5 Things We'll Never Ask Claude to Do
Earlier this year, we published our Humanifesto. It sets out four commitments that guide how we use AI at Within People: Be More Human, Create More Freedom, Enable More Impact, Be Ever Mindful.
We believe the evolution of work is super human, and when people no longer need to act like machines, work gets to be more human than ever. While AI can expand what's possible, it doesn't change what we believe.
Under Be More Human, we made three promises: we won't use AI as a standalone coach, we won't let it author our visions or values or strategy, and we won't delegate decisions and accountability to it.
There's a difference between what a tool can do and what we're willing to hand over. Claude could attempt most of what's on this list, and likely do it well. That's exactly why the line matters. It's not a workaround for something AI can't manage. It's a choice we're making on purpose.
"AI can't" is a claim about capability, and capability keeps expanding, so that kind of claim ages badly and starts to sound defensive. "We won't" is a claim about values. It holds regardless of how good the tools get, because a values statement is about who we choose to be, not what's possible. Years ago, we defined what human-centred leadership actually looks like. What follows draws on that same thinking.
1. Decide what we stand for
Values are a choice about who we're willing to be when it costs us something, not a research output. Claude can lay out every option and make the case for each one. What it can't do is hold a conviction, because it has nothing at stake. Our Humanifesto exists because we sat in a room, argued it out, and decided. That part was never delegable.
2. Say what we don't know
Ask Claude a question and it answers in the same confident register every time, whether it's certain or guessing. That's not a flaw we can coach out of it. Admitting "I don't know" or "I got that wrong" is one of the harder things a leader does, and one of the most trust-building. Vulnerability is the willingness to be seen without the armour of certainty, especially in front of the people counting on you to have the answer. Call it a soft skill and you'll miss what it actually costs.
3. Know what to say, and what to hold
Claude will answer any question put to it, fully, immediately, with the same confidence regardless of what's actually at stake. It doesn't weigh whether now is the moment, whether this person is ready to hear it, or what saying this out loud will cost the relationship. That weighing is the job. It takes courage to name the thing that needs naming, and the judgement to know when. Neither of those live in a transcript or prompt response.
4. Read what isn't being said
Our whole craft is catching the thing under the thing. The shift in energy when a group hits the real issue. The silence that means something. Our Six Steps facilitation framework works because a human is feeling the room in real time and choosing the next move from what they sense, not from a script. Claude can flag tone. It can't sit inside what's underneath it. Coach, don't play. You have to be there to coach.
5. Stay invested
Clients don't come back because we were efficient. They come back because someone stayed invested after the deliverable shipped, checked in when nothing was due, remembered what actually mattered to them. Claude can help us prep faster, draft sharper, clear the busywork. It can't be the reason someone trusts us enough to call again. That part is built in small human moments no model will ever be in the room for, and we're not handing it over.
What won't your team hand to AI?
We're building a short checklist to help you define your own answer, the specific things your team won't hand to AI. Subscribe to our newsletter and we'll send it your way.
