The financial services sector is famously risk-averse. It has to be. But right now, the industry is facing a disruption that feels less like a ripple and more like a tidal wave. Generative AI has arrived, and for many in the insurance and financial sectors, the pressure to adopt is overwhelming.
Everyone is talking about it. Every board is asking for it. But between the shiny demos and the ambitious promises, there is a stark reality: getting AI in financial services to actually work in a regulated, complex environment is hard, and often requires specialist AI consulting services to move beyond experimentation and into production.
We recently sat down with Philip Lobatto from Microsoft, Andrew Charge from First Central, and our own Chief Data & AI Officer, Jodie Ashwood, to talk about what is actually happening on the ground. No fluff, just the honest truth about where the industry stands, the shift from Copilots to AI Agents, and why so many businesses are struggling to move past the pilot phase.
The Reality of AI in Financial Services and Insurance
What is the current state of AI adoption in the sector?
It is a tale of two technologies colliding.
Philip Lobatto, Executive Advisor for Insurance at Microsoft, points out a crucial distinction. The insurance industry has used “classical AI” (machine learning, deep reasoning, predictive modelling) for decades. It is the backbone of risk calculation and pricing, especially in dynamic markets like the UK.
But since late 2022, Generative AI has entered the room. This is different. It brings natural language capabilities that open up entirely new ways to interact with data. Philip highlights that the “sweet spot” is where these two worlds intersect, combining the mathematical accuracy of machine learning with the conversational power of GenAI.
However, despite the excitement, success is not guaranteed. Philip shared a sobering statistic from an MIT study: 95% of GenAI pilots fail. They don’t fail because the technology is broken; they fail because companies try to avoid friction. They attempt to layer new tech over old cultures without making the hard decisions required to change how they operate.
Why is there such a high failure rate for AI pilots?
Culture eats strategy (and technology) for breakfast.
This is especially true in insurance, a business built on precision. “There is a culture of accuracy,” Philip explains. “Generative AI is a probabilistic system, which is something different than an exact outcome.”
Andrew Charge, Digital Workspace Director at First Central, agrees that the hurdle is often mindset rather than software. He compares the current AI race to the moon landings, everyone is rushing to get there, but few are discussing the training required to survive the journey.
“It’s a mindset shift,” Andrew says. “It’s no longer technology pushing out updates… all of us need to get this journey.” He warns that simply dropping tools like Microsoft 365 Copilot onto a workforce without preparation is a recipe for disaster.
“If you can ride a bike, and I give you a Ferrari, it looks lovely. Sounds great. But you don’t know how to drive a car… It’s going to be great for your Instagram photos, but it’s not actually going to achieve anything.”
Andrew Charge, Digital Workspace Director at First Central
Agents vs. Copilots: The Next Evolution
What is the difference between a Copilot and an AI Agent?
We are moving from chat to action
We are currently in what Philip calls ‘phase one’, the era of the Copilot. This is the human with an assistant. You ask a question, you get an answer. It is useful, but it requires constant human prompting.
Phase two is the era of Agents. This is where the technology shifts from passive to active. Instead of just answering questions, an army of AI agents will perform tasks for you, coached and overseen by humans.
Philip described an “orchestrator” model used by a health insurer in Asia. Imagine a conductor leading an orchestra of specialised agents, one medical agent, one compliance agent, one claims agent. They work together to process information, with the human stepping in only for oversight and final decisions. This agentic workflow is where the real efficiency gains lie, particularly when organisations adopt agentic AI models designed to operate across complex, regulated processes.
Where are the best use cases for this technology?
Claims, customer service, and underwriting are the big three.
For First Central, the strategy involves a pragmatic split between front-office and back-office functions. Andrew notes that while the front office focuses on customer value, the back office offers huge potential for efficiency that is often overlooked.
Philip shared several examples of this in action:
- Markerstudy uses transcription summarisation to save 3-4 minutes per call across 800,000 calls a year. That is massive aggregate saving.
- AXA UK used GenAI to review 70,000 multi-page documents in a week (a task that would have taken humans over a year) to identify specific building risks (RAAC).
- Manulife combined machine learning insights with GenAI to send personalised, relevant letters to customers with a high propensity to buy, revitalising their outbound engagement.
How did First Central approach their AI rollout?
They refused to buy the hype.
Andrew shared a refreshing perspective on resisting pressure. When the buzz around AI peaked, he faced huge demand to buy Copilot licenses for everyone immediately. He said no.
“I didn’t think the business was ready,” Andrew admits. “I didn’t think we had the training in place… I think if I just bought them, they’d have sat there on my budget for the last year.”
Instead, First Central focused on creating space for learning. They recognised that their staff were already suffering from “meeting burnout” and simply adding a new tool wouldn’t help. By prioritising digital literacy and cultural readiness first, they ensured that when they did deploy the tech, it would actually be used effectively.
How can organisations scale AI successfully?
You need a structured, scalable approach.
Jodie Ashwood, our Chief Data & AI Officer, argues that organisations get stuck because they treat AI as a series of disconnected experiments. To scale, you need a structured ecosystem, an AI Factory.
This approach focuses on three core components:
- The Journey to Value: Clearly mapping AI strategy initiatives to business KPIs so you aren’t just doing “technology for technology’s sake.”
- Trusted Architecture: Ensuring your data platform, security, and governance are robust. As Jodie puts it, “Rubbish in, rubbish out.” You need to bring legal and compliance teams along for the ride early.
- The Scale Engine: Creating a standardised environment for rapid prototyping. This allows you to test ideas quickly, fail fast if necessary, and move successful pilots into production using custom AI solutions built for regulated environments, with reliable guardrails in place.
The Future of Intelligent Operations
What is the one thing leaders should do right now?
Stop sprinkling AI and start strategising.
Philip’s advice is stark: “If you’re going to sprinkle AI across the company in every division, you will not move the needle.”
Success requires bold decisions. It requires leadership to identify specific, high-value areas where AI can transform the process, rather than just adding a slightly smarter spellchecker to everyone’s laptop.
For Andrew, it comes back to the human element. The technology is evolving daily, and the static learning models of the past, where you learned a system once every three years, are dead. Continuous learning is the new standard.
Getting It Right
The potential for AI in financial services is undeniable. From agentic workflows that handle claims in minutes to hyper-personalised underwriting, the future looks efficient and intelligent.
But as our panellists made clear, you cannot buy digital transformation off the shelf. It takes more than licenses. It takes a willingness to tackle the unsexy work: fixing your data, training your people, and reshaping your culture to embrace friction rather than avoid it.
At Transparity, we believe in straightforward advice and steady hands. If you are ready to move beyond the hype and build an AI strategy that actually delivers, we are here to help you navigate the journey.
