Google Wants AI to Be a Creation Engine, Not Just an Information Engine (Maureen Heymans, Google)
“AI is changing who gets to turn an idea into something real.”
What the printing press did for literacy, AI may do for creativity.
That is the hopeful idea I kept turning over while talking with Maureen Heymans, who works on learning and education at Google across products like Search, YouTube, Gemini, NotebookLM, and Classroom. Google is, whether it set out to be or not, one of the most powerful education organizations on the planet. So I wanted to understand where she thinks this moment could lead, and what kind of learning Google is trying to make more possible.
I came away from the conversation energized.
The most interesting shift in Maureen’s thinking over the last year is that she now sees AI less as an information engine and more as a creation engine.
She told me about teachers, many with no technical background, vibe coding apps for music, history, and cooking. I loved that image. AI is beginning to change who gets to build, who gets to experiment, and who gets to turn an idea into something real.
We also got into the harder questions:
At a moment when students increasingly expect AI to hand over the answer, how much productive struggle will learners actually tolerate, and how do you design for that tension?
What still stands between today’s study companion and a genuinely great tutor?
Can AI help schools assess the things we keep saying matter most, like creativity, critical thinking, communication, and agency, instead of just making the old tests faster?
What should educators do with a growing, and in some cases well-earned, backlash against technology in the classroom?
What I appreciated about Maureen is that she does not talk about AI in education as a race to maximum efficiency. She talks about it as a chance to make learning more active, more ambitious, and more human.
If a company with Google’s scale is going to shape how millions of people learn, I am glad someone with that orientation is helping steer.
On AI as a Creativity Engine
ALLISON: What’s something you’ve changed your mind about in the past year?
MAUREEN: For me, AI’s most interesting use is not as a search engine, but a creation engine. At Google, we saw early on that people were using AI as an information engine—better than search in some ways, because you can use natural language, ask anything, have a conversation. But over the last year, we’re seeing people use AI for creativity more and more.
That’s exciting, because AI is removing the skill barriers to building software. You don’t need to know code, or even like coding, to be creative with the medium. That’s enabling a huge shift in how good solutions get built.
A case in point: Google hosted the National Teacher of the Year event a few weeks ago, and I was floored by what teachers showed us—they’ve been vibe coding some remarkable apps across domains, from music to history to cooking, often without technical backgrounds. Their solutions focus largely on making learning more visual and interactive, like the culinary teacher using AI to create videos explaining the thermodynamics of food science.
ALLISON: I love this thread because our philosophy on education at the studio is grounded in experiential learning. We believe it builds the skills we’ll need to thrive in an AI future. Those skills look like problem definition, hypothesis development, and experimentation. It also makes learning relevant, which increases motivation. AI gives us the ability to imagine something and bring it to life within hours, sometimes minutes. It’s for all of us: teachers, learners, professionals.
MAUREEN: Absolutely. That perspective is actually how I got into learning and education.
During the pandemic, I enrolled my kids at Prisma, an online academy offering project-based learning. I saw the power of that model first-hand; their learning was applied to real-world problems and it built their math, research, presentation, and creativity skills simultaneously. Best of all, their motivation spiked. Since then, I’ve been exploring how to use AI to make project-based learning more accessible to teachers everywhere.
On Google’s Ed Tech Thesis
ALLISON: Suppose Google succeeds wildly in their work on AI and learning. What changes for education when you do?
MAUREEN: AI advocates often worry that AI will only reach a fraction of our population. But one of Google’s unique assets is reach. We’re seeing billions of users come to Google for educational needs—an incredible place to begin.
On user experience, we want to ensure users don’t just come for a quick answer, but stay for deep, curiosity-driven understanding. We want to create opportunities for learners to get active: hands-on projects, conversations about a topic, simulations. The idea is to motivate learners to stay curious past an isolated session and embark on a more connected, collaborative journey. We hope to create a generation of explorers and problem solvers.
ALLISON: What’s a debate you keep having at Google about building AI for learning?
MAUREEN: The perennial tension is always “How much struggle is productive?” We want to help get students unstuck, but without overhelping. That’s such a hard thing to understand, because it’s different for every learner, every experience. Not every struggle is productive, but we also don’t want to remove all the friction in pursuit of efficiency either.
ALLISON: In my view, this is the central question for AI and learning. We’ve known for decades that learning happens when people are uncomfortable—when they are struggling. AI is obviously so good at just providing the answer. Interfaces can’t make the challenge so difficult that learners go outside the environment for answers — the struggle has to be right-sized for their motivation.
Walk us through how Google is experimenting with getting this balance right, and what kind of research is woven into those iterations.
MAUREEN: The holy grail in ed tech is demonstrating efficacy, but that takes time—we also need quick feedback on what is and isn’t working. Finding that balance is difficult, but we are continuously engaged in both: efficacy proof, often partnering with pedagogy experts to define learning principles and create evaluation frameworks, and A/B testing to see how a feature or tool drives outcomes—or even defines one.
On what we’re assessing: we anchor on 21st century durable skills like motivation, agency, and critical thinking. So many school assessments test knowledge acquisition—that’s not where we think the future of education lies.
At Google, when we were building a product called Guided Learning—a Socratic-style tutor within Gemini—we did some testing before launch. Our tests have focused on how the tool encourages deeper reflection, and we saw some early results suggesting meaningful growth in students’ ability to solve related but novel problems on subsequent topics. One unexpected benefit: human tutors reviewing outputs throughout testing reported learning new pedagogical practices themselves. I think that’s just such a fascinating look at how we can learn with the tools in really deep, important ways.
ALLISON: What is the hardest unsolved technical problem between today’s AI study companion and a truly world-class tutor?
MAUREEN: We are getting better at customization as AI improves at understanding a student’s context: curriculum, grade level, reading level, and performance within a session. But we have much more to learn about the root causes of student misconceptions. How can AI know if a student missed an answer due to a silly mistake, inattention, or a fundamental misunderstanding of a prerequisite concept?
We also lack good insight into a student’s long-term learning journey—how they’ve progressed to their current level of skill or understanding. There is significant research underway to clarify longitudinal student progress, both within Google and beyond, and it’s critical to supporting students across years, from school to college to career. Without that understanding, a truly high-quality personalized AI tutor remains out of reach.
On Evolving Assessment for 21st Century Durable Skills
ALLISON: In my view, there is a strong problem-product fit for AI and assessment. What do you trust as proof of student understanding? How can AI help assess the things that matter most?
MAUREEN: I value assessments that get students demonstrating higher-order skills—the application, analysis, synthesis, evaluation, and creation components of Bloom’s Taxonomy—rather than sticking to lower-order knowledge and comprehension alone. When students apply knowledge to novel problems and create novel solutions, they are strengthening critical thinking, creativity, and agency—those 21st century durable skills—as well as contributing to knowledge and comprehension mastery.
AI gives us the opportunity to evolve assessment accordingly. I’d like to see AI used to facilitate more Socratic questioning as students research, interpret, and argue through complex problems. It’s hard for a teacher to do that one on one (although many great teachers do); in this use case, AI supports critical analysis across the project, not just in office hours.
Project-based learning is another good example. At the K-12 level especially, it’s far too time-consuming for teachers to create deeply personalized learning programs for each learner and each project, often across multiple disciplines. I’d love to see students taking the lead on creating projects, collaborating with teachers to design the learning pathway—another great use case for AI.
Finally, AI excels at feedback. Teachers give incredible feedback, but it takes time—often nights and weekends—and fatigue is real. It’s not reasonable to expect teachers to provide in-depth feedback on every piece of work a student generates. AI accelerates that process effectively, and when you compress time-to-feedback, you often get better, higher-order learning.
ALLISON: Those are great examples, and good ways of thinking about assessment-as-learning and learning-as-assessment—ultimately indistinguishable from one another, which is of course the dream of formative assessment.
What are your thoughts on summative assessment—high-stakes tests finely tuned to be reliable, valid, and unbiased, where comparability is essential not just between students, but across cohorts over time? Summative assessments determine how incentives and funding flow. What’s your vision for pushing that forward with AI?
MAUREEN: Standardized tests assess content knowledge, but not the skills we’ve been describing. AI could be powerfully leveraged to enable a more diverse set of formats beyond multiple choice—open response questions that assess students’ problem-solving processes and creative thinking. That’s hard to make consistent if assessed by hand; it’s doable with AI.
My hope is that by aligning summative assessment with higher-order skills, we create the conditions for better learning holistically. When I talk to students, I find that they want to learn and are curious by nature. We need to teach them how to use AI to pursue that curiosity—and schools, teachers, and parents may also need to embrace how students are already doing this for themselves.
ALLISON: It strikes me as unfortunate that just as we’ve developed tech-enabled superpowers to make learning more experience-based, more agile in cultivating creativity and agency—more human—we’re seeing a powerful backlash from parents to technology in the classroom. To a certain extent, it is justified: much of ed tech is not designed with these outcomes in mind. A decade ago, time spent using technology at school was a positive metric; that’s not what I’m seeing in national headlines or my own community anymore.
How are you personally squaring your focus with this growing skepticism toward technology, and especially Big Tech, in the classroom?
MAUREEN: Technology is not the solution for every learning moment, just as efficiency is not the goal for every learning experience. But AI can make many components of learning more efficient. That leaves more time for those important learning activities that require patience, endurance, and connection with other learners. At Google, we work with teachers often to put tools within the right process—and we learn quite a lot from them in doing so.
Just last week, I visited a design-thinking course at Design Tech High School in Redwood City. Students were working on a project to understand their user, and I was curious how the teacher was encouraging AI use at different steps. She wanted them to start brainstorming on sticky notes, by hand, before relying on AI. Once they’d done that first step, they could use AI to elaborate, get feedback, research, and prototype.
The sequence really matters: this teacher avoided letting students offload right away, instead using a collaborative, analog medium that’s still highly effective at getting people thinking together. And students were able to create a real solution in a fraction of the time—accelerating their overall learning about design too.
Before swearing off ed tech, I would encourage parents to get curious about how technology is used at their kids’ school. Ask questions, be open to teachers’ innovation with the tools, and take the time to understand what and how your child learns with technology. Then take that understanding and ask for the tools—and more importantly, the outcomes—that you want for your child.
A great way to get started: check out Raising Kids in the Age of AI, a podcast we produced in collaboration with aiEDU; it covers career preparation, responsible AI use, being a citizen in an AI-powered world, and how educators are using the tools right now.
One Small Signal
ALLISON: What’s a small signal to which we should be paying more attention?
MAUREEN: I’m inspired by how many young people are becoming early adopters of AI for exciting, high-value things. I love seeing kids with zero coding background getting remarkably creative with the tools—becoming the CEOs of their own businesses, building fundraising apps for causes they care about, using AI to create content for a YouTube channel or vibe coding their own interactive adventure game.
It’s not happening evenly, or everywhere. But I’d love to empower more students down this entrepreneurial path, adopting the experimental mindset that lets them unleash their imagination and develop something truly novel. It’s exciting to see these technical barriers collapse.
ALLISON: Motivation peaks when we build something in the world that creates value for others. Whether it’s traditional tech entrepreneurship or small business, Main Street entrepreneurship, or the gig economy, it feels good to create something and see people use it to make their lives better. Young people are seeing this moment as an opportunity to do that.
MAUREEN: I believe in the power and potential of young people’s ideas: I find them to always be the most creative thinkers. They are willing to challenge the status quo and to ask for the things they want, even what we all collectively need. I can’t wait to watch this generation of problem-solvers take the wheel.