6 questions to ask before implementing your benefits AI plans

Editor's note: This blog post is a direct follow-up to Episode 2 of Benefits Innovation Lab. If you haven't checked it out yet, be sure to register to view or listen on demand. And download the accompanying free worksheet to get the most out of this article.

There are so many ways your benefits team can leverage artificial intelligence (AI) to improve efficiency and boost benefits outcomes for your people. Once you've identified the first problem you'll solve with AI, it's time to get the rest of your organization on board.

After all, implementing AI doesn't happen in a vacuum. You'll likely need support or buy-in from folks in finance, the legal department, or the C-suite. Even in smaller organizations, you'll want to let other team leaders know what you're doing with AI.

To prepare for those conversations, we'll walk you through six questions you should be ready to answer. We'll also look at an example AI use case for a benefits team: turning your benefits handbook into an AI-powered chatbot that can answer employee queries.

First, let's talk about AI committees

Your organization may already have AI infrastructure in place. As many as one in five companies have a formal AI committee¹-a cross-team panel dedicated to governing and standardizing AI use in the organization.

These committees typically review and approve AI proposals from various departments, weighing in on topics like data management, compliance and legal considerations, and ethical questions associated with AI use. The committee may include representatives from:

• Executive leadership

• Legal and compliance departments

• Information security

• Data science

• Stakeholder groups, such as client working groups

AI committees will consider your benefits team's proposal and ask questions about important topics (like those we'll cover below).

Even if your company doesn't have an AI committee, you'll likely be having similar conversations with team members outside the benefits department. So, let's start exploring how you can prepare.

Question 1: Who are the key players in your AI proposal?

The first thing you'll want to do is decide who will play what role in requesting and managing your benefits team's AI proposal. Typically, you'll want to designate:

• A requestor. This person will present your AI idea to the AI committee or the individuals you'll need support from outside your benefits team. They'll answer questions about the proposal and help steer it through the approval process.
• An executive sponsor. This person in upper management will be responsible for the AI use case at an executive level.
• An owner. This person will implement or manage the AI system once it's in place.

Let's take a closer look at how to answer these questions by exploring an example from the world of benefits.

Our example: We'll imagine we're proposing an AI chatbot that will use our company's benefits handbook to answer employee questions. In this case, our requestor will be the benefits team leader. The executive sponsor would likely be the head of the Human Resources or People team. And the owner would either be the benefits team leader or someone on the benefits team tasked with employee communications. In a smaller organization, all three roles might be filled by the same person.

Question 2: What's your elevator pitch?

You should be prepared to give a brief but informative description of your proposal. Be sure to describe the specific problem you're solving with AI and who the main audience will be. Specify whether the AI system or tool will be used internally or externally. This will help the AI committee determine the level of risk and which guardrails they might want to put in place.

You should also be sure to outline the benefits of the project. Pointing out how your proposal aligns with the strategic goals of the company will illustrate how your work ladders up into broader organizational aims. You should also be ready to discuss how your intended audience will benefit from the use of AI.

Our example: An elevator pitch for a benefits chatbot might look like the following:

Based on our recent employee satisfaction survey, we know 60% of our employees wish they knew more about their benefits. As a benefits team, we spend several hours a week answering relatively simple questions about how to use our various benefits offerings. Although we provide a benefits handbook, some employees may find it hard to locate the answer to a specific question.

We propose training an AI-powered chatbot on our benefits handbook to answer these questions from employees and their families. When fully implemented, the chatbot will allow employees of the company, their spouses, and dependents to get answers to simple queries in an accessible way. This will make it easier for many people to access and use their benefits, maximizing benefits outcomes for the organization and laddering up to our broader goal of retaining 80% of existing employees over the fiscal year.

Question 3: What are the risks associated with your AI project?

No AI use case is risk-free. The people working with you to implement your AI proposal will likely have specific questions about how your use case will impact things like data management, company finances, and employee privacy.

It's unlikely you'll be asked to provide high-level technical details, but answering a few questions up front will help those with the necessary expertise make important decisions about how your proposal should be implemented.

Here are some details you should be ready to provide:

• The extent to which AI will make decisions on behalf of users. Does your use case rely on AI-driven algorithms to make a call on important items? Will it restrict itself to recommendations? Or will it merely provide information?

• Whether AI will interface with or produce content for external audiences. Will AI be used exclusively by internal team members, or does it also take on the associated risk of communicating with people outside the organization?

• The nature of the content AI will produce. If you're using generative AI, or an AI tool that produces content, be ready to describe what kinds of content that could include and whether your organization might consider it to be its intellectual property.

• How the AI project might affect company financial outcomes. Could your AI project have any material effect on how the company makes financial decisions? Or might it come with tax or compensation implications?

Be sure to consider data usage especially. AI requires data to be effective, and that always comes with risk. Be ready to provide information on the kind of data your proposal requires, how the AI use case will use that data, and what broad measures might be implemented to protect it.

Our example: A benefits chatbot is a relatively light use of AI and doesn't touch many of these areas. The chatbot is not meant to make decisions for employees, but rather answer simple queries they may have about the program or their plans. It's also relying on existing, documented company information rather than employees' private information.

However, given that the chatbot is trained on the company's benefits handbook, its output will likely be considered the company's intellectual property. It may also be worth considering the error threshold for answers given to employees. You may want to include a disclaimer that answers are for general use only, and that employees should reach out to the benefits team if they need specific or detailed help.

Question 4: What technology do you require?

Your organization will want to know if you have a specific AI platform or tool you'd like to use. Or if you'll require the support of data scientists and engineers to build something custom or to make specific modifications.

Be ready to talk about what conclusions you've arrived at and why, as well as whether any contracts or licensing agreements are required. If you're undecided or need help, be prepared to talk about some of the specific requirements you need. This way, your company's Chief Information Officer (CIO) or information technology (IT) partners can conduct a capability assessment and help you arrive at a solution.

Our example: A benefits chatbot can take advantage of existing tools and frameworks built by AI vendors. You can work with the AI committee to determine the right partner, or come to a specific proposal yourself. IT partners can also help make any modifications you feel you'll need.

Question 5: What roles will humans play in the oversight of AI?

AI always works best when implemented alongside human support. The specific role humans will play depends on your AI use case. For example, you should consider whether AI output will require human review or validation, as well as ensuring human time is allocated toward supporting the primary users of the AI tool or system

Our example: A benefits chatbot should require little human support, but humans should be available for initial testing and refinement as the chatbot is trained. Once implemented, benefits team members should be ready to verify or expound on answers when employees ask.

Question 6: How will you know your AI use is successful?

Lastly, you should make sure you know what success looks like. All organizations thrive on data, and yours will likely want to know how an AI tool or system is being measured. Think about the key metrics that will let you know your team goals and broader organizational goals are being met.

You should also be prepared to discuss the specific steps you'll take to make sure team members and AI users will have the skills and knowledge they need to use the AI tool or system.

Our example: To measure the success of an AI-powered benefits chatbot, look to employee feedback about the tool. You could measure satisfaction rates through the tool itself, as well as determine whether the average time your benefits team spends answering employee questions is reduced.

A benefits team implementing a chatbot will also want to first train its team members and then the broader employee audience on how to use the tool, while establishing expectations on what the chatbot can and cannot answer.

Succeeding with your AI proposal

The more time you take to round out your AI proposal, the better you'll be able to discuss it with folks outside the benefits team.

These first conversations are key for determining the overall success of your project. Even if an item doesn't come up in discussions with your AI committee, it may prove useful as you move through the implementation process.

Stay tuned to the Remark blog and to Benefits Innovation Lab for more insights on how to improve AI use in your benefits team.

Public link

Please use the above public link if you want to share this noodl on another website.