I work on AI every day, so saying we canceled Claude is a pretty ridiculous way to open this newsletter. Especially when the entire industry seems obsessed with getting every employee to use more AI.
But somewhere along the way, we started treating opening a prompt box like an accomplishment. People were spending hours perfecting prompts, building tools nobody needed, and generating an incredible amount of work for someone else to read.
We were using more AI, but I wasn't convinced we were getting more done.
So we decided to try something different. We canceled Claude, stopped handing everyone another tool to figure out, and started looking at the work we actually wanted AI to own.
And it turns out we weren't the only ones with an AI productivity problem.
We Canceled Claude & I Think We're Better Off.
We cancelled Claude and I think we might be better off. There, I said it. Some of you won’t get past that sentence without opening Claude to draft a rebuttal. That’s kind of my point.
Before you freak out, I’m not saying it’s forever. I wanted to see what happened when we stopped treating the prompt box like a life support machine.
And holy smokes, the dependency was pathetic.
- Ten minutes getting Claude to write an email they could’ve written in three.
- Two hours fighting it over a deck they could’ve built in one.
- AI hackathons that produced garbage nobody used.
- Training session after training session teaching people how to prompt.
We got so busy teaching people to use AI that we forgot to ask whether any of it was growing the business.
So we canceled the licenses and stopped the insane token spending. Honestly? It felt pretty damn good.
Then we got serious about where AI actually helps. Our AI ops team built AI Employees around work people actually do.
- One answers questions from the CRM and updates records after calls.
- One role plays with reps before tough sales conversations.
- One builds prospect lists.
- One follows up with inbound leads before they go cold.
The list goes on.
“Couldn’t they do all that with Claude?”
Sure. And every rep could spend the afternoon building their own version, fixing it when it breaks, and calling that a productive day. Instead, our AI ops team builds and supports it once. The reps use it and get back to selling.
Maybe we give Claude back to everyone. But a few weeks off the drug has made one thing painfully clear: a lot of people have confused opening Claude with doing their jobs.
And if you think that's bad, wait until you look at what happens when everyone starts sending their AI-generated work to everyone else.
Your Productivity Is Becoming Someone Else's Problem
AI slop grenades are exploding across companies, killing productivity and it’s our fault.
Here is how we got here. CEOs handed ChatGPT and Claude to every employee and said, “Use AI for everything.” So they did.
Now your team sends meeting recaps longer than the meeting, your manager forwards a 30-page report they haven’t read, and both somehow end with the same threat: “Would love your thoughts.”
The person sending it feels like a freaking genius because AI saved them an hour. Meanwhile, everyone else loses five hours reading, interpreting, summarizing, and trying to figure out what the hell any of it means.
Shopify CEO Tobias Lütke calls that an AI slop grenade and they are everywhere and in every company.
If I’m honest, I’ve thrown a few myself. And I gotta admit, it feels pretty damn good. You dump some half-baked sh*& into Claude, hit send before thinking, and walk away feeling super productive.
Then one lands in your inbox and you have to ask Claude what the hell Claude just wrote. hahah
That’s when you look around and realize everyone is throwing grenades at each other, the whole damn company is on fire, and the CEO is standing in the middle celebrating because AI adoption is up 300%.
It’s messy, really messy.
And that's the problem with measuring AI productivity at the individual level. Saving one employee an hour sounds fantastic until five other employees have to spend their afternoon dealing with what that person produced.
But there are real opportunities to use AI in ways that remove work instead of creating more of it. Sales is one of the clearest examples I've seen.
Most Sales Teams Are Starting in the Wrong Place
Most sales teams have the dumbest possible AI strategy.
I work on AI in sales every day, and I’m still figuring sh*% out. But after working through this with customers, I’m convinced there’s a better order of operations to get ROI from AI:
FIRST, STOP IGNORING BUYERS AFTER 5PM. Have AI respond to inbound leads, qualify them, and book meetings for your SDRs. It’s a small job but a great place to start.
In one customer case study, the human-plus-AI process produced 14% more closed-won deals than the AI-only process. I said deals, not SQLs. Double qualification could give you fewer SQLs and better opportunities.
THEN, GIVE AI A PORTION OF INBOUND. Let it qualify buyers and book directly with AEs. If meeting quality holds, explain why every lead still needs to pass through a human SDR before reaching someone who can run the deal.
THEN TACKLE OUTBOUND. Let AI research accounts and help reps practice. Feed your best buying signals to humans and the next tier to AI. For high-volume campaigns, let AI handle replies and book the buyers who actually engage.
THEN LET IT DEMO THE PRODUCT. A real, personalized demo where the buyer asks questions and sees what matters to them. Spare me the prerecorded video with a chatbot stapled to the corner or some clickable BS demo. Im talking about an AI Employee walking through a real demo of your product.
FINALLY, LET IT HELP CLOSE. Let an AI employee take the sales call, handle the buyer’s questions, and get the deal done.
Now, word of caution: If you’re selling jumbo jets to the king of Saudi Arabia, keep your best humans on it. No AI will help you there. If you’re selling to SMBs and making interested buyers wait until Tuesday for a first conversation, what the hell are you doing?
I wrote a full article on this here.
Of course, putting better AI around your sales team only gets you so far. There's another piece of this that I think we're going to see become increasingly important.
AI Is Making Great Reps Dangerous
AI is turning great sales reps into killers and bad reps into confident idiots.
I'm sorry to say that anything times zero is still zero. AI just lets bad reps suck with better slides, borrowed confidence, and a trail of AI slop behind them.
But a great rep surrounded by powerful AI? It's terrifying what they can do.
And I think that's an important distinction. Giving everyone the same AI tools doesn't suddenly give them the same skills. We can make it easier to produce emails, presentations, and prospect lists, but that doesn't mean we're making anyone better at selling.
Stop Celebrating AI Adoption
That's what canceling Claude forced us to confront. We'd spent so much time getting people to use AI that we'd stopped paying attention to whether it was helping them do their jobs.
I don't particularly care how many prompts our sales team writes. I care whether we're responding to buyers we used to ignore, whether reps are walking into better conversations, and whether our best people are closing more deals.
And I think that's where a lot of companies are going to get stuck. They'll keep spending money on licenses, celebrating adoption, and generating more work for everyone else while convincing themselves they're becoming more productive.
If your entire AI strategy is getting everyone to use AI, you've confused the tool with the job.
