Why being pro-AI is making me more humanist

The more I use AI, the more humanist my view of work seems to become. That sounds contradictory only if being pro-AI means wanting to automate as much human activity as possible. I don’t think it does. I’m increasingly interested in AI precisely because it forces a harder question: once more of the mechanical layer of knowledge work becomes cheap, what should people actually spend their time doing?

Most arguments about AI still push you towards a crude choice. Either the technology is transformative and we should push it as far as possible, or we defend the things humans already do against automation. I don’t find either position very useful. I’m very pro-AI, but much of what I’ve written over the past few months has really been about protecting the human part of work rather than protecting every task humans currently perform.

That distinction matters because AI can already do a great deal of competent production. It can propose structures, retrieve information, generate alternatives, analyse patterns, search old work, compare evidence and produce drafts. Pretending otherwise doesn’t help writers, marketers or anyone else whose job includes those activities. Nor am I sentimental about preserving repetitive work simply because a person used to be paid to do it.

What interests me is what becomes more important as that layer gets cheaper.

The scarce layer is judgement

A lot of my recent work has ended up circling the same point from different directions. In AI can make thinking less linear, I argued that the useful boundary isn’t between a supposedly pure human idea and an AI-assisted one. AI can participate in ideation by surfacing contradictions, proposing alternative structures, retrieving old thinking and showing connections I would not have reached quickly on my own. The boundary I care about comes later: I still have to decide whether the connection is meaningful, whether the evidence supports it and whether I am prepared to put my name behind the conclusion.

That is why I keep coming back to judgement. AI can expand the possibility space, but it cannot outsource responsibility for what you choose from it. It can help search for what might be worth believing without deciding what you believe.

The same pattern appears in AI didn’t replace editorial judgement. It made it more valuable. Once searching, sorting, comparing and drafting become easier, the bottleneck moves. The difficult part is no longer producing enough plausible options; it is deciding which one deserves to survive. More AI does not necessarily mean fewer editorial decisions. In my experience it often means more of them, because the cost of generating alternatives has collapsed.

That changes the economics of knowledge work. The value moves away from mechanical competence and towards significance: knowing what matters, what is supported, what connects to something else and what should be ignored.

I don’t think we should romanticise the loss

There is an uncomfortable side to this argument. A lot of people built careers around skills that AI is making cheaper, and those skills were real. They took years to learn and, for many people, were the job rather than an incidental part of it.

I wrote about this more personally in Being an ageing content writer may have made me comfortable with AI. I have spent a long time writing professionally, so when AI makes first drafts easier I do not experience it as the removal of something I still needed to prove to myself. My satisfaction had already moved upstream towards finding the angle, deciding what mattered and working out whether a piece should exist at all.

That makes me unusually comfortable with a change that may feel much harsher to someone earlier in their career, and that bias matters. Telling people to “move upstream” is easy when you already had years of practice doing the work that taught you how to get there.

That is the unresolved problem behind Raising the floor is not building a ladder. AI can raise the floor by making competent production widely available, but competence is not the same thing as expertise. If junior people no longer spend years doing the lower-level work through which previous generations acquired judgement, organisations need another way to build that judgement. Otherwise we risk creating a system with a very high floor and no obvious route upwards.

So a humanist position on AI cannot simply be “let the machines do the boring work and let humans do the interesting work”. That formulation skips the difficult institutional question of how people learn to do the interesting work in the first place.

Better AI should change the shape of jobs

This is also why I don’t think the most important AI story is productivity in the narrow sense. The obvious corporate use case is to make every employee produce more: more articles, more presentations, more analysis, more emails and more output passing through systems that were already full of output.

That is technically possible, but it is not especially ambitious.

In AI is making your job description less useful, I argued that AI makes it easier for people to look beyond the formal boundary of a role, understand adjacent problems and contribute in areas they do not technically own. That does not eliminate management. It makes context and direction more valuable because the employee has a larger space in which to operate.

The job description becomes less like a complete instruction set and more like an outer boundary. A manager helps explain what matters and where the constraints are. AI expands what someone is capable of doing inside that space, while the individual still needs to decide whether the work is worth doing at all.

That sounds more human to me than using AI simply to squeeze another twenty per cent of production out of the same job design.

The work moves upstream, but people still have to get there

The same shift is visible in editorial careers. In Editorial jobs are moving upstream, I argued that the valuable work is increasingly appearing before and around production: narrative development, executive thought leadership, customer evidence, competitive positioning, research synthesis and deciding what an organisation should say rather than simply producing the asset after that decision has already been made.

AI accelerates that shift because it makes the production layer easier to access. But again, there is a trap in treating this as automatic progress. Work only becomes more human if organisations actually redesign roles around those higher-value activities. If they keep the old structures and simply increase output expectations, AI does not liberate people from mechanical work; it turns mechanical work into a higher-volume target.

That distinction is becoming more important to me than whether any particular task can technically be automated.

Pro-AI does not have to mean pro-output

I used to think being enthusiastic about AI meant being enthusiastic about productivity. I am less sure now. The two can easily pull in opposite directions.

If the prize from AI is simply higher output per employee, then we have built a sophisticated way of making humans behave more like machines. We will measure the technology by how many more words, slides, campaigns, tickets and analyses one person can generate, even when nobody needed more of them.

The more interesting possibility is almost the reverse. AI could reduce the amount of human attention spent on the mechanical layer of production and make more room for interpretation, relationships, experience, judgement under uncertainty and the ability to change your mind when the evidence changes. Those things are difficult to put into a productivity dashboard, which may be precisely why they have often been squeezed out of work.

This does not require believing that humans possess some mystical quality machines can never approach. It only requires noticing where responsibility still sits. An AI system can generate a plausible answer, but a person or organisation still decides whether to act on it, publish it, trust it, reject it or accept the consequences when it is wrong.

That responsibility is not an inefficiency to automate away. It is part of what makes work human.

Let machines become better machines

The more capable AI becomes, the less interested I am in defending every task humans currently perform. Some of those tasks will disappear, some will become cheaper and some will stop being useful altogether. There is no virtue in preserving friction for its own sake.

But there is also no reason to assume that removing friction automatically produces better work or better lives. That depends on what we choose to do with the capacity we gain.

A genuinely human-centred version of AI would not ask people to compete with machines on speed, volume or consistency. It would use machines for the things machines are becoming very good at and create more space for people to exercise judgement, build expertise, understand context, form relationships and decide what matters.

That is probably why I have ended up simultaneously more enthusiastic about AI and more sceptical of productivity culture. I do not want humans to win by becoming better machines.

I want the machines to become better machines, so humans have less reason to behave like them.