Artificial intelligence (AI) projects that achieve true, measurable transformation can be extremely challenging to execute. Nonetheless, AI’s benefits have become clear: improved time to market, increased business and IT agility and flexibility, and positioning for the future. These benefits can lead to improved employee, agent, and customer satisfaction.
However, these projects may require more financial investment and organizational attention than originally anticipated, and they are at high risk of running over budget and schedule while failing to deliver promised benefits. As insurers move from initial AI proofs of concept, they will need to consider a number of factors around AI project readiness:
- Business readiness: The carrier must ensure that the stakeholder business units are ready for the project and have a clear vision of what the project should achieve, as well as clear guiding principles and responsibilities for decision-making.
- IT and technical readiness: Solutions architects need to map out the required architecture, including open-source architecture, create an inventory of required interfaces, and establish principles and processes to govern technology and model decisions around training, drift, and hallucination. These architects may conclude that, in some cases, small language models that have more restrictions on the type of data they can utilize but avoid hallucinations may be a better fit architecturally for the problems to be solved.
- Data readiness: Garbage in, garbage out. Insurers must ensure that the data is well organized, easily accessible, and cataloged with the appropriate business rules to ensure responsible AI outcomes around personally identifiable information and personal health information. If this isn’t done, the results of the model could be incorrect, misleading, or expose the firm to regulatory fines and reputational damage.
- Program readiness: Carriers need to staff their AI programs with key roles such as pod leader, prompt engineer, model mechanic, data engineer, and data scientist. They also need to work with vendor and systems integrator partners to define the scope of the project and plan execution. But where to get these folks? This is a challenge many carriers will face. The solution inevitably will come from unique recruiting programs on college campuses, summer internships, and re-training of staff who have an interest or are in similar adjacent roles today, where their skills can be augmented.
- AI maturity: Insurers need to understand how close the organization is to becoming an “AI Master” from a people, process, and technology perspective. The insurer needs to determine if it is prepared at an enterprise level to make informed business decisions utilizing AI or would rather use AI in niche applications, e.g., customer service piloting, application submission process optimization, enhanced claims adjudication, computer-assisted underwriting). It will initially be difficult for the organization to assess its maturity, and this maturity at midsize to large insurers may be uneven across business lines.
Addressing the above challenges will not be easy. Datos Insights recommends a central point of coordination for AI in general, as opposed to each department or division “doing their own thing.” Proofs of concept will help determine in which areas organizations should perform AI projects.
The enterprise needs to start looking at the AI big picture (i.e., the items listed above). Once a full AI project or projects are chosen, the means to achieve success should already be in place. The alternative to this thoughtful approach is a more hype-driven, opportunistic approach to AI, which will inevitably lead to higher costs and delayed deliveries, discrediting the technology and the IT organization that drove it.
For more on generative AI, see Datos Insights’ report Top Five Questions Insurer CIOs Have About Generative AI, September 2023. Please also feel free to reach out to me for a conversation around AI readiness in your own organization.
