What does AI cost you?
I saw a post on Threads last week, and I haven’t been able to stop thinking about it.
The author (@jaym.imbrium.arts on Threads), clearly articulates the pain many of us are experiencing with AI. It’s three posts long, and it does more work than most of what I’ve written about AI adoption this year.
“So, we are required to use AI at work. Full stop. Also to report how we are using AI. Yesterday, I had a task it could actually help with: turning screenshots into a spreadsheet. It did that. I spent a few hours adjusting and formatting data to make something nice.
Opened sheet in a meeting today. 98% of my changes were gone. They don’t exist. Anywhere. And the sheet isn’t saving new changes I make.”
“I had to copy the data out, and now I have to rebuild the spreadsheet. This task will now take more time than if I had done it manually.
And yes, I am reporting this, and every other issue. Bc it does not save time.”
“We had a 90 minute training on AI usage yesterday.
‘When you’re doing a task, stop and think about how AI might be useful. You’ll probably find a way it can work for you.’
AI is a cult. It has some small improvements to workflows, yes, but it is a flitchy, broken, expensive, planet-destroying mess, and I am so incredibly tired of hearing how it will make our workloads so much lighter.
It does not. It makes them heavier.”
You can read the whole thread here.
I agree with every line of that, including the parts that aren’t comfortable for me.
Because what the author is describing isn’t AI failing; it’s a rollout failing. There are companies doing this well. There are a lot more doing it exactly like this, where the people on the receiving end are just trying to survive it, check the box, and get back to their real work.
And yes, I sell AI enablement. Companies like this one are precisely who I sell it to. What I want is for people to still be good at their jobs in three years, not to sort of master the shiny object this quarter. That only happens if you start with the foundation, and the foundation is your people and your processes, not your platform. That’s the order in the 5P Framework by Trust Insights™ (Purpose, People, Process, Platform, Performance), and the order isn’t decorative. Platform is fourth on purpose. Almost everything in that thread is what it looks like when an organization starts at four.
“It does not. It makes them heavier.” That’s a measurable claim, by the way, and I’ll show you in a minute that the data backs it up.
Mostly, though, I read that thread and felt caught because I write to the person who ran that training. I write to the person who signed off on the mandate. I don’t usually write to the person on the other end of it, and that person is the one paying for all of this.
The mandate is what you say when you don’t have a strategy
“Use AI, and report back on how you’re using it.”
Read that again as an instruction. What it’s actually asking is for every individual employee to go find the use cases, evaluate the tools against their own work, absorb the failures when they come, and then file a report on all of it. They’re doing that on top of the job they were already hired to do, with ninety minutes of training behind them and no change to their workload, their deadlines, or their paycheck.
That’s not an AI strategy. That’s the discovery phase of an AI strategy, assigned to the people with the least authority and the least time, and reframed as empowerment.
I’m not the only one landing here. Harvard Business Review ran a piece in January on why AI output quality is falling apart inside companies, and the first root cause they name is unclear AI mandates. Not skills. Not tools. The mandate.
Nobody did this maliciously. Leadership got a directive with a number attached, usage is the only thing that’s easy to count, so usage became the goal. The person who ran that ninety-minute training probably knew it was thin and had nothing better to give.
But look at what it produces. Somebody loses hours, rebuilds a spreadsheet by hand, and the only thing that reaches the report is that they used AI. Which, technically, they did.
Somebody gave you a bread machine
You didn’t have a bread problem. You have never once stood in your kitchen and thought, “what this house needs is more bread.” (Clearly that person is not me).
But it was a gift, so now it lives on the counter, in the spot where you used to make coffee. And every time that person comes over, they ask if you’ve used it yet. So on a Tuesday at nine at night, you make a loaf, so you have something to say.
Meanwhile, the dishwasher has been broken since March. You’re washing everything by hand. Nobody asks about the dishwasher because nobody bought you the dishwasher.
That’s what this feels like from the inside. Screenshots into a spreadsheet are a bread machine task. It got picked because it was the easiest place to start, not because it was the thing that hurt. And when it broke, the person still had the original problem, plus a rebuild.
Three costs, and your form has a field for none of them
The rework is real and somebody already measured it. BetterUp Labs and the Stanford Social Media Lab surveyed 1,150 full time US desk workers: 40 percent had received AI-generated work in the past month that looked finished and wasn’t. Average cleanup, about two hours per incident. They put it at $186 per employee per month, roughly $9 million a year in a 10,000 person company. Your usage report has a field for “did you use AI.” It has no field for “how long did it take me to fix.”
The upside doesn’t come back. A February survey of 1,250 corporate workers found 31 percent said their workload went up after AI showed up, nearly double the 16 percent who said it went down. Of those, 43 percent said it at least doubled. The gains, where they exist, get absorbed as capacity. So the honest version of the deal is: learn a new tool on your own time, eat the failures, produce more, and nothing else about your job changes.
The person reporting the problems becomes the problem. This is the one I take personally. I spent years being the person who asked the inconvenient question in the meeting, and I got labeled problematic for it. Not wrong. Problematic. So when I read “yes, I am reporting this, and every other issue,” I know exactly how that reads in a status meeting by the fourth week. That person is doing your quality assurance for free, and the reward for it is a reputation.
Here’s what makes that last one expensive rather than just unfair. That’s your data. That is the only real signal you have about whether any of this is working, and the way most organizations are set up, it gets filed as an attitude problem.
What to do next
Change the question on the form. Instead of asking whether people used AI, ask what it cost them to use it. Two fields will do: what you tried and how many hours you spent fixing what came back. The number you get that first month is going to be ugly, and it’s also going to be the most useful thing you see all quarter.
Take use case discovery back. Finding where AI fits in your operation is a leadership job because it requires seeing across departments, and nobody can do that from inside one of them. Handing it out as homework isn’t really distributing the work so much as distributing the blame for it not getting done.
Close the loop. Every problem somebody reports should get a reply from a named human, even when the reply is that you can’t fix it yet. One-way reporting teaches people to stop reporting, and then you’re flying blind and calling it adoption.
Check whether anything came back. If output went up and nothing changed for the people producing it, what you have isn’t an adoption problem, it’s a fairness problem, and no amount of enablement is going to solve that one for you.
Want to learn more about how we help? Get more information here.
The moral of the story
AI isn’t failing because people are resistant. It’s failing because we handed the hardest part of the work to the people with the least power to do it, then measured them on whether they looked busy doing it.
The person in that Threads post isn’t a laggard. They’re the best AI analyst in that company, and they’re reporting for free, and somebody is about to write them up for negativity. (But I hope not; I’ll certainly help defend them.)
What’s your organization actually asking people to report about AI?
Reply to this email or join the conversation in our Free Slack community, Analytics for Marketers!
– Katie Robbert, CEO
What does AI cost you?
I saw a post on Threads last week, and I haven’t been able to stop thinking about it.
The author (@jaym.imbrium.arts on Threads), clearly articulates the pain many of us are experiencing with AI. It’s three posts long, and it does more work than most of what I’ve written about AI adoption this year.
“So, we are required to use AI at work. Full stop. Also to report how we are using AI. Yesterday, I had a task it could actually help with: turning screenshots into a spreadsheet. It did that. I spent a few hours adjusting and formatting data to make something nice.
Opened sheet in a meeting today. 98% of my changes were gone. They don’t exist. Anywhere. And the sheet isn’t saving new changes I make.”
“I had to copy the data out, and now I have to rebuild the spreadsheet. This task will now take more time than if I had done it manually.
And yes, I am reporting this, and every other issue. Bc it does not save time.”
“We had a 90 minute training on AI usage yesterday.
‘When you’re doing a task, stop and think about how AI might be useful. You’ll probably find a way it can work for you.’
AI is a cult. It has some small improvements to workflows, yes, but it is a flitchy, broken, expensive, planet-destroying mess, and I am so incredibly tired of hearing how it will make our workloads so much lighter.
It does not. It makes them heavier.”
You can read the whole thread here.
I agree with every line of that, including the parts that aren’t comfortable for me.
Because what the author is describing isn’t AI failing; it’s a rollout failing. There are companies doing this well. There are a lot more doing it exactly like this, where the people on the receiving end are just trying to survive it, check the box, and get back to their real work.
And yes, I sell AI enablement. Companies like this one are precisely who I sell it to. What I want is for people to still be good at their jobs in three years, not to sort of master the shiny object this quarter. That only happens if you start with the foundation, and the foundation is your people and your processes, not your platform. That’s the order in the 5P Framework by Trust Insights™ (Purpose, People, Process, Platform, Performance), and the order isn’t decorative. Platform is fourth on purpose. Almost everything in that thread is what it looks like when an organization starts at four.
“It does not. It makes them heavier.” That’s a measurable claim, by the way, and I’ll show you in a minute that the data backs it up.
Mostly, though, I read that thread and felt caught because I write to the person who ran that training. I write to the person who signed off on the mandate. I don’t usually write to the person on the other end of it, and that person is the one paying for all of this.
The mandate is what you say when you don’t have a strategy
“Use AI, and report back on how you’re using it.”
Read that again as an instruction. What it’s actually asking is for every individual employee to go find the use cases, evaluate the tools against their own work, absorb the failures when they come, and then file a report on all of it. They’re doing that on top of the job they were already hired to do, with ninety minutes of training behind them and no change to their workload, their deadlines, or their paycheck.
That’s not an AI strategy. That’s the discovery phase of an AI strategy, assigned to the people with the least authority and the least time, and reframed as empowerment.
I’m not the only one landing here. Harvard Business Review ran a piece in January on why AI output quality is falling apart inside companies, and the first root cause they name is unclear AI mandates. Not skills. Not tools. The mandate.
Nobody did this maliciously. Leadership got a directive with a number attached, usage is the only thing that’s easy to count, so usage became the goal. The person who ran that ninety-minute training probably knew it was thin and had nothing better to give.
But look at what it produces. Somebody loses hours, rebuilds a spreadsheet by hand, and the only thing that reaches the report is that they used AI. Which, technically, they did.
Somebody gave you a bread machine
You didn’t have a bread problem. You have never once stood in your kitchen and thought, “what this house needs is more bread.” (Clearly that person is not me).
But it was a gift, so now it lives on the counter, in the spot where you used to make coffee. And every time that person comes over, they ask if you’ve used it yet. So on a Tuesday at nine at night, you make a loaf, so you have something to say.
Meanwhile, the dishwasher has been broken since March. You’re washing everything by hand. Nobody asks about the dishwasher because nobody bought you the dishwasher.
That’s what this feels like from the inside. Screenshots into a spreadsheet are a bread machine task. It got picked because it was the easiest place to start, not because it was the thing that hurt. And when it broke, the person still had the original problem, plus a rebuild.
Three costs, and your form has a field for none of them
The rework is real and somebody already measured it. BetterUp Labs and the Stanford Social Media Lab surveyed 1,150 full time US desk workers: 40 percent had received AI-generated work in the past month that looked finished and wasn’t. Average cleanup, about two hours per incident. They put it at $186 per employee per month, roughly $9 million a year in a 10,000 person company. Your usage report has a field for “did you use AI.” It has no field for “how long did it take me to fix.”
The upside doesn’t come back. A February survey of 1,250 corporate workers found 31 percent said their workload went up after AI showed up, nearly double the 16 percent who said it went down. Of those, 43 percent said it at least doubled. The gains, where they exist, get absorbed as capacity. So the honest version of the deal is: learn a new tool on your own time, eat the failures, produce more, and nothing else about your job changes.
The person reporting the problems becomes the problem. This is the one I take personally. I spent years being the person who asked the inconvenient question in the meeting, and I got labeled problematic for it. Not wrong. Problematic. So when I read “yes, I am reporting this, and every other issue,” I know exactly how that reads in a status meeting by the fourth week. That person is doing your quality assurance for free, and the reward for it is a reputation.
Here’s what makes that last one expensive rather than just unfair. That’s your data. That is the only real signal you have about whether any of this is working, and the way most organizations are set up, it gets filed as an attitude problem.
What to do next
Change the question on the form. Instead of asking whether people used AI, ask what it cost them to use it. Two fields will do: what you tried and how many hours you spent fixing what came back. The number you get that first month is going to be ugly, and it’s also going to be the most useful thing you see all quarter.
Take use case discovery back. Finding where AI fits in your operation is a leadership job because it requires seeing across departments, and nobody can do that from inside one of them. Handing it out as homework isn’t really distributing the work so much as distributing the blame for it not getting done.
Close the loop. Every problem somebody reports should get a reply from a named human, even when the reply is that you can’t fix it yet. One-way reporting teaches people to stop reporting, and then you’re flying blind and calling it adoption.
Check whether anything came back. If output went up and nothing changed for the people producing it, what you have isn’t an adoption problem, it’s a fairness problem, and no amount of enablement is going to solve that one for you.
Want to learn more about how we help? Get more information here.
The moral of the story
AI isn’t failing because people are resistant. It’s failing because we handed the hardest part of the work to the people with the least power to do it, then measured them on whether they looked busy doing it.
The person in that Threads post isn’t a laggard. They’re the best AI analyst in that company, and they’re reporting for free, and somebody is about to write them up for negativity. (But I hope not; I’ll certainly help defend them.)
What’s your organization actually asking people to report about AI?
Reply to this email or join the conversation in our Free Slack community, Analytics for Marketers!
– Katie Robbert, CEO
|
Need help with your marketing AI and analytics? |
You might also enjoy: |
|
Get unique data, analysis, and perspectives on analytics, insights, machine learning, marketing, and AI in the weekly Trust Insights newsletter, INBOX INSIGHTS. Subscribe now for free; new issues every Wednesday! |
Want to learn more about data, analytics, and insights? Subscribe to In-Ear Insights, the Trust Insights podcast, with new episodes every Wednesday. |
Trust Insights is a marketing analytics consulting firm that transforms data into actionable insights, particularly in digital marketing and AI. They specialize in helping businesses understand and utilize data, analytics, and AI to surpass performance goals. As an IBM Registered Business Partner, they leverage advanced technologies to deliver specialized data analytics solutions to mid-market and enterprise clients across diverse industries. Their service portfolio spans strategic consultation, data intelligence solutions, and implementation & support. Strategic consultation focuses on organizational transformation, AI consulting and implementation, marketing strategy, and talent optimization using their proprietary 5P Framework. Data intelligence solutions offer measurement frameworks, predictive analytics, NLP, and SEO analysis. Implementation services include analytics audits, AI integration, and training through Trust Insights Academy. Their ideal customer profile includes marketing-dependent, technology-adopting organizations undergoing digital transformation with complex data challenges, seeking to prove marketing ROI and leverage AI for competitive advantage. Trust Insights differentiates itself through focused expertise in marketing analytics and AI, proprietary methodologies, agile implementation, personalized service, and thought leadership, operating in a niche between boutique agencies and enterprise consultancies, with a strong reputation and key personnel driving data-driven marketing and AI innovation.