Can AI write your agency's thought leadership?

It can draft it. It cannot supply the thing that makes it publishable - a named person's specific position, drawn from work only your agency has done, that a reasonable reader could disagree with.

That distinction decides whether the output is worth running. Everything else about the question is detail.

What it genuinely does well

Being honest about this matters, because agencies that pretend otherwise lose credibility with their own teams.

A current model will structure an argument you have already made, cut a 1,400-word draft to 900 without losing the spine, generate six headline options where you would have managed two, and catch the paragraph where you changed position halfway through. It is very good at the second draft and the fourth. It is fast at the mechanical passes that used to eat an afternoon.

Used that way it raises the floor of what an agency can publish, and it lets a small team keep a cadence that would previously have needed a hire. That is real, and the agencies refusing to touch it are paying for the refusal.

The three things it cannot supply

A position. A model produces the consensus of everything written on a topic, which is by construction the average view. Thought leadership earns attention by departing from the average view in a specific direction. Asking a system optimised for the middle to generate the edge is asking the wrong tool.

Proprietary evidence. The reason an editor runs an agency byline is that the agency knows something the editor's readers do not - what happened in eleven pitches last year, why a category's buyers changed their brief language in March, what the client actually said when the work tested badly. None of that is in the training data. It is in your people, mostly undocumented, and getting it out of them is the work.

Accountability. A byline is a named human staking their reputation on a claim. That is what makes it worth reading and what makes it quotable. No model can be accountable for a sentence, which is why the ones it writes unaided read as though nobody would mind if they were wrong.

The test an editor applies

Trade editors do not screen for whether a piece was drafted with AI. They screen for something that catches most AI-drafted pieces anyway.

Read your opening two paragraphs and ask whether any of your competitors could have published them without changing a word. If they could, it is not thought leadership - it is category summary, and every publication already has more of that than it can run. Getting into Campaign was never about writing quality. It is about whether the argument belongs to you.

The question is not whether a machine wrote it, but whether anyone decided it.

The tell-tales of a piece nobody decided

These are the patterns that get bylines rejected and LinkedIn posts ignored. They are worth knowing because they are easier to spot in your own drafts than to design out in advance.

The piece surveys a debate rather than taking a side in it, usually closing on the observation that the truth lies somewhere in the middle. The byline belongs to someone who could not defend the argument in a meeting, because they did not make it. Every claim is true and none is specific - the numbers are industry-wide rather than yours. The position drifts between paragraphs, because no single person held the whole thing in their head. And the piece could run under a competitor's logo with a find-and-replace.

Any one of those is survivable. Three together and the piece is doing reputational damage, because the readers you want are precisely the ones who notice.

This is getting harder, not easier

The volume of competent, unremarkable content went up sharply when drafting got cheap, and the filters tightened in response. LinkedIn now gives readers a button to report exactly this kind of post. Editors who used to read the first three paragraphs now read the first two. The engines that cite sources in AI answers reward a clearly stated position and skip a hedge.

So the cost of publishing something forgettable has risen. It is no longer neutral. It teaches the people you want as clients that your agency does not have a view, which is a harder impression to reverse than having published nothing.

What this means for how you staff it

The useful reframe is that AI changed which part of the process is scarce, not how much the process costs. Drafting was most of the visible effort and is now close to free. Deciding what the agency thinks was always the bottleneck and is now the entire job.

That points somewhere uncomfortable for agencies who have cut writing roles on the assumption that the tool covers it. The tool covers the part that was never the constraint. What it does not do is sit with a founder for ninety minutes and extract the four opinions they hold strongly and have never written down, then work out which one a trade editor will run this quarter.

That extraction is a craft, it is mostly interviewing, and it is the reason a thought-leadership programme costs what it costs.

Use the model for the drafts. Keep a named human responsible for the argument. If the argument is the part your agency is struggling to pin down, that is the conversation to have.