AmTrust Financial Services, Inc

Artificial intelligence (AI) has swept American industry by storm. Most of the leading companies in the insurance industry are now allocating a significant portion of their IT budget to AI. While I’m certainly not an AI skeptic (AI was used to help me craft, edit and proofread this article), I believe that many AI projects will not deliver the promised return on investment, especially if that AI investment displaces resources that were previously allocated to data infrastructure, data governance and data quality. Additionally, investments in training and upskilling humans will be a necessity for the success of many AI projects. Humans need to be trained to build, collaborate with and adopt the recommendations of AI systems. Only by combining strong data governance with empowered, well-trained employees can organizations fully reap the benefits of AI.

Data is the fuel for modern AI systems

Data is the key ingredient to train AI systems. Yet, many companies are enthusiastically pursuing AI without having their underlying data infrastructure in order. AI models can fail if they’re fed incomplete and inaccurate data. Data is too often stored in silos and not easily combined and merged. Many insurers, both large and small, rely heavily on Excel spreadsheets for storing, analyzing and merging data. In addition to the ability to access and merge the data used to train AI models, the absence of robust data management practices that includes accurate documentation in the form of data dictionaries and data lineage, can make it overly challenging and time consuming to compile an accurate and complete dataset to feed into an AI model.

“AI has the potential to improve employee job satisfaction and work/ life balance if applied correctly. It can reduce the need to perform tedious and repetitive tasks, thereby freeing people to focus on more complex, rewarding and creative aspects of their roles, potentially improving job satisfaction and retention.”

Moreover, much of the data consumed by AI must be obtained and input by humans such as underwriters, claims adjusters, call center agents and risk control specialists. These professionals must be trained to ask the right questions and document the appropriate details as consistently as possible, as poorly captured data can lead to flawed AI outputs and missed opportunities. Quality audit programs will continue to be a vital way to ensure that personnel perform their jobs consistently and correctly, helping to maintain uniformity in data collection across teams and offices.

Artificial intelligence and Human intelligence complement each other

Despite rapid improvements, AI is not yet advanced enough to supplant most human decision-making—at least not for the foreseeable future. Currently, only tasks that are both simple, and where the cost of making an error is relatively low, can be exclusively turned over to machines while incurring a reasonable level of risk (for example fully automating simple auto physical damage claims, answering basic customer inquiries about their claim or policy). Even for these tasks, human expertise is necessary to train and monitor the performance of the AI system. For more complex tasks humans will need to be the ultimate decision makers, but those decisions can be aided, enhanced, and accelerated by collaborating with AI if the human decision makers have confidence in the AI’s recommendations.

For AI to be successful on a large scale, it will be critical for employees to buy-in to its adoption. Unfortunately, many employees have the perception that AI is simply a tool for eliminating jobs. AI has the potential to improve employee job satisfaction and work/life balance if applied correctly. It can reduce the need to perform tedious and repetitive tasks, thereby freeing people to focus on more complex, rewarding and creative aspects of their roles, potentially improving job satisfaction and retention. Senior leadership will need to build trust with employees to get them to buy-in and fully enlist their cooperation to get the most value out of AI projects.

To win over regulators as well as employees, AI based decisions must be transparent and explainable. Employees will have more confidence in adopting the recommendations of an AI system if they can understand the data and logic behind the recommendation and can explain it to others if necessary. Additionally, understanding the AI’s reasoning will enable users to suggest improvements and enable a feedback loop to continuously improve the AI’s results. Moreover, insurance regulators have signaled that they are unlikely to accept explanations for decisions that lack supporting reasoning or rely solely on “black box” algorithms.

The Next Step in AI-Enabled Growth

AI has the potential to bring enhanced productivity and profitability to insurance companies. However, the success of these AI initiatives will require more than just investing in and implementing the technology itself. To unlock the full value of AI, leaders must ensure their data is accurate, accessible and well-governed—and their teams are equipped with the skills and confidence to collaborate with AI systems.