Better listening or better watching? Employees can tell

Any communicator who reads 2026 trend reports will walk away with a clear assignment: stop relying on the annual survey, start measuring sentiment in real time. Watch how people interact with your content. Track time spent reading, repeat views, where they drop off. Read the signals.

An employee sees the same activity and calls it something else. They call it being watched.

Both of them are describing the same tool.

That is the problem nested quietly inside this year's most-repeated piece of internal comms advice. The push to "measure sentiment beyond the survey" sounds like better listening. Run it forward a few steps and it starts to look a lot like the thing employees already distrust.

What Meta just walked into

The tool built to watch employees couldn't keep what it captured. That's not a Meta problem. That's every collected record's problem.

In April, Meta rolled out a program called the Model Capability Initiative. It captured how US employees used their computers to help train AI agents: keystrokes, mouse movements, click locations, screen content, and according to internal documents, the contents of emails and direct messages sent to US staff. More than 1,500 workers signed a petition calling it an "Employee Data Extraction Factory".

Meta's first response, in early June, was to soften it. Employees could pause collection for up to 30 minutes at a time and request an exemption. Note what that was not. It was not "we will stop watching." It was "you can ask us to stop watching, sometimes, for half an hour."

Then the watching backfired. On June 18, the data the tool had been harvesting, including private conversations, performance information, and meeting transcriptions, was accidentally exposed to the entire company. Meta logged it as a SEV 2 incident, patched it within four hours, and the patch did not hold. Only after that leak became public did Meta pause the whole program.

Sit with that narrative.

The petition did not stop it.

The trust damage did not stop it.

What stopped it was the surveillance data leaking to everyone, proving the company could not keep safe the thing it insisted on collecting. The pause is a reaction to a breach, not a change of heart. The program is suspended, not killed.

That sequence matters for communicators, because the softened version Meta tried first is exactly where a lot of "employee listening" already lives. We do not collect keystrokes. We collect who opened the message, who finished it, who came back twice, who never clicked. We tell ourselves it is engagement data. The employee on the other end did not get a pause button either. And every record we keep is a record that can leak.

The trust math is completed

This is not a hypothetical reaction. Gartner predicts that by 2028, 40% of large enterprises will use AI to monitor and shape employees' mood and behavior, scoring the tone of messages in tools like Slack and email. The trust cost lands not when the monitoring starts, but when employees find out it has been happening all along.

And the data rarely stays in the building. A Northeastern University study examined nine workplace monitoring platforms and found all nine shared worker information with outside parties, including Google, Facebook, and Microsoft. Information from those tools reached more than 145 domains tied to advertising and tech companies. The pitch for most of these tools is insight. The byproduct is exposure.

So when a vendor or a trend report tells you to instrument your comms for richer sentiment data, the honest follow-up question is not "can we?" It is "what happens to this data, who else sees it, and what do we tell employees about it?"

Why this hits internal comms specifically

Here is the trap. The "listen more" mandate gets handed to internal comms because listening sounds like our job. We are the empathy department. We are supposed to know how people feel.

But the tools that promise to tell us how people feel are monitoring tools. Sentiment scoring, message-level engagement tracking, AI that flags frustration in internal channels. When IC adopts them, we become the team holding the surveillance apparatus, whether or not that was the intent.

That is a real shift in what our function is. We can be the people who help employees feel heard, or we can be the people quietly measuring them. Those are not the same role, and employees can tell the difference faster than we can.

The reframe most use to make this comfortable is the word "listening." It implies consent. It implies a conversation. But listening is something people opt into. Monitoring is something done to them. If an employee would be unsettled to learn exactly what you track and where it goes, you are not listening. You are surveilling with a friendlier label.

Where the line is drawn

You do not need to throw out measurement to stay on the right side of this. You need to be able to pass one test: would your employees be fine with it if they knew the full picture? If the answer is no, the metric is not worth the trust it costs.

A few things that keep you on the listening side of the line:

• Tell people what you measure, before they find out some other way. The trust damage comes from discovery, not collection. A short, plain explanation of what your platform tracks and why removes the ambush.

• Measure the message, not the person. "Did this announcement land across the org" is a comms question. "Who specifically is disengaging" is a monitoring question. The first improves your work. The second builds a file on people.

• Ask before you infer. A question an employee chooses to answer beats a behavior you captured without telling them. The survey you mock as old-fashioned at least had the decency to ask.

• Know where the data goes. If your tool ships engagement data to third parties, that is not a footnote. That is the story, and you own explaining it.

• Keep a pause button that means something. Meta's 30-minute version looked like a fig leaf, and it took a full-blown data leak to actually halt the program. If employees have no real way to opt out of being measured, you do not have consent. You have compliance, until something breaks.

The trend reports are right that the annual survey is a weak instrument. They are wrong to assume the fix is more collection. The fix includes better questions asked in the open, not quieter data gathered in the background.

Internal comms spent years arguing it should be “at the table” because it understands employees.

It would be a strange ending to earn that seat and use it to watch them.

Written by Chuck Gose, founder of ICology.