AI and discipline.
Ten posts. The same argument, made several different ways.
Every AI failure I have cleaned up over the last two years had the same root cause. Not a bad model, not a bad prompt, not the wrong platform. A weak operating posture that the AI amplified. Vague segmentation became a thousand off-target sends. An incoherent brand voice became a thousand off-voice posts. A reporting framework with three known problems became a dashboard with three thousand. The model did exactly what models do - it took the existing posture and multiplied it.
I have stated the underlying argument several different ways across the years and the writing. AI is a partner, not a leader. AI is the fuel, not the driver. AI works fast but not well. AI is the tool, not the craftsman. AI memory is trash, and AI cannot be held accountable for what it forgets. Pick whichever frame is useful for the conversation you are in. Each is true. None of them is the line. What sits underneath all of them is the same: the driver is still a marketer who knows what good looks like, owns the lifecycle the AI is touching, and is willing to delete the output when the output is wrong. That is not a limitation of current technology. It is the nature of the work.
The posts below were written over about four months, in the order they needed to be written. The reading order below is different - it is the order I would give them to a CMO who is about to make a decision about where AI belongs in their stack. Start at the top.
Recommended reading order
- Start here. The plainest version of the argument.
AI is your accelerant, not your driver
Most "AI marketing" failures are not AI failures. They are operating-discipline failures that an LLM amplified at scale, and the cleanup costs more than the work it replaced.
April 22, 2026 - Understand the difference between a retrofit and a rebuild before you buy anything.
What “AI‑native” actually means in a marketing ops stack
Most stacks calling themselves “AI‑native” are the same stacks they were eighteen months ago with a chat box bolted on. The label is doing more work than the architecture is.
March 11, 2026 - A practical framework for where AI belongs - and where it does not.
When to trust AI in your marketing ops, and when not to
There is no single right answer for whether to put AI inside a marketing operation. The right answer depends on whether the work is reversible, observable, and bounded.
March 25, 2026 - The missing role most teams do not know they need.
Why your AI workflow needs an editor
The cheapest, highest-leverage role you can add to an AI-heavy marketing team is not another generator. It is a senior editor with the authority to delete.
April 1, 2026 - What happens when you skip the review step.
The cost of a vibe‑coded automation
A workflow generated in twenty minutes by a confident LLM and a marketer in a hurry will eventually meet a quarterly review. The cleanup is almost always more expensive than the original build would have been.
April 15, 2026 - Where the asymmetry has moved, now that content volume is table stakes.
Real‑time marketing in an AI‑saturated market
When everyone can generate a campaign in twenty minutes, the differentiator is no longer the campaign. It is the response time on the things the campaign produces.
April 8, 2026 - The broader case for why this position exists at all.
Discipline is the product
In a market where every agency is selling AI capabilities, the firms that will still be standing in five years are the ones that sold operating discipline the whole time.
March 18, 2026 - The foundation that AI touches - and that AI cannot fix if it is already broken.
Why your CRM is your real product
Whatever the company sells, the CRM is the system the company actually runs on. Treating it like a long‑lived product instead of a back‑office tool is the difference between a marketing operation that compounds and one that resets every two years.
February 18, 2026 - Applied rigor before the next purchase decision.
The audit we run before recommending any new tool
Before we recommend buying anything, we run the same short audit on what the team already owns. More often than not the new tool is not the answer, and the audit pays for itself before the renewal.
March 4, 2026 - The lifecycle work that predates AI and will outlast it.
The extra mile: why the leads you gave up on are still deciding
Most sales follow-up stops at 60 days. Roughly half of all closed B2B business comes from leads that took 3–6 months to decide. The lifecycle stage that spans this window is never crowded, because everyone else gave up at the same time you did.
February 25, 2026
The operating discipline is the differentiator.
AI compresses the time between intent and execution. A disciplined operation ships faster; an undisciplined one fails faster. The firms still standing in five years will be the ones that built the discipline first and layered AI on top of it. The other order does not work.