For years, a healthy engagement score was the closest thing HR leaders had to proof that a company was working. That proof is getting weaker, because the score only measures half of engagement.
Across more than 11 million employee survey responses we have collected at CultureMonkey, median employee Net Promoter Score (eNPS) fell from +39 in 2022 to +28 in 2026. Over the same years, engagement scores barely moved.
People still like their work. Fewer of them would recommend their employer, and two numbers that are supposed to move together have come apart.
I believe AI is the biggest reason, and the timing backs it up: advocacy peaked in 2022, the year generative AI went mainstream, and hasn't recovered since. AI has changed how work gets done, and the way most companies measure their people hasn't caught up.
Engagement is whether what people care about can turn into work that counts. How people feel is only the first half.
The score most companies run measures how people feel and goes no further. What it leaves out is org intelligence: how well an organization notices what's broken, gets that signal to the people who can fix it, then decides and learns, measured through what employees say.
That is engagement. It is the half we stopped asking about.
Engagement surveys earned their place. They gave HR a shared language for morale and a reason to invest in managers, recognition and growth. I'm not arguing for scrapping them.
They were built for a time when the biggest risk was people checking out. Today, the bigger risk is an organization that can't keep pace with people who are ready to move faster.
Most CHROs have seen the result: engagement holds steady, action plans get written, and six months later the same frustrations show up in exit interviews. The score held because nothing it asked about had changed.
The Half We Measure Looks Fine
We then looked inside five organizations that run their employee surveys with us, one each in automotive AI software, insurance brokerage, construction, aviation cargo handling and food and beverages.
Each scored within 0.35 points of its industry benchmark on engagement. Each trailed its industry on eNPS, by 4 to 20 points.
A leader looking only at engagement would see five companies on par with their peers.
The wider benchmark shows where the gap sits. Of the 22 workplace drivers we track, communication has the strongest association with eNPS, yet it ranks 16th on score.
In all five organizations, communication scored below its industry benchmark. It was the only driver that did. Of all 22, it's the one that measures how the organization works.
Employees Already Tell You Which Half Is Broken
We analyzed more than 13,000 open-text comments that employees at those five organizations wrote in their surveys.
When employees described what's good about their jobs, 37% of comments named engagement themes, such as belonging, flexibility or growth. Just 9% named how the organization works.
When asked what they'd change, the pattern flipped: 32% named how the organization works and 24% named engagement themes. Employees praise one half and ask you to repair the other.
The comments split the same way. Employees write warmly about the people they work with, and the frustrations they name are about how the organization runs: information that arrives late or not at all, tools that slow the work down, and processes that get in the way.
One employee, asked what they enjoy most, captured both halves: "I work with an amazing team in spite of the frustrations we feel with so many changes and limitations."
Information flow alone came up in 15% of change comments. Pay came up in 9%. Pay still matters; rewards is the lowest-scoring driver in 11 of the 14 industries we track. It's no longer the only thing employees ask for.
AI shows up most clearly where it's closest to the work. At the AI software company, comments about AI tools on the equipment question rose from 16% to 29% in one year, while the score on that question edged up from 7.8 to 7.9.
The survey said things were improving. The comments said a new gap was opening.
These findings come from five organizations and one benchmark, and our comment tagging remains directional. Survey data also can't fully separate AI from layoffs, return-to-office mandates and slower hiring.
Still, the pattern held across industries that have little else in common.
Four Questions That Measure Org Intelligence
Your engagement survey stays as it is. Add the questions it was never built to ask, one for each step. They belong on the same score sheet.
1. Notice: Ask where work gets stuck. Add one open question, such as "What slows you down?" Report the share of answers about how the organization works next to your engagement score. Test statement: "When something goes wrong in my work, it gets raised quickly."
2. Route: Treat communication as a system. Measure whether information reaches people in time, across teams. Test statement: "I get the information I need from other teams, when I need it."
3. Decide: Make ownership visible. One construction company we studied rebuilt its survey around simplicity and ownership. Its two lowest scores showed leaders exactly where decisions stall. Test statement: "It's clear who makes the decisions that affect my work."
4. Learn: Spread what works. If good ideas stay inside one team, the organization keeps relearning the same lessons. Test statement: "When a team finds a better way to work, the rest of the company hears about it."
Give each step an owner, so when a score drops, someone owns the fix. And if you're asking people to work faster with AI, give them the tools before you ask for the gains.
For years, a healthy engagement score was proof enough. It isn't anymore. The employee who loves the team and fights the system is telling you exactly where to look.
Org intelligence is the half of engagement most surveys skip. Measure how people feel, and measure whether that feeling reaches anything. Keep the survey. Change the questions.
