AI Growing Pains
Companies are focused on what AI can do. Few are asking what it should do.
“AI can already do all of the jobs that we, as humans, do.”
That's what Klarna’s CEO Sebastian Siemiatkowski told Bloomberg late last year, right before he replaced the entire customer service department with AI chatbots, cutting the workforce by roughly 40 percent.
The result? They boasted $10 million in savings thanks to “AI efficiencies,” but reported $99 million in losses for Q1 this year.
So, you do the math.
Siemiatkowski now admits that cutting costs wasn’t the best idea. “Cost unfortunately seems to have been a too predominant evaluation factor when organizing this. What you end up having is lower quality.”
Try not to point and laugh at Klarna. The reality is a lot of companies are facing similar pressures to do more with less.
Two Paths to AI
There are two ways companies think about the value of AI.
“This could save us so much money.”
Or,
“This could help us do so much more.”
That’s the difference between efficiency and cutting costs. Companies often conflate the two, perhaps because so many use the word “efficiencies” to spin budget cuts when things get hard.
But there’s an important distinction.
A cost-focus is reductive. They look for ways to cut costs to improve profits, and quality becomes a casualty in the process.
An efficiency-focus is additive. They want to augment their team’s talent to make them more efficient and effective, without sacrificing quality.
Remember the Triple Constraint? "Good, fast, cheap—pick any two."
AI can do fast and cheap. But it struggles with the good part. You still need people with taste and experience to tell it what's good.
Of course it’s tempting to cut costs with AI. But don't cut the good out of the equation. The companies making headlines are learning the hard way that there's a line you shouldn’t cross.
The Quality Line You Can’t Cross
I've been thinking about where that line sits, and I keep coming back to the customer experience. It's the last bastion of quality.
Prospects and customers will put up with a lot if you have a good product, but if you push it too far, they’ll walk out the door.
Take Duolingo. They replaced human translators and content writers with AI. Sounds smart if you’re cost-focused, but it’s backfiring. People are calling the AI output “boring” and, worse, it makes mistakes. The brand is losing the fun, quirky personality that made Duolingo stand out in the first place.
Quality slips, and people aren’t happy. The customer experience suffers, so they leave.
When you automate interactions that have historically been human-to-human—like sales, customer service, or brand—it feels disingenuous and cheap. Like the company doesn't care enough about its customers to warrant spending money on them.
People notice the robotic tone in writing, sure. But it's way more pronounced when talking at an automated agent or chatting with a bot that clearly doesn't understand your question.
That kind of customer experience is the last thing you should automate. Your customers have enough robots in their lives already.
What Smart Teams Automate
So what should you be automating?
All the behind-the-scenes, workflow heavy stuff that doesn't directly touch the customer experience but absolutely drains your team's energy.
Data processing.
Administrative workflows.
Content tagging.
The spreadsheet swamp of sadness that bogs your people down and keeps them from their best work.
Automate that, but don’t sacrifice the customer experience.
You can use AI to cut creative talent and surrender your content to the same cheap robot labor that every other company employs. Or you can use AI to finally free your team from mundane tasks, clearing the way for them to build memorable, meaningful work that reminds people there are still humans behind your brand.
I know which one I'd choose. I'm betting you do too.
Worth Noting
AI accountability gap. Maybe it’s time to move from AI adoption to AI governance. McKinsey's latest research shows only 27% of companies review all AI-generated content before it goes public, while 47% of organizations have experienced negative consequences from using AI.
The market strikes back. Google announced its March 2024 update achieved a 45% reduction in low-quality, unoriginal content from search results, exceeding their 40% target. Mediavine (a major ad network) terminated publisher accounts for overusing AI content, stating they "do not monetize low-quality, mass-produced, unedited, or undisclosed AI content."
Try This
Want to find safe places to automate? Try this short exercise with your team:
Have everyone write down their three most draining tasks
For each task, ask: "Would customers notice if we automated this?"
Pick one or three where the answer is "no" and automate it for 30 days
Track time saved and team satisfaction
If both metrics improve and no customers complain, you've found your sweet spot for automation.
Napkin Notes
“The words in advertising are like the windows in a store. You must be able to look right through them and see the product. If you see the window, it's dirty-and you're going to see yourself, or you're going to see the smear. You're not going to see the product, and you're going to lose.” – Eugene Schwartz
Google announced AI mode, and I appreciate how Garrett Sussman breaks down the key points.
The CEO of Anthropic says that AI could wipe out half of all entry-level white-collar jobs. This from the guy who’s building the tools.