Data Society built a Slack-based HR agent so employees can ask benefits and policy questions and get an answer immediately, instead of digging through a folder or tracking down the VP of People. Global VP of People Catie Maillard says the trade she made was real, not theoretical: a lower-touch HR org that catches fewer informal signals, in exchange for support that doesn’t depend on someone being willing to bother a busy person. She’s not pretending the trade-off doesn’t exist. She’s arguing it’s the right one, and that every team rolling out an agent needs to name its own version before flipping it on.
What the agent actually does
The use case is deliberately narrow: routine questions that used to mean opening a Google folder or messaging Maillard directly.
“Instead of having to go look in a Google folder or reach out to me to figure out where a process lives, employees can just Slack an agent: ‘Hey, I’m trying to change my benefits. When is open enrollment?’ Or even, ‘I want to learn more about communicating to executives. Do we have any content for that?’”
Catie Maillard, Global VP of People, Data Society
Data Society isn’t an outlier here. IBM has found that 62% of employees already turn to virtual assistants for HR needs, and Gartner projects that conversational AI platforms will handle 75% of routine service inquiries as adoption matures. Airbus’s recruiting chatbot, Bessie, resolves 80% of candidate questions automatically, including the 40% that come in outside working hours. The pattern across these deployments is the same one Maillard describes: most of what lands in an HR inbox isn’t a judgment call, it’s a lookup.
The trade-off she names before you ask
Maillard doesn’t frame the agent as a clean win. She names the cost directly.
“There are trade-offs to this that we’ve already considered. One is that we now have a lower-touch HR org, so I miss the opportunity to connect with an employee where I could learn something was happening in the organization. That’s a trade-off I’m making to get employees answers more quickly and, honestly, give them more support.”
Catie Maillard, Global VP of People, Data Society
That lost opportunity isn’t small. Gallup’s research across 2.7 million employees and more than 2.5 million work units found that managers account for at least 70% of the variance in team engagement scores. A lot of that influence isn’t built in scheduled one-on-ones. It’s built in the ad hoc moment when someone mentions something is off and a manager or an HR partner actually hears it. Every question an agent absorbs is one fewer chance for that moment to happen.
Why she made the trade anyway
The counterweight, in Maillard’s telling, is that the human channel she’d be protecting wasn’t actually open to everyone in the first place.
“A lot of times it’s a barrier to entry just to reach out to me, knowing I’m busy or have a lot going on. So we took that trade-off: I might have fewer ad hoc touchpoints with folks, but at least employees will have a better service overall. But you’ve got to consider that. What are the trade-offs and the impacts on all the stakeholders when you’re rolling out an agent?”
Catie Maillard, Global VP of People, Data Society
The friction she’s describing shows up in survey data too. In a 2024 MyPerfectResume survey of nearly 1,000 U.S. workers, 67% said they had trouble getting a timely response from HR, and 85% said they hesitate to approach HR about a work issue at all. Whatever the reason, busy schedules, unclear ownership, or reluctance to be a bother, the effect is the same: a question that needed an answer in five minutes sits unanswered for days, or never gets asked. An agent doesn’t fix the underlying trust gap those numbers point to. It does remove the specific friction Maillard names: the discomfort of interrupting someone who’s visibly slammed.
What the agent doesn’t replace
The trade-off only holds if the touchpoints being lost get replaced by something else, not just left as a gap. Maillard’s framing puts the burden on the rollout, not the tool: know what signal you’re giving up, and build a deliberate way to catch it, whether that’s a pulse survey, a skip-level cadence, or a standing check-in that isn’t contingent on someone deciding they’re worth the interruption. An agent that resolves the “when is open enrollment” questions frees an HR team to spend the time it gets back on exactly that kind of proactive outreach. It only works out that way if the team actually spends it there.
Frequently asked questions
What does Data Society’s HR Slack agent actually handle?
Routine, lookup-style questions: benefits and enrollment logistics, where a policy or process lives, and where to find existing content, such as resources on communicating with executives. It’s built to answer instantly in Slack rather than route to a person.
What’s the main trade-off of adding a self-service HR agent?
Fewer ad hoc, informal touchpoints between HR and employees. Those moments are often where a manager or HR partner picks up on something happening in the organization before it becomes a bigger issue.
Why do employees hesitate to reach out to HR directly?
Survey data backs up the barrier Maillard describes: a 2024 MyPerfectResume survey found 67% of U.S. workers had trouble getting a timely response from HR, and 85% hesitate to approach HR about a work issue at all.
Does an AI agent replace HR’s role in catching employee problems?
No. It absorbs routine questions so HR’s limited time isn’t spent on lookups, but Maillard is explicit that this doesn’t happen automatically. The lost informal touchpoints have to be replaced with a deliberate mechanism, not assumed away.
What should people leaders do before rolling out an HR agent?
Name the specific trade-off for their org, not a generic one. Identify which touchpoints the agent will absorb, decide whether those touchpoints were actually reaching the employees who needed them, and build a replacement for whatever signal gets lost.
Next steps
Before rolling out a self-service HR agent, name your own version of this trade-off, and decide what will replace the informal touchpoints it absorbs. Data Society’s AI Advisory work helps HR teams map those stakeholder trade-offs before launch, and AI upskilling programs help teams build the habits that catch what the bot won’t.
