Skip to content


Even though the insurance sector is experiencing increasing adoption of artificial intelligence (AI), there’s still a long road left to travel. In addition, those insurers that achieve widespread AI adoption may see uneven success. This brings us to the SIU.

Introducing AI to an insurer’s SIU may seem like an easy sell – investigators need to do less work in order to find more fraud – but this underestimates the potential drawbacks. Research from Gartner shows that up to 85% of AI implementations fail. How can your deployment beat the odds?

Change management: laying the groundwork for AI in the SIU
Even the most technically sophisticated AI program will fail to deliver if no one decides to use it. What’s more, even the most technically sophisticated users might hesitate when confronted with the idea of using an AI tool. Lastly, SIU leadership has no way of knowing which users will be enthusiastic about AI, and which will be less so.

Going in blind only opens up the door to unpleasant surprises. To succeed at change management, leaders need to understand how their users will react to change.

One way to start is by conducting a series of interviews with key stakeholders such as senior investigators. The goal here isn’t to convince them to accept AI. Instead, the purpose is to:

  • Let them know that a change is on the way
  • Learn what users dislike most about the existing fraud detection system
  • Understand what workflows they consider to be essential
  • Identify what their "Ideal State" is
  • Identify where potential resistance to a new solution is
  • Create a change management plan to address all areas


This is important because distrust is one of the most common reactions to AI. AI will usually change the SIU in a way that will eliminate a number of existing workflows. Problems arise when these changes either A. fail to eliminate processes that users dislike or B. eliminate processes that users like. B. is the most common problem with AI, because oftentimes the missing process makes it harder for users to double-check the AI’s work. When this happens, no trust is possible, and users will reject the solution.

By taking these meetings, SIU leadership will learn the most effective ways to implement AI. They can make certain that AI eliminates most or all of what was bad about the existing system, while working with the vendor to make sure that users can replicate their essential workflows under the AI solution.

Demonstrating change management leadership
When the SIU leadership demonstrates engagement during the change management process, they can help to forestall many of the common objections. Empathy is key. Leadership should do their best to acknowledge that even though they’re doing their best to preserve essential workflows, things are going to change, and there is going to be a learning process.

Here, it’s important to anticipate common objections. At Shift, we often hear that employees are suspicious of the idea that AI will make their work easier and more efficient. They understandably worry that this means they’ll be asked to do a lot more work in the same amount of time. 

We try to get ahead of this by saying that this usually isn’t the case – the job functions may get easier, but since they also produce much better results, the SIU is rarely if ever pressured to work harder.

Although every organization is different, it’s still possible for leaders to anticipate objections and demonstrate engagement. If you’re planning a busy schedule of changes over the next few years, it may make sense to get a change management certification, which will help you put a more formalized process in place. Alternatively, you can ask if the representative from your vendor has similar credentials.

Choosing the right AI solution
Even if your AI solution performs exactly as advertised, the absence of certain features may make it difficult for your employees to trust.

Some AI solutions operate out of what’s known as a “black box.” This means that it can be difficult to understand why the AI chooses to flag some alerts rather than others. That’s because the algorithms it uses aren’t designed to explain how they weigh certain inputs.

The SIU performs sensitive work. Nobody wants to investigate an alert – much less investigate potential crimes – without a thorough understanding of why they’re being asked to do so. Therefore, opting for an explainable AI, one that lists the reasons why it’s flagging alerts, is a better choice in terms of adoption.

Shift Technology offers fully explainable AI solutions that make life a lot easier for the SIU – but change management is another big part of our offering. We work closely with our partners and create highly customized implementations that reflect the needs of the end users. That means we achieve a high level of customer success – helping spur even greater adoption of AI within the insurance industry.

For more information about Shift Technology and how we’re driving AI-powered solutions to detect insurance fraud, request a demo today.