One of the most surprising things about running structured skills conversations - whether human-led or AI-supported - is what people share. When you give someone unhurried space to talk about their work, what they enjoy, what they are good at, and what they would like to do next, they often share things that are not strictly about skills at all.

Sometimes that is positive. The hidden hobby. The unexpected qualification. The volunteer work that reveals a whole side of someone their colleagues never see.

Sometimes it is not. Stress. Burnout. Concerns about a manager. A welfare issue they have not raised because there has never been a sensible place to raise it.

This second category has to be handled with extreme care - and when it is, it becomes one of the most valuable things about doing skills conversations properly.

Why concerns get raised in skills conversations

The reason is simple. A skills conversation is usually one of the few times a year when an employee gets dedicated time to reflect on their work, in a relatively low-stakes setting. They are not being assessed. They are not being asked to defend anything. The space invites honesty.

And often, what comes out is what has been on their mind. The colleague situation that has been wearing them down. The workload that has been unsustainable. The capability they have not had the chance to use because of a difficult dynamic.

This is not a bug in the process. It is one of its quiet superpowers - if it is handled properly.

What "handled properly" looks like

The conversation has to be safe

People will only raise concerns if they trust what happens next. That means the data has to be encrypted, the access has to be controlled, and the route forward has to be clear. If raising a concern feels like it will lead to gossip, retaliation, or being labelled "difficult", it will never get raised again.

The right people need to see it

A welfare concern flagged in a skills conversation is useless if it disappears into a system. It needs to reach a designated, trained safeguarding lead who knows what to do with it. Often this is HR, sometimes it is occupational health, sometimes it is a specifically trained welfare officer.

The response has to be human

Whatever flagged the concern - human or AI - should never be the one to act on it. The right response is always a real person reaching out to the employee in an appropriate way, on an appropriate timescale, with appropriate confidentiality.

The employee should know what happens

If someone shared a concern, they should be told what was done with that information, even if the answer is "we noted it and will check in next month". Silence breeds anxiety. Transparency builds trust.

What to look for

The welfare signals that emerge in skills conversations are often quiet:

  • Mentions of being "drained", "overwhelmed", or "always behind"
  • Talking about their work in past tense ("I used to enjoy...")
  • Mentions of conflict or difficult relationships at work
  • Reluctance to discuss their current role at all
  • Mentions of stress, anxiety, or sleep problems
  • Long descriptions of administrative burden
  • Mentions of caring responsibilities or personal challenges affecting work

None of these are necessarily concerns. All of them might be. The point is to notice and check in - not to diagnose.

Why AI-supported conversations can be especially valuable here

People sometimes share things with an AI that they would not share with a human. Not because the AI is more trusted, but because it is less judgmental. There is no risk of a colleague's reaction, no future eye contact in the office, no worry about how the conversation will be remembered.

This is a double-edged thing, and we approach it carefully. But when handled well - with proper consent, secure data handling, and a clear escalation path to trained humans - it can surface concerns that would never have come up in a manager 1:1.

The principle

Skills conversations should be conducted with the assumption that they will sometimes reveal things that are not skills. The tools, processes and people around the conversation need to be ready for that. Done well, this transforms a skills system from a workforce planning tool into something that genuinely cares for the people in it.

Done badly, it makes people feel surveilled. There is no shortcut. The infrastructure of trust matters enormously.