complex systems plus organisational transformation plus effective governance is genuinely non-trivial. Don't pretend otherwise.
These five mistakes are the difference between Firm A and Firm B. Between incremental efficiency and transformational advantage.
SLIDE 16: The New Formula
Let me give you a formula to take away.
The new formula for success: Agentic operational capacity—reimagined workflows delivered by a hybrid human-AI workforce—divided by total cost of productive capacity. Multiplied by expanded strategic options—new business scope enabled by near-infinite capacity. Equals superior commercial performance—new unit economics and competitive advantage that compounds over time.
Contrast that with predominant AI thinking today: AI tools times isolated use cases, divided by operational myopia—automating legacy workflows. Multiplied by traditional strategic options. Equals incremental efficiency.
Same inputs. Radically different outcomes.
The question is: which formula are you running?
SLIDE 17: The Agentic AI Accelerator
So what do you do next? We've designed a process specifically for this.
Step one—Strategy Hypothesis. One month. We work with your leadership team to build a clear hypothesis about where agentic AI creates value in your business.
Step two—Rapid Prototypes. Three months. We build and deploy 2-3 working agentic solutions that prove the value and build internal capability.
Step three—Controlled Scaling. Ongoing. Systematic expansion across your organisation with proper governance and change management.
Underpinning all of this—the Agentic Transformation Operating System. The infrastructure that allows you to deploy wave after wave of agentic capability.
We build with you. We transfer capability to you. You own everything. No perpetual dependency. Details here.
SLIDE 18: What Will You Achieve
So let me bring it back to where we started.
In 2028, will you have unbeatable agentic economics? Or will you be explaining why you didn't move?
The signals are clear. The economics are proven. The technology works.
The only question is whether you'll be Firm A or Firm B. What’s your ambition?
