Intro

Let’s be honest – insurance is one of the industries where many businesses still operate on decades-old infrastructure. And while it isn’t exactly known for moving fast, that’s changing.
AI is no longer a buzzword. It’s becoming a real competitive advantage. And for insurers willing to lean in, it’s opening up possibilities we’ve never had before – faster claims, smarter underwriting, better customer experiences, and yes, lower costs.
But here’s the catch. This isn’t about slapping AI on top of old systems and calling it innovation. It’s about rethinking how things work; connecting the tech to the heart of your operations, and making it work with your people, not instead of them.
The insurers that get this right won’t just keep up. They’ll lead. So – where do you start?
Let’s break it down.

Integrating AI with existing IT systems

When it comes to bringing AI into the insurance industry, the first question isn’t “What model should we use?”, or “Where do we begin?”. Here are a few tips to get you started

Create a service/customer journey blueprint to map technologies in use

There’s a lot of excitement around AI and for good reason. But in the rush to implement it, many businesses are skipping a critical first step – understanding the systems they already have in place.
A smart place to begin is breaking down the service or customer journey. This means looking closely at how people actually use your product or service. From the first interaction to what happens after it’s over. When you map this out as a service blueprint, you’re not just tracking the customer’s path; you’re also seeing the technology and processes behind each step. That’s where the opportunities for AI start to show.

Let’s take a claims journey.

A customer has an accident, files a claim, tracks its progress, and eventually gets a payout. On the surface, that looks like four steps. But under the hood, there’s policy data being pulled, risk assessments being triggered, fraud checks happening in parallel, and communications going out across multiple channels. Mapping this out reveals which technologies are already doing the heavy lifting – and where friction still exists.
Right now, many companies are bolting on AI without asking if it’s actually helping the user. That’s a mistake. If the goal is a better customer experience, then AI has to be linked to a clear need. The blueprint helps you see that need.
Creating one doesn’t just involve the tech team. It requires collaboration across departments; marketing, operations, support, anyone involved in shaping the service. It starts with understanding your customers, not just in demographic terms, but in how they actually use what you offer.

If your company hasn’t started with AI yet, don’t jump straight to tools or vendors.

Start by creating detailed customer personas. Then map out their journey: before, during, and after they use your service. Identify every interaction, and how it’s handled today. This brings different teams into alignment and gives you a shared view of where improvements are possible (and where AI could genuinely add value).
Internally, the same principle applies. Take a close look at your processes. If you don’t have anything clearly defined, that’s your starting point. Once you map them, you’ll be able to see which parts are ready for improvement and which might benefit from AI.

Decide on the goals for AI – operational, customer-facing, or other

Before rolling out AI, ask the obvious, but often skipped, question: “What exactly are we trying to improve?”
Internally, AI can streamline underwriting, flag anomalies in claims, or help teams do more with less. Externally, it can speed up quoting or offer round-the-clock support. But you can’t aim at covering all of these areas, all at once. Start by choosing a focus, operational or customer-facing, and build from there.
Also, remember that automating too much of the customer experience in a high-trust industry like insurance could backfire. As Marina Belezina, Aviva’s Group Director of Strategy and Innovation, puts it, just because AI can do something doesn’t mean it should. Even though they incorporate AI, they try to stay clear-eyed about where it adds value, and where humans can do the job better.
Aviva uses a hybrid model: AI handles initial analysis of documents and data (even complex ones like medical records), but real experts make the final call. Especially on sensitive or complicated claims.

Belezina lays it out across three horizons:

• Enhancement: using AI to make existing human-driven processes more efficient through internal tools.
• Transformation: automating parts of key workflows to free up human bandwidth.
• Reimagination: long-term goal of building automated ops and creating smarter customer propositions.

What ties it all together are clear goals, with specific KPIs. So, before you implement any form of AI, define what success looks like – hours saved, claims settled faster, fewer escalations, etc.

Run an assessment of your IT systems

Take a hard look at your current data management and systems, i.e., how data flows, where it’s stored, how clean it is, and whether your infrastructure can actually support AI at scale.
This is where most insurers hit friction. Legacy systems, siloed databases, and inconsistent formats aren’t exactly AI-ready, but they’re fixable.
I believe a good example comes from Allianz. Instead of tearing down decades of tech, they built the Allianz Data Platform. It acts as a central hub that connects data from across their global systems. Underwriting, claims, risk management and data on other areas all feed into one place. AI then pulls information from here instead of working in isolation. It uses real operational data, in real time, with consistent structure.
This foundation makes it possible to experiment and scale with confidence. Allianz started with AI chatbots to reduce frontline load, then expanded into risk assessment and claims. None of that would’ve worked without a clean, connected data backbone.

Decide where to host AI

Once you’ve committed to AI, the next decision is infrastructure. You can choose between cloud, on-premise, or hybrid. This choice shapes how easily your systems scale, how data flows, and how quickly new solutions can go live.
Each approach has trade-offs. Cloud offers speed and flexibility, which is ideal for experimenting and deploying models at scale. On-premise gives you full control, which matters when you store sensitive customer data and need to meet strict regulatory requirements. Hybrid sits in the middle, offering you a way to modernize without having to go through a complete rebuild.
For insurers with deeply embedded legacy systems, the latter is often the most practical option. You can keep core data and critical infrastructure on-premise, while running AI services like analytics or automation workflows in the cloud.
Legacy systems aren’t going anywhere overnight. Many still run on custom-built platforms that weren’t designed with modern APIs in mind. Hosting decisions should account for this. If your AI models need real-time access to claims data, policy histories, or customer interactions, make sure your infrastructure can support that connection without delay or data loss.
I take a deep dive into this topic in our article on Implementing AI in Insurance – I highly recommend giving it a read.

Keep humans in the loop (AI as support, not a replacement)

AI can process information faster than any claims team ever could. But faster isn’t always better, especially when precision and empathy matter.

In insurance, that matters a lot.

AI models, particularly those built on machine learning or deep learning, aren’t perfect. They can misread a document, misclassify a claim, or miss a nuance in a complex medical case. And if there’s no human in the loop, that error becomes the final decision, which can lead to unfair outcomes and unhappy customers.

That’s why human oversight isn’t just recommended; it’s essential. Especially in sensitive areas like claims assessment, where AI might scan medical records or accident reports, the final call should always come from a qualified expert – a physician, a claims adjuster, someone trained to apply judgment and context.

Many insurers already use this hybrid approach. AI handles the first pass, i.e., gathering data, spotting red flags, suggesting outcomes. Then a human steps in to review, verify, and approve. It’s not about slowing the process down. It’s about making sure it stays accurate, fair, and aligned with what customers expect.

And in insurance, expectations go deeper than speed. Policyholders want to know someone’s listening. That their situation is understood. Trust is the foundation of the industry and it’s hard to build that if customers feel like they’re being evaluated by an algorithm alone.

So, as you integrate AI, put clear procedures in place. Define when and how human validation happens. Make sure people know where the AI ends and the human begins. Because in this industry, keeping that line clear isn’t just operational – it’s ethical.

Step-by-step guide to implementing AI in your business

1. Understand your customers
Before anything else, develop a deep understanding of your customers.

• Create customer personas if you don’t already have them.
• Map out customer needs, behaviors, and expectations.
• Remember: your customers drive your market, and AI should serve their needs.

2. Map the customer journey
Build a service blueprint that goes beyond a typical customer journey.

• Include the before, during, and after stages of a customer’s experience.
• Identify every touchpoint and interaction.
• Link these interactions to the technology involved at each stage.

3. Identify opportunities for AI integration
Once the customer journey and systems are mapped:

• Pinpoint where AI could enhance the experience.
• Avoid using AI just because it’s trendy – look for genuine improvements.
• Think about both customer-facing and behind-the-scenes AI use cases.

4. Define clear goals
Establish SMART goals (Specific, Measurable, Achievable, Relevant, Time-bound).

• Decide if the goal is to enhance user experience or improve internal efficiency.
• Examples include:
  • Reducing time spent on repetitive tasks.
  • Increasing speed of customer support.
  • Improving product recommendations.
  • Reducing operational costs.

5. Assess and improve internal processes
Evaluate your internal operations:

• Document existing processes if none exist.
• Identify which ones could be automated, streamlined, or eliminated using AI.
• Understand where AI can reduce manual, repetitive work and free up staff for higher-value tasks.

6. Choose the right AI tools or build custom solutions
Decide whether to:

• Use existing AI tools (e.g., AI chatbots, analytics, automation platforms), or
• Build custom solutions tailored to your needs.

7. Set KPIs and define what success looks like
Before launch, determine how you will measure the success of your AI initiative

• Possible KPIs
  • Time saved
  • Number of customer issues resolved
  • Increase in customer satisfaction
  • Revenue growth
• Ensure all stakeholders are aligned on these goals.

8. Start small, then scale
Begin with a pilot project:

• Focus on one clear problem or opportunity.
• Measure results.
• Refine the process based on feedback.
• Use early success to scale responsibly.

9. Prepare for change management
AI adoption can create fears around job displacement or redundancy.

• Communicate openly with teams.
• Reframe AI as a tool for support, not replacement.
• Invest in training and upskilling employees to work alongside AI.

10. Reimagine your business model
AI isn’t just about automating tasks – it’s a chance to rethink your entire business.

• Ask: Could we serve our customers differently with AI?
• Explore ways to differentiate from competitors.
• Look for new market opportunities or business models enabled by AI.

11. Focus on seamless, not visible, AI
Whether or not the AI is visible to users doesn’t matter as much as:

• How well it improves the experience.
• Ensure it feels natural and frictionless.
• Avoid “slapping on AI” just for appearances – customers value outcomes, not buzzwords.

Examples of AI Applications in Insurance

Here are the areas where we have seen leading insurance companies create early impact.

Underwriting

BCG ran a study of international P&C insurers and found that AI can boost efficiency in complex lines of business by up to 36%. The main area that drives this positive outcome is underwriting.

Once again, Allianz shines bright as an example here.

On top of the above-mentioned data platform, they piloted an AI tool called BRIAN, designed to help underwriters quickly access essential information across guidelines and documentation. Instead of searching through lengthy documents, underwriters can ask BRIAN direct questions and get short, accurate answers in real time.

After a pilot involving nearly 3,000 queries from 190 users, Allianz scaled BRIAN into full production. The tool now covers all Property & Casualty and Specialty line underwriting documents.

Customer service

If you’re looking for a practical entry point into AI, customer service is a strong place to start.

In BCG’s research across more than 20,000 insurance service and operations employees, AI-driven tools have delivered productivity gains of over 30%. Not from full automation but from giving human workers better tools.

What brought the biggest boost? Knowledge assistants. These tools help frontline staff find the right information faster. Whether it’s policy details, claims history, or regulatory requirements. In fact, nearly two-thirds of the productivity gains came from these assistants alone. That makes them an ideal starting point for insurers thinking about AI – low risk, high return, and immediate relevance.

Once that foundation is in place, more tools can be layered in:

• Document generators to help craft personalized policy communications or claims letters faster
• Call transcribers to capture conversations in real time, reducing manual note-taking and improving records
• Sentiment analyzers to flag when a customer interaction is going off track giving agents a chance to step in, not just respond.

Claims

Claims is where AI can make some of the biggest impact, fast.

There are two main paths emerging. The first applies to complex claims. These are cases that still require expert judgment but involve a lot of manual steps like extracting first-notice-of-loss (FNOL) details, reviewing documentation, or routing files to the right teams. Here, standalone AI tools can take over repetitive tasks, helping claims handlers focus on decision-making. What’s the result? Cost reductions of up to 20%, and claims processed up to 50% faster.

The second path is full automation, but only for the right use cases. Simple claims, like minor auto incidents or low-value property damage, are increasingly handled through end-to-end AI workflows. From data capture and triage to decisioning and settlement, these journeys are being rebuilt with automation at the core. And it’s working. Insurers are seeing real-time resolution for up to 70% of simple claims, and operational cost cuts between 30% and 50%.

But beyond efficiency, this shift is also changing the experience. Customers don’t just get faster outcomes; they get more transparency, fewer back-and-forths, and a process that feels modern. That matters in a moment when trust and ease are just as important as speed.

What are the roadblocks in enterprise and insurance adoption of generative AI?

Generative AI holds promise, but adoption in insurance faces serious roadblocks.
First, experimentation isn’t production. Tools like ChatGPT are great for quick answers but enterprise-grade insurance work demands security, precision, and integration. In complex claims or underwriting, 99% accuracy isn’t enough. Errors carry real financial and legal risks.
Second, there’s the issue of data privacy. You can’t risk proprietary underwriting rules or customer data leaking into open models. That means using secure, enterprise platforms built for insurance-specific tasks.
Third is the skill gap. Generative AI isn’t plug-and-play. It requires new technical expertise and insurance context, old methods don’t apply.
Finally, time and cost pressures are real. Executives want results now, but deploying safe, high-performing generative AI solutions takes planning and the right foundation.
That’s where partners like Clurgo come in.
In short, successful adoption depends on doing it right – securely, accurately, and with domain-specific tools and experts.

AI will set the pace for innovation in insurance

Generative AI is a breakthrough moment for insurance. For the first time, technology can read, understand, and act on complex documents the way a human would – only faster, and at scale. Early adopters are already gaining an edge, cutting costs, and transforming how work gets done. From broking to claims to reinsurance, the potential is massive. The tools are here. What matters now is action. Insurers who move with purpose, who build smart, secure, AI-driven processes – won’t just catch up. They’ll lead.