Burnout Predictor Identifies Staff Too Exhausted to Resist More Work

We commissioned the burnout predictor because we wanted to identify colleagues who needed help before they had to ask for it. At the demonstration, the system identified them. Then somebody from staffing asked whether we could sort the list.

Hereish, our wellbeing platform, had assessed late-night activity, cancelled leave, weekend logins and the increasingly brief answers people gave when asked how things were going. It produced a red indicator beside an associate whose last three replies had been ‘Fine’, ‘Fine’ and a thumbs-up.

‘This person needs an intervention,’ the people director said.

‘How quickly do they usually accept an assignment?’ asked the head of a practice with an intervention of its own.

The product demonstrator checked. Eleven seconds.

There was a different quality of attention in the room after that.

By the second demonstration, the dashboard included a filter called Likely to say yes. It correlated long working hours, repeated interruptions and a declining tendency to negotiate deadlines. The red indicators remained. Their meaning had become more encouraging.

I asked the implementation lead why a prevention tool was connected to the allocation system.

‘We needed to demonstrate business value,’ she explained. ‘The original presentation only demonstrated a problem.’

The revised workflow was impressively joined up. When a lawyer crossed the threshold for sustained overload, Hereish notified the people team, offered the lawyer a breathing exercise and alerted three partners to the availability of somebody unlikely to query a Friday deadline.

No data had to be entered twice.

Our policy says wellbeing information must be used to support employees. The project team therefore added a field requiring every new allocation to specify its supportive purpose. The most popular selection is Maintaining a sense of professional belonging.

The people director proposed a simpler intervention: stop assigning additional work to those the model had identified as overloaded. This was recorded as a potential phase-two enhancement, dependent on commercial feasibility.

We have also discussed consent. Employees originally agreed to use the platform because they understood it would help the firm respond to excessive workloads. The consent screen did not anticipate a disagreement about what ‘respond’ would mean. Legal has asked for that word to remain.

The model correctly identified vulnerability. The business selected a different use case.

Yesterday, the associate with the three brief replies received another matter, a notification congratulating her on being a valued team member and a request to complete our wellbeing survey.

She accepted all three in nine seconds. The dashboard says the pilot is improving.