AI is sold on saving time. Whose time it saves, and what that time gets spent on, is settled before anyone opens the tool, in the defaults someone wrote for it. If we want those defaults to be fair to women, the answer isn’t to sort users by gender. It is to build around the individual, on a backbone skill that knows the pressure on people’s time is uneven.
That’s where Paolo and I ended up after this morning’s techUK breakfast, Closing the gap: Women, work & AI adoption. It opened with Rachel Seftel, Head of Research and Impact at the Cherie Blair Foundation for Women, in conversation with Sheila Flavell, President of techUK. That conversation gave us the phrase and the idea we kept coming back to: time poverty.
What we heard
For their report, Adopted, not embedded, Rachel’s team surveyed more than 3,000 women entrepreneurs in low- and middle-income countries. Most now use AI, and the benefit they mention most is time. But they start with less of it, spending around 18 hours a week on caregiving, and the time AI frees up tends to go back into existing demands rather than into growing the business.
We both felt that a lot of the barriers discussed during the panel would be familiar to anyone trying to get a team using new tools, men included. Time, however, was the exception.
The discussion afterwards
Every time we work with a company, we write what we call a backbone skill. It is a set of instructions that tells every agent we build who the organisation is, how it works, and what it cares about. Every agent built for that client refers back to it.
So we asked a simple question: should the backbone skill be feminist? After some internal debate, we agreed that it doesn’t need to know anyone’s gender. It should understand the inequality and bring that understanding to everyone it works with. The uneven split of work at home isn’t only a women’s issue; it is also the man who isn’t doing his share.
Should the backbone skill be feminist?
We have started developing profiles inside our own platform: a running picture of what each person is working on, which the system uses to decide what is worth telling them. For now the profile is built mostly from your work. It could just as easily include how you want to spend your time. The school run. No calls after four. Friday afternoon kept free. The gym before work.
Treating all women the same would be as unfair as treating all men the same. Some people have children and some don’t. Some fathers would rather do the school pickup than stay late. A profile built around the person makes room for all of that without asking anyone to declare a category.
The Adopted, not embedded report points the same way. The gains go to those with the time and support to build AI into how their work actually runs, and its recommendations include cutting the time and mental effort that using AI takes in the first place. A tool that hands a time-poor person one more thing to read or react to has missed the point. One that takes something off their plate, and knows what their week looks like, hasn’t.
Where the backbone skill comes in is the defaults. In the piece Paolo wrote with John Zealley last week, they argue that the value of a good system lies less in what it tells you than in “the hundred things you are not told”. Deciding what to leave out is a judgement. Every judgement carries the assumptions of whoever designed it.
A feminist backbone skill, then, is one that knows the pressure on people’s time is uneven. When it builds a picture of how someone wants to work, the starting point isn’t one person’s idea of normal.
Where we’ve got to
This is an idea rather than a finding: our immediate response to a morning that touched on exactly what we’re working through at Activate Intelligence. What should an agent understand about the people it works for, and whose assumptions does it carry into everything it does?
Time is where our conversation started, but a feminist backbone skill is about more than time. It gives every agent a feminist lens: an awareness of the inequalities that shape how people work, who carries the load at home, who gets taken seriously, and whose time gets treated as spare. Each person’s profile is then read through that lens. It isn’t a category anyone gets put in. It’s the context the agent brings to everyone.
It’s early, and it’s the question we’re taking back into our work.
With thanks to techUK and the speakers: Rachel Seftel (Cherie Blair Foundation for Women), Sheila Flavell CBE (techUK and FDM Group), Sue Daley OBE (techUK), Maryam Yaqub (Fawcett Society), Riham Satti (MeVitae), and Cat Pinchard (Lloyds Banking Group).