Insurtech Insights London : How to Leverage a Near-Infinite Synthetic Workforce
Building Agentic AI Teams: How to Scale with Synthetic Workers in 2025
Lessons from the AI-Risk Accelerator and live insurance case studies
The Biggest Risk? Falling Behind on Agentic AI
The most urgent risk facing modern businesses — especially in insurance — isn’t cyber threats or bad AI investments. According to Simon Torrance (Founder of AI-Risk), it’s being outcompeted by rivals who deploy agentic AI better and faster than you.
In this fast-moving talk, Simon and Frank P share real-world frameworks, definitions, and case studies on how companies are already scaling agentic teams — replacing costly consultants and overworked staff with coordinated synthetic AI workers.
What Is Agentic AI — Really?
At its core, agentic AI is about autonomous digital workers that can:
Perceive a problem
Reason and plan
Collaborate with humans and each other
Take action (via tools or systems)
Learn and improve through memory
These are not just smarter chatbots. They’re team-based AI agents that act like real employees, complete with roles, objectives, and personalities.
Simon outlines five core elements of agentic AI:
LLM Brain: One or more large language models at the centre.
Persona: A job description, behaviour style, and goal.
Planning Capability: Can break down tasks and sequence them.
Action Layer: Connects to external systems and tools.
Memory: Learns over time — faster and more consistently than humans.
Case Study 1: Replacing Consultants with AI
One major enterprise ran a standard procurement risk assessment. Normally this would involve 6 weeks of work from a team of consultants, costing around £500,000.
Using agentic AI:
A team of 30 agents were created in 2 weeks
They delivered 50 scenarios (vs 1 from the humans)
Task completed in 1 hour
Results were deeper, faster, and cheaper — and repeatable
These weren’t assistants helping humans. These were full replacements for expensive, knowledge-heavy consulting roles.
Case Study 2: An AI-Run Claims Department
Frank P shared a real-world deployment: building a virtual insurance claims team made entirely of AI agents.
Key roles:
Customer Care Agent (Claudine)
Claims Experts (Raul, Victor)
Team Facilitator (Ellen) – coordinating the workflow like a human manager
Tools Used:
Slack for communication
Swarm-based algorithms for decision-making
Low-powered LLMs (not cutting-edge models)
No strict workflows — just collaboration
Results:
No new human hires
Significant cost reduction
3x productivity in claims categorisation
Higher NPS than previous setups
Even staff morale improved (one human staffer even requested to be “cloned”!)
This team didn’t replace people because of cost — it replaced them because the company couldn’t hire fast enough, and AI proved more reliable.
From Tools to Teammates
Many companies still use AI as a “tool” — to write content or summarise documents. But the future lies in agentic teams, not just productivity hacks.
Frank compares it to assembling a team of superheroes:
“You don’t just give one person a power suit — you build a team.”
The key? Team dynamics.
AI agents need structure, a facilitator, and alignment with human goals to work well. Just like a real team.
How It Works (Technically)
This agentic framework isn’t built on the “most powerful” LLMs. In fact, it uses simpler models — because the magic lies in collaboration, not model size.
Key principles
Run agents in Slack (or Teams, WhatsApp, etc)
Use swarm logic to simulate team decision-making
Encourage human-in-the-loop collaboration
Let agents learn through doing, not just fine-tuning
This is artificial life in action — and it’s live in production.
Why Strategy Must Come First
Simon ends with a crucial warning:
“Do not start with technical experiments. Start with strategy.”
AI teams must align to business goals. What would it mean to triple your operational capacity in 5 years? What if your competitor does it before you?
The Agentic AI Accelerator was created to help executives get this alignment, fast — combining education, experimentation, and operating model transformation.
Final Thought: Who Manages AI Teams?
The biggest shift might not be technology — but leadership.
“If AI agents become your colleagues, who manages them?”
Forward-thinking companies are already training managers to lead blended teams of humans and AI. They’re investing in people who understand collaboration, oversight, and optimisation in this new world.
The question isn’t whether you’ll use agentic AI.
It’s whether you’ll lead it — or be left behind by those who do.
Want to see the tech in action?
Follow Simon Torrance on LinkedIn or visit ai-risk.co. A live demo and event series will showcase how agentic AI teams are already transforming insurance, banking, and operations in 2025.