Competence Is No Longer the Differentiator: Building a Publishing System Around My Own Work

The uncomfortable observation

For most of my career I believed the work would speak for itself. Deliver well, and the right people notice.

That was never quite true, and around the start of this year it stopped being even approximately true. The reason is not mysterious. When a capable model can produce a migration plan, a RAID log, a sprint report or a stakeholder email in thirty seconds, the market value of being able to produce those things falls. Not to zero, but far enough that it no longer separates you from anyone.

What does not fall is the value of judgement that has survived contact with reality: knowing which of those three migration plans will actually get compliance sign-off, and why the other two die in week four. That knowledge is genuinely scarce.

The problem is that scarce knowledge is invisible by default. Mine lived in Jira tickets behind a corporate SSO, in status decks that went to eleven people, and in meetings that left no trace. Ten years of it. Somebody evaluating me could see a CV claiming judgement, which is exactly what every other CV claims.

So the gap was not competence. It was that competence had no public surface.

What I was actually solving

Not “build a personal brand”. I find that phrase slightly embarrassing, and it describes an outcome rather than a problem.

The problem was mechanical. I had roughly forty things I genuinely knew and could defend — why pure scrum fails in multi-tenant SaaS, how to make a client-facing ticket that stops scope creep, what agentic AI actually means for delivery. Having the view took no time at all. Turning the view into 1,500 publishable words took four to six hours, and I do not have four to six spare hours in a week that already contains a delivery job.

So the maths of the honest version was: forty good ideas, one publishable article a month at best, and a three-year runway to say what I already knew. By which point half of it would be stale.

The bottleneck was never the thinking. It was the distance between having a position and having it written down well enough to put my name on.

The system, in plain terms

Four stages, and only one of them is automated.

1. The source is always work I actually did. Every article traces back to a real delivery problem, a migration, an escalation, a reporting format that failed and the one that replaced it. This is the constraint that makes everything else defensible. I do not write about topics; I write about things that happened to me.

2. A generated base draft. Since April 2026 I have started most articles from a draft produced in iBeVisible, the content platform I have access to through work. I give it the position I want to argue and the angle, and it returns a structured piece with the surrounding context filled in, the definitions, the standard counter-arguments, the section scaffolding, and the search terms real people actually use for the topic, which is consistently not the phrasing I would have picked myself.

That last part turned out to matter more than the drafting. I would have written “utilisation reporting”; the words people search are closer to “why my team looks busy but delivers nothing”. Same article, entirely different reach.

3. The rewrite, which is not optional. Roughly a third of every draft gets replaced before it goes out. That third is the whole point of publishing. A generated draft can tell you that stakeholder sign-off is a common migration risk. It cannot tell you that we cut sign-off from weeks to days by pre-briefing regional compliance leads as a group and handing them automated content-diff reports, because that happened to me and exists nowhere in any training corpus. Every number, every named failure, every “I would do this differently now” is typed by hand.

I also cut. Generated prose reaches for balance, on the one hand, on the other hand and balance is what makes writing forgettable. If a paragraph does not commit to something, it goes.

4. Publish to my own site first, distribute second. The article lands here, at a URL I control, and only then goes out to the professional platforms. That ordering is deliberate and I will come back to it.

The decisions I would defend

Own the destination. Publishing exclusively on someone else’s platform means renting your credibility from a company that can change its ranking algorithm, its reach, or its entire business model without telling you. The canonical version of everything I write lives on a domain I pay for. Social platforms get a copy and a link, and they are a distribution channel, not an archive.

Cadence beats quality, up to a point. Weekly for eight months, longest gap eight days. Not because every one of those 33 pieces is excellent, they are not but because a body of work reads differently from a scatter of posts. Thirty-three articles on adjacent problems demonstrate a coherent point of view in a way that three brilliant ones cannot. Consistency is itself the evidence.

Automate the slow part, never the credible part. The tool compressed drafting from hours to minutes. It did not, and could not, compress the part where I decide what I believe. I have seen people automate that step too, and the result reads exactly like what it is: confident, fluent, and about nothing. The failure mode of AI writing is not bad prose. It is that the prose is fine and the person has vanished from it.

Write for the machines that answer questions, not just the ones that rank pages. This is the part most professionals have not adjusted to yet. Increasingly the first thing that reads your work is not a person and not a search crawler, it is an assistant, summarising an answer for someone who will never see your page. That changes what publishing has to do. The site emits structured metadata, a clean feed, and an llms.txt describing what is here, because content that cannot be parsed cannot be cited. It is the same instinct that governs stakeholder management in regulated industries: make the reader’s job trivial and you get a decision instead of a delay.

Say what you would not say in a meeting. The most-read things I have published are the ones where I described something that went badly. Nobody remembers a summary of best practice. They remember the person who explained the specific way their process broke.

What actually changed

The honest answer is that the outcomes are early and compounding, not dramatic.

The measurable shift is in discovery. My work is now indexed against the specific problems I solve, and more usefully, AI assistants surface and cite it when people ask about those problems. That is a genuinely new distribution channel, and it rewards depth and structure rather than volume and recency. Being the written-down authority on a narrow set of delivery problems turns out to be a far better position than being loud about a broad one.

The second shift is internal, and I did not anticipate it. Writing publicly about how I run delivery changed how I am treated inside my own organisation. Colleagues cite the articles back to me. A position I have argued in public is a position people have already read before the meeting starts, which means the meeting starts further along. The brand did not just open external doors; it raised the resolution of my internal one.

The third is harder to quantify and probably matters most. Committing to defend a position in public every week forces you to actually hold positions. I understand my own work better than I did in December, because writing is where you find out whether you know something.

What I would tell someone starting

• Being good is now the entry fee, not the differentiator. AI has commoditised competent execution. The remaining scarce thing is judgement you can prove you own, and proof requires a public record.
• Your experience is the moat; everything else is scaffolding. Automate the scaffolding without hesitation. The moment you automate the experience, you have published nothing.
• Publish on ground you own. Distribution is rented. The archive should not be.
• A body of work outperforms a viral post. Thirty-three adjacent articles say something no single article can, and they keep saying it while you sleep.
• Write to be cited by machines as well as read by people. The discovery layer changed. Structure your work so it can be quoted accurately by something that is answering on your behalf.
• The point is not visibility. It is that the people who would benefit from what you know currently cannot find you, and that is a solvable, mechanical problem.