<img height="1" width="1" style="display:none;" alt="" src="https://px.ads.linkedin.com/collect/?pid=4958233&amp;fmt=gif">
Article
5 min read
Richard Pugh

Generative AI (Gen AI) promises to transform the insurance industry. AI models are already in use improving efficiency, reducing fraud and offering tailored solutions that delight customers. But, as we’ve learnt over time, generative AI can ‘hallucinate’ and get things wrong, with costly implications. Consequently, insurers have been cautious in applying AI to core business processes. 

 

To address this, Endava has embraced a new concept called agent-based AI, or agentic AI. Flexible enough to work in real-world situations and deliver results with a high degree of accuracy and reliability, our agentic AI industry accelerator, internally called Morpheus, is a first-of-its-kind solution that could shape the industry.  

 

In this blog, we’ll discuss the concept of agentic AI and how our solution can deliver on the generative AI promise.

 

 

Teams of AI agents working together

 

As a colleague put it, every CEO knows that diverse teams do better work; together the team is greater than the sum of the parts. With this in mind, agentic AI builds a team of AI agents that ‘talk to one another.’ Each AI agent has a different perspective and role. You create a healthy tension between team members, passing information back and forth until they agree the result is correct.

  

To provide context, you could set up a team of agents like below to assess an insurance claim. To represent the complexity of decision-making and the different perspectives each agent has different roles and responsibilities.

 

Morpheus Agentic AI campaign infographic V5

 

 

This team of diverse perspectives creates the checks and balances to deliver quality results.

 

Imagine having teams of highly skilled workers, who understand all the rules and objectives of a business and understand how to achieve their tasks using a set of tools, that you can deploy with little cost against any process. You could assign an agent to every customer in your business just to ‘know’ them and look after them.

  

One of the biggest challenges insurers will likely face is knowing where to start. We're already helping insurers understand these capabilities and their impact.

 

  

Agentic AI builds on four advances in generative AI

 

Generative AI is a technology built on large language models (LLMs), huge statistical models that ‘guess’ the most likely next word based on the words that came before.

  

When they are guessing based on common phrases “the cat sat on the …” they tend to be pretty accurate “…mat”. But when they have to guess more unusual phrases “the llama sat on the…” the results are more varied. To illustrate this, Microsoft Copilot gave me “…top of the hill”, “…patchwork quilt” and “…edge of the meadow”.

 

To make generative AI more reliable (without sacrificing creativity and flexibility) you need added functionality. To achieve this enhanced functionality, agentic AI builds on four generative AI advances.

 

  1. 1. Agents have roles: When you ask ChatGPT a question, it returns a vanilla result, without bias or constraints. We can now create agents that have specific roles, knowledge, objectives, biases, constraints etc. Imagine the marketing potential of a chatbot that has the objective of positioning you as a leader in the insurance industry, who understands your mission and tone of voice etc.
  2.  
  3. 2. Agents use tools: Agents have been able to use the web on our behalf for a while, but agents can increasingly use a broad set of tools, such as email, CRM systems, office tools etc. You could create an agent that interacts with customers via email or LinkedIn, who updates your CRM and is good at booking sales meetings in calendars.
  4.  
  5. 3. Agents follow workflows: Agents can be given business processes to follow. Unlike conventional process automation, these workflows can be flexible and include value judgements rather than strict ‘true/false’ conditional statements that often fail to reflect real world situations. 
  6.  
  7. 4. Agents can collaborate: The big advance in agentic AI is that we can create ‘teams’ of agents that can collaborate on complex tasks, like a renewal process, or an onboarding process, or a claims process. Any stated goal. For example, ‘perform a first pass analysis of an insurance claim and suggest a resolution / escalation with rationale.’ 
  8.  
  9.  

Collaborative insurance automation

 

Think of our AI industry accelerator as a factory that builds teams of agents and gives them domain knowledge, roles and business processes so that they can collaborate on complex tasks.

 

This team of agents can have an informed discussion and together reach the correct conclusion. (the below will also use a graphic to represent it)

 

  • Each AI agent in the team has a different role 
  • The agents debate every detail from their individual perspective 
  • Unlike a human debate, there is no emotion or entrenched positions 
  • They are not trying to win, they just want to quickly find the best outcome 

 

As a result, it can use AI to automate business processes that are too complex and subtle to be accomplished with conventional process automation.

 

The AI agents are used to automate difficult tasks that require domain knowledge and flexible thinking. Teams of AI agents are built to collaborate and challenge each other and to guarantee reliable results.

  

They can be deployed robustly in almost any enterprise environment, thanks to potent reference architecture, including being able to interface with all major LLM models and cloud platforms. The ways in which results are reached are reported with transparency enabling use in strictly regulated industries. 

 

 

Empowering our insurance customers with agentic AI

 

Our AI industry accelerator is a proven, enterprise-ready framework on which solutions can be built. We’re already using agentic AI to solve real world problems for insurers, helping them to eliminate administration costs and gain competitive advantage. 

 

Our AI experts provide commercially-driven support that guides deployment, helping you to ideate, prioritise, scope and deliver measurable results for your business.  

 

We have a proven methodology roadmap for use cases, with check points along the way to demonstrate value. We’re excited about the opportunity agentic AI presents for insurers and coupled with our insurance domain knowledge, we're here to help you make agentic AI a reality.  

 

If you’d like to learn more about data and AI download our e-book Leveraging Data in Insurance: developing an effective data + AI strategy, or get to know our first-of-its-kind AI solution that combines the power of data and multi-agent autonomous teams.

 

 

Agentic AI use cases 

Use Case 1: Onboarding

We are working with an insurer to automate their onboarding process. Agentic AI understands the rules of the onboarding process, collects the appropriate information and follows the correct admin workflows to make onboarding easier for the customer and more efficient for the insurer.

Use Case 2: Insurance Claims

We are developing a claims processing use case where agentic AI balances the needs of all stakeholders to resolve the claim quickly and correctly.

 

Use Case 3: Data Extraction

For this project, each data interaction is meticulously logged by agentic AI, tagged with metadata and is fully transparent, allowing for verification of actions, timing and rationale, for complete clarity. Human oversight is possible through a rich governance layer applied to all activity.

 

Data integrity

Shifting to a data integrity model offers greater trust and flexibility for businesses. Agentic AI embodies this approach, founded on the notion that initial inputs may not be flawless but can evolve into reliable and actionable assets through processing, thus enabling more impactful outcomes.

 

 

No video selected

Select a video type in the sidebar.