Policy Bot turns one. Meet the Policy Agent.

“Do I have to tell HR if I’m dating my coworker?”

One year ago, we shipped Policy Bot because of questions exactly like that one. Every workplace generates them by the hour. Nobody wants to read a 40-page handbook to answer one, and let’s be honest about the dating question in particular: nobody wants to ask a human that, either.

Last year, Ethena’s Policy Bot made the answer a chat window instead of a confession. Free, opt-in, trained only on your approved policies and handbooks. Every answer cites the exact policy it came from, and when a question falls outside your policies, it says so instead of improvising. The willingness to say “I don’t know” remains an underrated feature in both chatbots and coworkers.

A year of real questions

Since launch, we’ve talked to hundreds of organizations using the Policy Bot, and somehow, the smart chatbot graduated from “new feature” to “the thing the whole team has bookmarked.”

We were surprised to learn that teams were going off-book in ways we didn’t even anticipate: HR teams adopted it for themselves before employees ever saw it. One HR leader at a government contracting firm told us she uploaded the entire handbook the day she got the launch email, and now her whole team keeps it bookmarked. “I like that it can be a standalone website,” she said. “I can link directly to the bot.”

Before Policy Bot, employee questions had three destinations: the Slack channel where HR answers the same one every week, a spelunking trip through the HRIS, or a DIY bot project someone had to babysit forever. An HR VP at an e-commerce company put it plainly: “Right now it’s just me and the other HR VPs fielding all the questions.”

What customers asked for

Then the feature requests started, and they were remarkably consistent. Call after call, customers asked for the same four things: show us what employees are asking, because the question log doubles as a compliance radar. Keep policies living, so the bot answers from today’s version instead of last year’s PDF. Let a human check the AI’s answer before it sends. And when a policy changes, make the training change with it.

Customers also invented uses we never saw coming. Our favorite came from a compliance lead at a government services nonprofit who pointed the bot at his own policies to hunt for loopholes: “I started redrafting the policy, then fed it in and asked: is the loophole still there? And it would say no, you fixed it. Perfect.” Red-teaming your own handbook wasn’t in our launch plan. It absolutely should have been.

The bot grew up

Those four asks became the roadmap. Over the year, they shipped one at a time:

• A review mode where AI drafts the answer and a human approves it before it sends.
• Policies uploaded as links, so the bot reads the current version and never a stale PDF.
• Question reporting that turns the log into the compliance radar customers kept describing.
• A training builder that kicks in once a policy is uploaded

The Policy Bot outgrew its original purpose: it read policies, showed teams what was missing, and fed what it learned back into training. That’s more than a bot. That’s a Policy Agent.

Earlier this summer, we made it official. The Policy Agent scans policies for gaps, overlap, and unclear guidance. It drafts updates in the company’s own policy voice and routes them to compliance for sign-off. Policy Bot lives inside it, still fielding employee questions, which now double as signals for where policies need work.

One thing held constant through every release: the agent reads, flags, and drafts, and the compliance leader approves. Nothing ships without human sign-off.

Customers who previewed the agent this spring skipped the philosophy and went straight to work. “You could ask it questions,” a chief compliance officer at a software company told us. “‘Can I give this gift to this particular person?’ And it would check the policies and say, hey, I need a little bit more information.”

Where it goes next

The Policy Agent is one of four agents on the Ethena platform, and they talk to each other. Say employees start disclosing prediction-market accounts. The Disclosure Agent notices the trend, the Policy Agent drafts the policy you don’t have yet, and the Training Agent turns the approved version into a scenario your team will actually remember. One summary lands on your desk. One approval sends it everywhere it needs to go.

The agents draft. You decide.

A year ago, the pitch was “your employees will stop Slacking HR about the dating policy.” Still true. Still free to turn on. But answering questions turned out to be the warm-up. The real job is catching policy issues before they become your problem, and that’s the job Policy Bot grew up to do.

If you want to see the Policy Agent working with your own policies, let’s talk.