Good enough

Most of the prose produced in most organisations is, at best, mediocre.

Business cases, status update reports, performance reviews, job descriptions, policies and procedures, terms and conditions, onboarding documentation, strategy documents, change request forms, tender responses… the list goes on.

Bilge, for the most part, the lot of it. Documents over which hours of effort have been poured, knowing that the chance of any of it being read is low. Written by people who generally aren’t professional writers, they often end up as cut-and-paste jobs from multiple sources – Frankenstein’s monsters of corporate prose.

There is a narrative today that suggests generative AI tools aren’t good enough because they produce generic, only superficially competent text. That’s a fair criticism, as far as it goes. But it assumes that the alternative – human-written workplace documents – represents some higher standard worth protecting. That perspective can only have been gained by never having to read the things.

These documents come from a particular moment in organisational history. There was a time when professionals existed specifically to take others’ ideas and craft them into coherent prose. The typing pool. The secretary. Roles whose entire purpose was to make organisations legible to themselves and others. Then came personal computers, word processors, networked file storage – and with them, the assumption that everyone could and should do it themselves. Most people couldn’t, particularly. But they had to anyway.

And so we arrive at a familiar pattern: a new technology emerging to clean up the mess left by the previous one. LLMs are, it turns out, reasonably well suited to exactly this kind of work. Assembling disparate inputs into something coherent and readable. Doing it quickly. Producing output that, while unlikely to trouble any literary prizes, is probably better than what it replaces.

That last point is worth sitting with. If LLMs take over the writing of documents that few people read, written by people who’d rather be doing something else, producing text that serves a bureaucratic function more than a communicative one – what exactly have we lost? And more usefully: what might we do instead with the time and effort that work has been consuming?

That question doesn’t have an obvious answer yet. But it’s a more interesting one than whether AI writing is good enough. For most workplace documents, good enough has always been the ceiling, not the floor.