AI-First Isn’t the Flex You Think It Is: Lessons Growth-Stage Businesses Can Learn From the Klarna and Duolingo Mess

What Klarna and Duolingo got wrong, and how growth-stage teams can get it right

Every few years, the tech world gets caught up in a wave of transformation. And right now, that wave is AI.

You’ve heard the predictions: AI will replace knowledge workers, collapse entire departments, and rewrite how we build and sell. Some of it’s true. A lot of it’s noise.

But what’s not being talked about enough is this: AI won’t harm your business on its own—your decisions about AI might.

Just ask Klarna and Duolingo.

Both companies made bold moves to go “AI-first.” Both prioritized cost savings and speed—focusing on short-term gains and shareholder optics. What they didn’t fully consider was the long-term impact: on buyer experience, employee trust, brand reputation, and operational consistency.

Now, both are managing the consequences—revising their strategy in public, fielding backlash from their communities, and trying to rebuild what they disrupted too fast.

This isn’t a teardown piece. It’s a reality check.

Because here’s what I’ve learned leading go-to-market strategy for growth-stage companies: The tech isn’t fully there yet—and more importantly, neither are most teams.

AI might impress in a demo, but it doesn’t understand context, nuance, or emotion the way your people do. Not yet. And definitely not without guidance. If you don’t take the time to understand what your buyers and workers expect—and align your systems accordingly—you risk moving faster in the wrong direction.

Let’s talk about what went wrong, what’s working instead, and how you can move forward with clarity.

Klarna and Duolingo’s AI-First Missteps

Over-reliance on AI had led to lower quality service

Klarna’s Overcorrection

In February 2024, Klarna announced that its AI assistant had handled two-thirds of customer service chats in its first month, effectively replacing 700 full-time agents. The company highlighted faster resolutions and reduced costs.

But by May 2025, they walked it back. CEO Sebastian Siemiatkowski publicly admitted that over-reliance on AI had led to lower quality service. Klarna started rehiring humans, including remote support reps, to restore trust and improve consistency.

Long-time users ended their learning streaks in protest.

Duolingo’s Brand Backlash

In April 2025, Duolingo declared itself “AI-first,” reducing contractor roles and baking AI usage into team KPIs. It didn’t go over well. They quickly got dragged on social media. Long-time users ended their learning streaks in protest. Concerns about content quality, cultural sensitivity, and trust surfaced immediately—and the brand took a hit.

The Common Thread

Both companies moved too fast. They prioritized efficiency without testing for impact. They led with what AI could do, not what their people and users actually needed (or wanted). From the outside, it definitely looked like a profit over people play.

What the Women x AI Community Already Knows

The tech isn't there yet.

While headlines chase the flashiest wins, there’s a smarter, more grounded conversation happening—one I’ve had often with leaders like Jenny Kay Pollock, Jedidah K., Aprelle Duany, Pallavi Sharma, and Fleur Prince. All of whom have guested on Revenue Remix to share their insights, and the community around these conversations is growing fast.

Side note: I found the Women X AI community after Klarna announced in May 2024 that they were laying off half their marketing team and going all-in on AI there as well. (I swear I'm not just dunking on them.) The increasing amount of profit-over-people policies in AI deployment was concerning. I wanted to find people who were talking about ethics, responsible use, and who were thinking about... people. And to risk over-using that word, I found my people. A group of women and a few men working with and in AI with all levels of technical capabilities.

We’ve all adopted AI into our workflows. We’ve all used it to reduce redundant tasks. And we’ve all said—out loud, repeatedly—that the tech isn’t fully there yet.

Here’s what we agree on:

• AI adoption must be human-centered.

• Profits should never be prioritized over people—whether workers or buyers.

• There will be backlash when companies rush deployment without ethics or oversight.

We’ve joked—only half-jokingly—that the next job wave might be “AI clean-up.” People hired to clean up the messes made by sloppy AI implementation.

And now we're seeing it happen.

If you’re going to automate, you need oversight. If you’re going to scale, you need structure. Even agentic AI won’t be able to operate fully autonomously. It will always need responsible humans behind it. That’s what we’re building toward. An AI-powered future of work that is deeply rooted in understanding of buyers and empathy for human workers.

And apparently Klarna and Duolingo missed that memo.

When Smart Strategy Meets Fragile Tech

Even when you think you do everything right, things can still go sideways.

In one client project, we rolled out an AI-enhanced personalization strategy using two leading tools: one for data enrichment, the other for automated outreach. The objective was clear: leverage tech tools to nurture a database of 14,000 contacts to help a small team manage enterprise level deals with large buying teams and even larger sales cycles.

The idea was solid—identify key buying group members, enrich the CRM, personalize the messaging, and launch a multi-step campaign.

What actually happened:

• About 20% of the enriched contact data was wrong

• Out-of-office replies were flagged as “high engagement”

• Long job titles made emails sound robotic and off-base

• In short, two famous tech platforms I'm not shady enough to name failed on their promises

We were sending messages that looked customized—but were off in key areas. And worse, the AI was stacking those errors. It read false engagement as buying intent and pushed those contacts deeper into a high-touch "push for the meeting" sequence.

Luckily, we caught it fast.

A human flagged the awkward replies and the false "high engagement" signals. We paused the sequence, re-imported the contacts, and rebuilt the logic. We also added a manual review step for all “highly engaged” contacts—because AI doesn’t know the difference between curiosity and sarcasm. Or out of office replies. Or the person who replies with "not interested" email instead of unsubscribing.

We didn’t scrap the strategy. But we dialed back the AI-enhanced and automated personalization—because even smart systems break when the inputs are off.

That experience didn’t shake my belief in AI. It reinforced my human-led approach.

What It Looks Like When You Do It Right

Inside the Smart Sales Playbook rollout that actually worked

One of my clients wasn’t chasing an AI headline. He was trying to transition out of founder-led sales—without losing the trust, process, and personalization that had helped the company grow. He came from operations. Managing a salesperson directly wasn’t in his comfort zone. But he also knew something that many founders don’t: sales isn’t a set-it-and-forget-it job. It needs coaching, structure, and regular course-correction.

So we built it

We created a full revenue playbook—personas, battle cards, messaging, partner strategy, success handoffs. And since I know that salespeople don't use those awesome playbooks after their initial training period, we took it a step further. We trained a custom GPT on that playbook and layered in my coaching frameworks. The result?

A GPT that sounded like me. A tool the rep could use to prep for meetings, brainstorm follow-ups, qualify faster, and stay focused.

He called it “a coach in my pocket.” And joked to his wife that he had "made it" and had his own personal assistant. And that’s exactly what we designed it to be.

The owner got confidence that sales would keep moving without him. The rep got a system that helped him perform independently. And the company no longer needed me—which is the point. Someone like me isn't supposed to be around forever.

How We Rolled It Out (And Why It Worked)

If you’ve ever rolled out a sales process—or introduced a new tool—you know most reps ignore both.

Here’s why this one landed:

1. We trained the system before we trained the rep. The GPT was fully prepped on the company’s real strategy, not generic templates. It knew how to act, how to coach, and how to respond like I would.

2. We emphasized safety, responsibility, and security. Before he did anything, the rep learned how to use the tool safely and responsibly. He was also schooled on what it's good at, and where it falters, and how he should lead the AI tool. Security wasn’t an afterthought—it was the first conversation.

3. We told him it was there to boost his superpower. He’s great at building relationships. And that’s not something AI will ever do. This tool freed him up to do more of what he’s best at—and he understood that from the jump.

4. We gave him a fast start. No clunky onboarding. Just a two-page quick guide and a prompt library.

5. We asked how it was going—and listened. He told us it helped him shorten meetings. Not because we told him to—because it made him better at prep, faster at follow-up, and clearer in the room.

That’s what adoption looks like when it’s done right.

AI Adoption Myths – A Mini FAQ

Here are the myths I still hear—and how I answer them.

“AI will replace our sales team.” No. AI supports the team. It doesn’t replace human judgment, relationship-building, or live deal strategy.

“Once we automate, we can reduce headcount.” That’s not the win. The real move is doing more with the same resources. You already have the team and the budget—AI should help you expand output, not slash ops.

“AI will tell us what to do.” It won’t. It reflects what you feed it. Strategy still has to come from humans.

“AI is a one-time setup.” Absolutely not. It’s an ongoing system that needs maintenance, feedback, and iteration.

“If we pick the right tool, we’ll be set.” Wrong. The tool is just the delivery method. Success comes from clarity, alignment, and real-time support.

One Thing You Can Do This Week (It’s a Short Week)

Pick one workflow your team dreads. (Maybe it's the thing you want them to do that they never do.) Ask: “What part of this could AI handle—without hurting the client experience or internal trust?” Then pilot one tiny change. That’s it.

You don’t need a big reveal. You need a clean experiment.

Because smart adoption doesn’t start with a tool. It starts with an objective.

Want to Build the Sales System Your Team Actually Uses?

If you’re building a modern GTM motion and want your sales systems to scale without breaking trust, RiseAI Sales Accelerator, the official name of my Smart Sales Playbook program, is your next move. This is where I blend the magic of an AI tool with human ingenuity to boost your team's superpowers.

And if you want to hear how women revenue leaders are approaching AI adoption with nuance, strategy, and actual empathy—check out my podcast.

The tech is moving fast. How you lead through it—that’s what matters.

Curious Where You Stand?

If your team is exploring AI—or already deploying it—but you’re not sure if your systems, messaging, and sales motion are actually aligned…

Start with an Alignment Scan.

It’s a fast diagnostic I built for growth-stage teams who want:

• A clear view of where AI is helping (or hurting)

• Insight into sales + marketing disconnects

• A roadmap for scaling smarter, not sloppier

It’s quick, it’s actionable, and it’s built for operators who want to move fast without breaking trust.

Let’s make sure your tools, team, and buyer experience are actually working together.

---

Rise of Us is a practice run by Summer Poletti, specializing in revenue growth: sales, strategic partnerships, customer success, marketing alignment. We generally work with financial services and SaaS companies from $2MM - $20MM ARR and help them plan and execute for their next stage of revenue growth. We concentrate on strategy, coaching, and organizational alignment.