2 Reasons AI Isn’t Working for Your Company (And What to Do Instead)
AI is the most hyped business tool of the decade. Every company wants to use it to get faster and better—in fact, one poll showed that over 75% of respondents said that their organizations use AI. But why are so few of us seeing dividends? Hans Peter Brondmo says that it’s because we’re looking at AI all wrong. The former CEO of Everyday Robots, Google’s AI-meets-robotics moonshot, this Lavin Exclusive Speaker has spent decades creating AI-powered robots that actually help us in our day-to-day (think: helper robots who clean cafes and pick up trash). Below, find the 2 reasons he says AI isn’t working for your company (yet), and 2 better ways to use AI at work. And then contact us to book Hans Peter to speak at your next event!
Mistake #1: You’re expecting a grown-up, but the tech’s still a toddler.
AI may command trillion-dollar valuations, but that doesn’t mean it’s mature. And that shouldn’t be a surprise to us: ChatGPT has only been around for a few short years. “When was the last time you talked to a two-year-old?” Hans Peter asks. “They hallucinate all the time. They tell stories, random stories. It’s wonderful to listen to, but it’s also a little bit fantastical.” The difference is that this particular toddler has attracted billions of dollars in investment, creating wildly inflated expectations for mature performance.
The Google robots demonstrated this learning curve in tangible ways. “They were pretty much at this two-year-old stage,” Hans Peter recalls. “They would bump things over and drop things on the floor and go ‘whoops,’ and then pick them up again and learn all these things with a lot of human supervision.” Yet despite their limitations, these robots eventually “learned” how to do routine tasks and began to provide genuine value to the Google campuses. And the same thing will happen with generative AI tools: “Over time, as they train and as they become more attuned to what the world looks like, they will hallucinate less, and they will tell less fantastical stories, and learn. What we’re dealing with here is technology in its infancy—and in spite of that, it’s remarkable how quickly things are happening.”
You wouldn’t expect polished, error-free performance from a two-year-old. So don’t expect the same of AI.
What to do instead: You can and should use AI for productivity and efficiency at work—just remember to treat it like a talented but inexperienced hire. Set clear guardrails around its use. Supervise outputs carefully, especially for mission-critical applications. Most importantly, hold employees accountable for the final results rather than letting them publish raw AI outputs. “Don’t let employees just do the quick thing and publish something that came straight out of the mouth of Gemini or Claude,” Hans Peter says. “Have them do the human layer on top, which is critical analysis and thinking.”
Mistake #2: You’re waiting for the future, but it’s driven by today’s experiments.
By the end of his tenure at Everyday Robots, Hans Peter had over a hundred robots working across the Google campuses: cleaning tables, sorting trash, and even making music and dancing. But these breakthroughs didn’t happen overnight. Each task took relentless iteration for humans and robots alike, even for something as seemingly simple as sorting trash into the appropriate bins. 20 physical robots practiced sorting trash in Google’s facilities—and at the same time, 245 million virtual robots continued learning in simulation. “I like to think of it as if they were dreaming during the night,” Hans Peter says, “and then the physical robot wakes up and gets a new model from that dreaming.”
AI development isn’t about perfect solutions but about continuous learning. Each failed attempt, each minor improvement, each small success builds toward genuine capability. And the same is true for those of us looking to leverage AI in the workplace: the companies waiting for AI perfection will find themselves far behind organizations that started learning years earlier.
Think of your AI journey like you’re looking across a deep valley towards a mountain range. You can see the destination clearly: a warm, inviting hut atop a mountain representing AI’s full potential. But reaching it requires navigating what Hans Peter calls an “alligator-infested valley” full of obstacles, wrong turns, uncomfortable challenges, and probably mosquitoes. “It’s going to take some effort to get there, and some tricky navigation.”
What to do instead: Set guardrails, yes, but don’t let that stop you from getting to know the tools. Start experimenting now, even if your first attempts feel clunky or incomplete. Create small pilot programs where failure won’t damage your business but success could provide meaningful value. Treat every instance of AI implementation as a learning opportunity rather than a finished solution. And don’t restrict employees from using AI tools—instead, teach them to use these tools effectively while maintaining critical thinking and quality control. “Have your business become a learning organization that learns how and where AI fits into your daily workflows,” Hans Peter says.
Interested in booking this leading keynote speaker on AI and the future of work?
Hans Peter sat down with us recently to explain why AI should take our jobs, and why the robot revolution “can’t come fast enough.” Watch his episode of our Lavin Voices podcast:
Mistake #1: You're expecting a grown-up, but the tech's still a toddler.
AI may command trillion-dollar valuations, but that doesn't mean it's mature. And that shouldn't be a surprise to us: ChatGPT has only been around for a few short years. "When was the last time you talked to a two-year-old?" Hans Peter asks. "They hallucinate all the time. They tell stories, random stories. It's wonderful to listen to, but it's also a little bit fantastical." The difference is that this particular toddler has attracted billions of dollars in investment, creating wildly inflated expectations for mature performance.
The Google robots demonstrated this learning curve in tangible ways. "They were pretty much at this two-year-old stage," Hans Peter recalls. "They would bump things over and drop things on the floor and go 'whoops,' and then pick them up again and learn all these things with a lot of human supervision." Yet despite their limitations, these robots eventually "learned" how to do routine tasks and began to provide genuine value to the Google campuses. And the same thing will happen with generative AI tools: "Over time, as they train and as they become more attuned to what the world looks like, they will hallucinate less, and they will tell less fantastical stories, and learn. What we're dealing with here is technology in its infancy—and in spite of that, it's remarkable how quickly things are happening."
You wouldn't expect polished, error-free performance from a two-year-old. So don't expect the same of AI.
What to do instead: You can and should use AI for productivity and efficiency at work—just remember to treat it like a talented but inexperienced hire. Set clear guardrails around its use. Supervise outputs carefully, especially for mission-critical applications. Most importantly, hold employees accountable for the final results rather than letting them publish raw AI outputs. "Don't let employees just do the quick thing and publish something that came straight out of the mouth of Gemini or Claude," Hans Peter says. "Have them do the human layer on top, which is critical analysis and thinking."
Mistake #2: You're waiting for the future, but it's driven by today's experiments.
By the end of his tenure at Everyday Robots, Hans Peter had over a hundred robots working across the Google campuses: cleaning tables, sorting trash, and even making music and dancing. But these breakthroughs didn't happen overnight. Each task took relentless iteration for humans and robots alike, even for something as seemingly simple as sorting trash into the appropriate bins. 20 physical robots practiced sorting trash in Google's facilities—and at the same time, 245 million virtual robots continued learning in simulation. "I like to think of it as if they were dreaming during the night," Hans Peter says, "and then the physical robot wakes up and gets a new model from that dreaming."
AI development isn't about perfect solutions but about continuous learning. Each failed attempt, each minor improvement, each small success builds toward genuine capability. And the same is true for those of us looking to leverage AI in the workplace: the companies waiting for AI perfection will find themselves far behind organizations that started learning years earlier.
Think of your AI journey like you're looking across a deep valley towards a mountain range. You can see the destination clearly: a warm, inviting hut atop a mountain representing AI's full potential. But reaching it requires navigating what Hans Peter calls an "alligator-infested valley" full of obstacles, wrong turns, uncomfortable challenges, and probably mosquitoes. "It's going to take some effort to get there, and some tricky navigation."
What to do instead: Set guardrails, yes, but don't let that stop you from getting to know the tools. Start experimenting now, even if your first attempts feel clunky or incomplete. Create small pilot programs where failure won't damage your business but success could provide meaningful value. Treat every instance of AI implementation as a learning opportunity rather than a finished solution. And don't restrict employees from using AI tools—instead, teach them to use these tools effectively while maintaining critical thinking and quality control. "Have your business become a learning organization that learns how and where AI fits into your daily workflows," Hans Peter says.
Interested in booking this leading keynote speaker on AI and the future of work?
Learn more about Hans Peter here, and get in touch with us to book him to speak at your event!
Hans Peter sat down with us recently to explain why AI should take our jobs, and why the robot revolution "can't come fast enough." Watch his episode of our Lavin Voices podcast:
https://youtu.be/x0x9EPIP7Ak
