Spending on generative artificial intelligence continues to grow. U.S. private investment in AI reached $109.1 billion in 2024, but Microsoft alone plans to spend $100 billion on AI in the next fiscal year. Amazon plans to spend a similar amount. Alphabet plans to spend $85 billion. Meta plans to spend about $70 billion.
All these investments are made with a singular goal: Can we use Gen AI to slightly improve the virtual assistants on workplace finance platforms?
For all the hype and spending, real world applications for Gen AI remain limited, in workplace finance and elsewhere, leading to the current Gen AI paradox. Eight out of ten companies report using Gen AI, but the same percentage report no bottom-line impact. Firms are all looking for ways to turn easily generated, inexpensive and quasi-reliable text into business results. For every USAA that can celebrate a popular A.I. assistant that helps its customer service workers answer questions more efficiently, there is an AI-powered vending machine that can be talked into giving away free tungsten cubes. A misaligned super-intelligent AI might take over the world by 2027. But right now Google’s top Gen AI customer success stories include an attempt to automate Wendy’s drive-thrus and using AI to generate recipes with your GE fridge.
With that paradox in mind, how are financial services firms reacting to Gen AI in their customer support flows?
CI’s Workplace Finance team recently published a report examining the virtual assistants (VAs) available from equity plan providers. The report covers VAs from six firms, three workplace equity plan providers and three retail brokerage firms, that provide equity compensation services. Below is a preview of two key findings from that report.
Identifying and Meeting Consumer Expectations for VAs
Large language models are swiftly changing user expectations for intelligent chat interfaces. While stock plan services providers do not need to replicate the full scope of these systems within their VAs, firms that offer VAs with relatively narrow comprehension capabilities may run the risk of frustrating users who expect sophisticated chatbot experiences.
There are two notable approaches that firms are taking to compensate for these limitations. Some integrate a wide array of preset prompts in an effort to constrain the types of questions users pose. Others rely on follow-up questions to narrow topics after an initial query.
E*TRADE takes the first approach. Leaning heavily on a set of prefabricated topic buttons to steer participants toward relevant content, the E*TRADE Assistant presents users with eight prompt options following its initial greeting, most of which lead to secondary selections of subtopic prompts that lead, in turn, to more specific requests and/or commands. While this interaction flow helps ensure that users focus on available content, it also creates an experience more akin to conventional menu navigation than the free-form conversation that characterizes consumer-grade AI tools. Reinforcing this impression, users can scroll back up to initial prompts at any time to effectively restart their conversation with another topic.
Notably, however, the VA can recognize most free-form queries that have an approximate button equivalent—for example, the VA gives the same response to “whats rsu?” as it does to the What Are the Types of Company Restricted Stock? prompt button—suggesting that the buttons serve more as guardrails than rigid input options.
In comparison, Fidelity’s workplace VA provides a much shorter list of initial prompts, informing users that short phrases or questions are most likely to produce relevant results. Once users provide an input, the VA has several response strategies, which include presenting subtopic follow-up prompts, links to potentially relevant site areas culled from the conventional sitewide search function, or—when the VA clearly identifies a query or phrase—explanations or directions that often involve hyperlinked resources. When the VA presents search results, users have the option to follow a View More Search Results link to open a full result page where they can explore additional links or try different search inputs.
Using VAs as Frontline Support and Filtering Inquiries
Stock plan providers are also strategically deploying VAs as frontline support filters to intercept and resolve high-volume routine inquiries before they reach human representatives. At the same time, these digital assistants can streamline the user journey to connect with a live CSR by segmenting requests based on topic and complexity, escalating to live chat on demand, and providing agents with request previews and chat histories to improve efficiency.
A comparison of firms’ strategies in positioning VAs in relation to more traditional support channels reveals that while firms are generally united in their efforts to employ VAs in a support capacity, they differ significantly when it comes to the specifics of the role.
Fidelity stands out for how it funnels participant support queries through its VAs, particularly within the NetBenefits workplace site. Specifically, the top of the workplace platform’s Customer Service page is dominated by an embedded VA interface, making it clear that it is the default first option for all support participant requests. While an area beneath the fold includes a link to connect with a live CSR, the same section also includes an additional nudge to engage with the VA that participants presumably just scrolled past.
Notably, when participants open the VA within another site area and request customer support, the VA provides a link to the same Customer Service page where participants are confronted with a second, embedded VA chat window. The VA’s strong natural language recognition suggests that the tool is equipped to recognize the type of colloquial phrasing participants might use with a live CSR.
While the firm clearly nudges participants to try to resolve issues using the VA before reaching out to a CSR, requests for customer service promptly trigger the option to begin a live chat. This prompt handoff may help alleviate the frustration of participants who find that live chat with a CSR is not immediately available from the dedicated Contact Us page.
Workplace Finance Monitor research can help any team looking to improve their VAs and overall digital customer experience across financial wellness, HSAs, and equity compensation. Contact us today to learn how you can gain access the virtual assistant report—with all four key findings and detailed firm profiles for all six firms and their VAs—and the full suite of Workplace Finance Monitor research.