Agentic AI - What is it, how it works, why is it important, best case studies?
Why Agentic AI Is the Buzzword of 2025
Agentic AI has gone from niche research to front-page tech in a matter of months. With OpenAI releasing customer service agents, and giants like ServiceNow demoing agent studios with thousands of synthetic workers, the conversation around AI has shifted — again.
But what is agentic AI? Why does it matter? And what are the risks and opportunities for businesses — especially in financial services?
Simon Torrance, founder of AI-Risk and seasoned strategist in tech and insurance, joined the London Fintech Podcast to unpack these very questions.
What Exactly Is Agentic AI?
In simple terms, agentic AI refers to autonomous AI agents — synthetic digital workers that don’t just assist humans but can think, plan, act, and even collaborate with other agents and people.
Unlike chatbots or co-pilots, agents have:
Goals and roles (like a real team member)
Reasoning and planning ability
Memory to recall past context
Tools to take actions inside and outside a system
Personalities for human-like interaction
These agents don’t just follow commands. They operate semi-independently to solve problems, complete workflows, and optimize systems — at speed and scale far beyond human teams.
Synthetic Teams: Real-World Use Cases
Simon shared two eye-opening case studies:
1. Ministry of Defence (UK): Cost Risk Assessments
Traditionally, the MOD would hire six consultants for six weeks to estimate costs and risks for a large capital project. Price tag: ~£500,000.
Instead, using 30 agents trained for the task, the same project was completed in 1 hour, producing 50 scenarios and a more thorough output.
2. Insurance Company: Fully Agentic Claims Department
With no access to affordable insurance talent, a startup created a full claims ops department using AI agents:
Agents managed customer queries
Collaborated in Slack channels
Reported to an AI manager agent
Delivered better net promoter scores and higher profits than traditional setups
These agents weren’t just automating forms — they were solving problems, making decisions, and improving profitability.
Why This Is a Strategic Shift — Not Just Automation
For Simon, the biggest risk in agentic AI isn’t reputational damage or hallucination. It’s strategic risk:
“The risk of being outcompeted by those who adopt and deploy AI better than you.”
Most financial institutions pride themselves on being “fast followers.” But with agentic AI, fast followers lose. The learning curve is exponential — early movers pull ahead rapidly.
As Simon put it, if you’re not building with agentic AI now, you’ll be competing against those who have 100M synthetic workers at their disposal by 2026.
Should You Build or Buy?
Here’s Simon’s framework for enterprise leaders:
Understand the art of the possible.
Most leaders don’t know what’s technically possible today.
Tie it to strategy.
Don’t chase shiny demos. Link agents to real business goals.
Define your tech stack.
Data, tools, intelligence, governance, user interface.
Avoid outsourcing your brain.
Be wary of handing over all agent development to hyperscalers.
Agents = collective intelligence. That’s your competitive edge.
What Roles Will Humans Play?
Simon sees a mixed future:
Some agents augment human teams, unlocking productivity.
Others will replace repetitive knowledge work — especially early-career roles like analysts, underwriters, and claims handlers.
Humans will need to collaborate with agents, not compete with them.
This echoes NVIDIA CEO Jensen Huang, who said they aim to scale from 50,000 employees to 100 million AI assistants.
Governance Is Not Optional
With great power comes great risk.
Simon warns that governance, ethical AI, and responsible design must be embedded at every layer — especially in financial services.
Think of it like GDPR: compliance must be baked in, not bolted on.
2025: The Year of Awareness
Simon’s prediction?
“2025 will be the year leaders realize how fast agentic AI is moving.”
Some will act. Most won’t.
Those who educate themselves, pilot real use cases, and align their strategy now will dominate their industries. Everyone else will fall behind — fast.
Want to learn more?
Simon and the AI-Risk team are hosting a live event in London on March 27th focused on Agentic AI. Follow Simon on LinkedIn or visit ai-risk.co to stay updated.
