Two things on your marketing team fell a quarter behind this year.
One is the internal AI workflow somebody’s been building. The thing that’s going to automate the reporting, or the briefs, or the outreach. It’s close. It’s been close for a while now.
The other is the unglamorous work. The articles that make you the obvious source on your topic. The mentions on the sites AI systems actually pull from. The reviews nobody got around to asking for.
Both are late.
Only one of them can be caught up.
And most leaders have no idea which one is which.
Hi, I’m Jeff Payne. You’re listening to The Jeff Payne Show, Episode 81: Nine Months You Can’t Backdate.
Let me tell you what’s actually happening inside marketing teams right now.
Ninety-one percent of marketing leaders say their teams use AI. That number won’t surprise anybody. Here’s the one that should stop you: two-thirds say their organization is building its own internal AI tools for marketing.
Building. Not buying. Not subscribing to. Building.
Somewhere in the last eighteen months, your marketing department quietly became a part-time software team. Nobody approved that. Nobody staffed for it. It doesn’t appear in the marketing plan, and it isn’t a line item in the budget. It just started happening, one clever shortcut at a time.
Now — is that work paying off? Ask your team, and they’ll say yes. They’ll mean it. And the research says they’ll probably be wrong.
In a survey of 3,200 leaders and employees, 85% said AI saves them between 1 and 7 hours a week. So the savings are real. That part isn’t in dispute.
Then the same study asked where those hours actually go. Roughly 37% of them go straight back out the door — correcting, verifying, and rewriting output that didn’t come back usable. Ten hours saved, four hours handed back.
And only 14% of people said they consistently come out ahead.
There’s a second leak, and this one’s worse because it doesn’t show up on the same person’s ledger. Somebody sends along work that looks finished and isn’t. In a separate study, 41% of workers said they’d been handed something like that in the past month. Each time, it cost the person receiving it just under two hours to sort out.
The sender saved 20 minutes.
On any dashboard that counts output, the sender looks great.
So here’s the first thing worth naming clearly. AI doesn’t delete work. It moves it.
It moves it from doing the task to building and maintaining the thing that does the task. Prompting, checking, correcting, patching it when a model updates and it stops working on a Tuesday.
That’s nothing. Some of it is genuinely worth doing. I build these things myself, constantly, for my own business and for clients.
But that’s not the expensive part.
Here’s the expensive part. Every one of those hours is drawn from somewhere. And it is almost never drawn from the urgent thing, because the urgent thing has a deadline and a person waiting on it.
It’s drawn from the work with the longest gap between effort and result. The work that won’t produce a visible number this month, or next month, or possibly this quarter.
Which is the authority work. Publishing enough depth on a subject that you become the obvious answer. Earning mentions on the sources these systems actually cite. Building a review record that means something. That is always the flexible budget. It’s always the thing that can wait a few weeks.
Now here’s the asymmetry, and this is the whole episode.
The workflow can be finished later. Delay it a quarter, and you have a tool that arrives in June instead of March. Scrap it entirely, and you’ve lost the hours, but nothing else. A delayed tool is just a delayed tool.
The authority work does not behave that way.
The nine months of citations you didn’t earn. The mentions that weren’t placed. The reviews that weren’t asked for while the patient or the client was still thrilled with you. Those aren’t sitting in a queue waiting for you to get back to them. That window opened, and it closed, and you weren’t in it.
You cannot backdate them. There is no sprint that recovers them. The record has a gap where those months should be, and it’s permanent.
I want to give you a concrete version of what I mean.
I’ve worked with Dr. Kamran Haghighat at Portland Perio Implant Center in Portland, Oregon, for about ten years.
In that time, we’ve published more than a hundred patient stories. One at a time. Not in a batch, not in a campaign — one at a time, over a decade. And I’ll be honest with you: almost none of them looked important when they went up. Any single one of them, on its own, is a rounding error.
Today that practice is the dominant name in periodontal search in its market. When somebody in Portland goes looking — on Google or, increasingly, by asking an AI system — the record is simply there, and it’s deeper than anyone else’s.
Here’s the part I want you to sit with.
There is no version of that outcome that gets produced in a sprint. No workflow builds it. No budget accelerates it. No amount of clever tooling in 2026 can make up for ten years of published evidence. It exists for one reason: somebody started a decade ago and never stopped.
And if we’d paused that for nine months to go build something internally, the practice wouldn’t be nine months behind schedule. The record would just be missing nine months, permanently, and no amount of catching up afterward would put them back.
So what do you do with this?
Not stop building. That’s not the argument, and it would be bad advice. Some of this work genuinely collapses a task that used to eat a day.
The move is to make the trade visible before somebody makes it for you.
Right now, in most organizations, the AI meta-work is invisible. It’s not on a plan, it’s not in a standup, and it’s not attributed to anyone. Which means it isn’t a decision — it’s a leak. The first thing to do is give it a name and a line, so that when your team spends thirty hours on an internal tool this month, thirty hours appear somewhere that you can see them.
And then apply one test to everything competing for those hours.
If this slips two quarters, is it recoverable?
If yes — it’s a tool, a workflow, an internal system — then it can wait, and it should be the thing that waits.
If no — if it’s an asset that only accumulates by being worked on continuously — it goes first. Not because it’s more urgent. Almost by definition it isn’t. It goes first because it’s the only one of the two with a real expiration date.
Three questions to sit with this week.
First: how many hours did your marketing team spend last month building or maintaining AI tooling? If you can’t answer that within a reasonable range, that itself is the answer.
Second: Of everything on your team’s plate right now, which items are genuinely recoverable if they slip by two quarters — and which are gone?
And third: nine months from now, what will exist that couldn’t have been built quickly? If the honest answer is nothing, then you already know where this year’s hours went.
Thank you for listening. I’m Jeff Payne. I will see you next time.
Your marketing team is quietly building software. The hours are coming out of the one asset that can’t be rebuilt later.
Your marketing department became a software team
91% of marketing leaders say their teams use AI. That figure surprises no one anymore. The number worth pausing on is the second one: roughly two-thirds say their organization builds its own internal AI tools for marketing.
Building. Not buying, not subscribing to — building. Somewhere in the last 18 months, a great many marketing departments became part-time software teams. Nobody approved it. Nobody staffed for it. It does not appear in the marketing plan, nor is it a line item in the budget. It simply started happening, one clever shortcut at a time.
Ask the team whether it is paying off, and they will say yes. They will mean it. The research suggests they are probably wrong.
Nobody approved it. Nobody staffed for it. It simply started happening, one clever shortcut at a time.
The savings are real. So is the leak.
In a survey of 3,200 leaders and employees, 85% reported that AI saves them between 1 and 7 hours per week. The time savings are not in dispute.
The same research then followed those hours. 37% of them go straight back out — correcting, verifying, fixing, and rewriting output that did not come back usable. 10 hours saved, close to four hours handed back. Only 14% of respondents said they consistently come out ahead.
A second leak is harder to see because it lands on someone else’s ledger. Work gets passed off as looking fine when it isn’t. In a separate study, 41% of workers said they had received something like it in the past month, and each instance took the recipient just under two hours to resolve. The sender saved 20 minutes. On any dashboard that counts output, the sender looks excellent.
This is one of the best written, most informative email newsletters I’ve ever read!
Thank you for sharing your expertise, and hope you’re doing well.
Beth Zimmerman
Founder, Executive Director
AI does not delete work. It moves it.
The work shifts from doing the task to building and maintaining the thing that does the task — prompting, checking, correcting, and patching it when a model updates and it stops working on a Tuesday.
Some of that is genuinely worth doing. The expensive part is not the hours themselves. It is where they are drawn from.
They are almost never drawn from the urgent thing because the urgent thing has a deadline and someone waiting on it. They are drawn from the work with the longest gap between effort and result: publishing enough depth on a subject to become the obvious answer, earning mentions on the sources AI systems actually cite, building a review record that means something. That is always the flexible budget. It is always the thing that can wait a few weeks.
AI does not delete work. It moves it.
The work shifts from doing the task to building and maintaining the thing that does the task — prompting, checking, correcting, and patching it when a model updates and it stops working on a Tuesday.
Some of that is genuinely worth doing. The expensive part is not the hours themselves. It is where they are drawn from.
They are almost never drawn from the urgent thing, because the urgent thing has a deadline—someone, a person, waiting on it. They are drawn from the work with the longest gap between effort and result: publishing enough depth on a subject to become the obvious answer, earning mentions on the sources AI systems actually cite, building a review record that means something. That is always the flexible budget. It is always the thing that can wait a few weeks.
It is drawn from the work with the longest gap between effort and result. That is always the flexible budget.
The asymmetry nobody is pricing
A workflow can be finished later. Delay it a quarter, and the tool arrives in June instead of March. Scrap it entirely and the hours are lost, but nothing else is. A tool delayed is just a tool delayed.No
Authority can recover or behave that way. The citations that were not earned, the mentions that were not placed, the reviews that were not requested while the client was still delighted — none of that is sitting in a queue. The window opened, it closed, and the business was not in it.
There is no sprint that recovers those months. The record has a gap where it should be, and it is permanent.
A delayed tool is just a delayed tool. A quarter of unearned authority is gone.
What a decade of compounding actually looks like
Dr. Kamran Haghighat at Portland Perio Implant Center in Portland, Oregon, has been a client for roughly ten years. In that time, we have published more than a hundred patient stories — one at a time, over a decade, never as a batch or a campaign.
Almost none of them looked important during the week they went live. Any single one, in isolation, is a rounding error. Today the practice is the dominant name in periodontal search in its market. When someone in Portland goes looking — on Google or, increasingly, by asking an AI system for the best Periodontist in the Portland area — the record is there, and it is deeper than anyone else’s.
No workflow builds that. No budget accelerates it. No amount of tooling in 2026 can make up for ten years of published evidence. It exists because someone started a decade ago and never stopped. Nine months paused would not have put the practice nine months behind schedule — the record would simply be missing nine months, permanently.
A delayed tool is just a delayed tool. A quarter of unearned authority is gone.
Make the trade visible before someone makes it for you
This is not an argument against building. Some of this work genuinely collapses a task that used to consume a day.
In most organizations, the AI meta-work is invisible — not on a plan, not in a standup, not attributed to anyone. That makes it a leak rather than a decision. Give it a name and a line, so that thirty hours spent on an internal tool appear somewhere a leader can see them.
Then apply a single test to everything competing for those hours: if this slips two quarters, is it recoverable?
If yes, it can wait, and it should be the thing that waits.
If no — if it only accumulates by being worked on continuously — it goes first.
Not because it is more urgent. Almost by definition, it is not. It goes first because it is the only one of the two with a real expiration date.
If it only accumulates through continuous work, it goes first — because it is the only one with a real expiration date.
THREE QUESTIONS WORTH SITTING WITH
How many hours did the marketing team spend last month building or maintaining AI tooling? If that cannot be answered within a reasonable range, that is itself the answer.
Of everything on the team’s plate right now, which items are genuinely recoverable if they slip two quarters — and which are gone?
Nine months from now, what will exist that could not have been built quickly? If the honest answer is nothing, the question of where this year’s hours went has already been settled.
Frequently Asked Questions
What is the AI productivity paradox?
It describes work that feels faster, while effort shifts to prompting, checking, correcting, and maintaining the system. Survey research finds that roughly 37% of the time AI saves is spent on rework, and only 14% of employees report net-positive outcomes consistently.
Why does building internal AI tools cost marketing teams more than expected?
Every workflow becomes a small permanent maintenance job owned by whoever built it. That job does not appear in the marketing plan or the budget, so the hours are drawn silently from other work — usually the work with the longest lag between effort and result.
Which marketing work becomes unrecoverable if it gets delayed?
Authority work. Citations, third-party mentions, and customer reviews accumulate only through continuous work. A delayed internal tool arrives later; a quarter of unearned authority cannot be backdated or produced in a sprint.
How should a leader decide which AI projects are worth the hours?
Ask whether the work is recoverable if it slips two quarters. Tools, workflows,kflows and internal systems can wait. Assets ty compound through continuous effort are not, so they should be protected first — not because they are urgent, but because they expire.
Episode informed in part by “The AI hours nobody on your marketing team is counting” by Kevin Indig and Amanda Johnson, published in Growth Memo. Survey figures referenced on-air are drawn from published research by Workday (conducted with Hanover Research, 3,200 leaders and employees), HubSpot’s AI Trends for Marketers report, and BetterUp Labs with Stanford Social Media Lab.
COMPLETE THE FORM TO
BOOK A STRATEGY CALL
"*" indicates required fields
COMPLETE THE FORM TO
BOOK A STRATEGY CALL
"*" indicates required fields
Subscribe and Share – WE APPRECIATE YOUR SUPPORT
