The Next Big Leap: AI and Personal Agents Transforming the Internet as We Know It
AI assistants powered by large language models are set to revolutionize how we navigate the internet. These smart agents could handle searches for products, services, and content, tailoring results to our needs. As AI agents are poised to become our main digital gateway, websites will likely evolve to prioritize AI-friendly design over human interfaces, reshaping the online experience.
These disruptive changes and their implications were at the heart of the fourth episode of the M01 Rocketeers podcast, in which we sat down with Tarmo Toikkanen, Senior Lead at the Data and AI Program in Sitra, and Pasi Ilola, Founder and CEO of Starcart.
The discussion began by listening to a future scenario—a research-based yet fictional story describing a potential future state. You can find the story and the company introductions at the end of this post.
AI agents redefining advertising
While the tech world has only recently begun buzzing about personal AI agents—tools capable of browsing countless websites to quickly find what humans want at the best price—Pasi Ilola‘s company, Starcart, has been pioneering this space for quite some time.
“The terminology for what we do is still evolving,” says Pasi. “What do we even call businesses like ours? The term ‘AI agent’ is actually the first way we’ve found to describe our work that resonates with people in the tech space. While it might not yet be fully understood by the broader public, it’s a step toward defining this entirely new field.”
Starcart’s early success showcases how this emerging AI agent model is shaping the future of online commerce, making it easier and faster for consumers to navigate the overwhelming digital landscape. Pasi added, “I’m a firm believer in the AI agent model for transactions. If an AI agent can handle all the heavy lifting—comparing products or services and completing the transactions for you—why wouldn’t you embrace it?”
Sarita Runeberg observed, “The AI agent model is fascinating because it’s likely to reshape the advertising landscape significantly. If agents are browsing countless sites on behalf of users, the focus could shift away from well-known stores with large marketing budgets. This change might impact resellers and place greater emphasis on brands advertising themselves directly.”
Pasi shared his vision for the future, predicting an agentic layer atop all online commerce. “The big question,” he noted, “is which companies will dominate this layer. There’s going to be intense competition as the opportunities are vast. Will the leaders be tech giants like Google or OpenAI, or new entrants like Starcart, or perhaps both? I believe this will be the major battleground of the future.”
This shift also raises pressing questions about advertising and the use of personal data. If AI agents establish this “agentic layer,” effectively replacing traditional ad platforms, what will the future of advertising look like? What systems, actors, platforms, and business models will emerge? Online commerce has long relied on collecting consumer data to fuel targeted promotions, but this dynamic may change significantly.
Pasi reflected on this uncertainty: “Where will businesses advertise their products or services, and will advertising even be possible in the same way? That’s still completely open. I’m not sure about models like earning by watching ads, but much of it hinges on the role of private data—whether we’ll stick with the current model of data aggregation or move in a different direction.”
If personal agents were to safeguard our private data, sharing only anonymized fragments with sellers, we might experience a new era of hyper-targeted, entertaining, and relevant advertising. Pasi reflected on this possibility: “If we introduce this agent layer between consumers and service or product providers, the agent layer could take on the role of retaining and protecting your data.”
Sarita raised an important counterpoint, questioning whether this would simply recreate a third-party system similar to the one we have today. “I think the collection of private data is something we can’t entirely avoid as long as we operate within this market-driven model,” she noted, emphasizing the complexities of balancing privacy with the realities of commerce.
Tarmo Toikkanen posed an intriguing question to Pasi about the advertising principles at Starcart:
“Pasi, you’re essentially providing consumers with a personal AI agent—a service akin to having a personal shopper. Does this make you the third party in the transaction, or are you the first party while the store becomes the third party? And when, at some point, brands like Nike approach you with offers to promote their products for a fee, how would you handle that?”
Pasi clarified Starcart’s position, explaining that legally, they act as facilitators of transactions between consumers and stores rather than resellers. “We operate with a model that, to our knowledge, hasn’t been done anywhere else. Our agents buy goods directly from websites on behalf of consumers, and we haven’t seen a similar execution elsewhere.”
He emphasized that Starcart doesn’t rely on advertising, a conscious decision to preserve the user experience. “We chose not to ruin our platform with ads, unlike some other services.”
The question of commercially neutral AI agents will undoubtedly become more widely discussed in the future. As AI agents evolve into our personal helpers and trusted advisors, it’s crucial that we can transparently understand the logic and principles guiding their actions. Transparency will be key to ensuring these agents remain impartial and serve our best interests without hidden agendas.
The future of social media: interoperable apps?
Social media platforms are on the brink of major disruption as concerns rise over data privacy, algorithmic biases, mental health impacts, misinformation, deepfakes, and ad-heavy content. Filter bubbles created by algorithmic loops will further narrow users’ exposure to diverse perspectives.
Our social media feeds are increasingly filled with both commercial and AI-generated content, leading some users to disengage. Sarita pointed out that platforms have encouraged sellers to create ads that resemble user-generated content to better connect with audiences.
Pasi noted, “The story format is the last major innovation in the entire online advertising space.” He also highlighted that we’ve reached a pivotal moment in the search engine and social media landscape, where competition is fiercer than it has been in the last 20 years. With major platforms like Facebook and X seeing declines—at least according to media reports—this surge in competition is ultimately a positive development, sparking innovation across the industry.
Tarmo highlighted an intriguing trend: “To my understanding, Facebook is currently experimenting with AI-generated content in certain regions. The reason? People aren’t posting enough anymore. So, now AI is stepping in to create posts that your friends might write, keeping users engaged. In many ways, it’s a fascinating yet bizarre development.”
A shift toward subscribing to social media platforms and content providers that prioritize data privacy could emerge as a significant future trend. When asked about the potential impact of the EU’s Digital Markets Act (DMA), which mandates platform interoperability with third-party apps, Tarmo expressed strong support, emphasizing that people should have the freedom to choose the platforms they want to use and seamlessly message friends across different networks.
Tarmo compared the DMA to the GSM roaming regulations of the 1990s, noting that just as telecom operators were once required to allow international mobile calls despite initial resistance, the DMA ensures that platforms like WhatsApp will have to provide APIs to allow interoperability with apps like Signal or Telegram.
He reflected, “Back in the ’90s, European legislators mandated that telecom operators enable people to make mobile calls across borders, despite claims it was technically impossible. Now, roaming is a standard feature, and nobody thinks twice about it. Interoperability is the beautiful future we’re aiming for. How and when we get there is, of course, another question, but that’s essentially the vision behind the DMA.”
Pasi reflected on how we’ve been building siloed messaging spaces for decades, with no new open communication platforms emerging since the GSM era. “So, will anything come out of the `Fediverse`? If major players like Meta were to join this space and make content easily shareable, perhaps it could take off and become the new standard.”
Sarita added that, compared to the ‘90s, we now live in a much more complex world. Today’s messaging apps offer a lot more functionalities and a multitude of nuances.
Tarmo shared how he had predicted the interoperability of instant messaging a decade ago: “I even made a 10-minute pitch presentation about it on YouTube. It hasn’t happened yet, but not because of technical limitations—it’s more about the reluctance of companies to make it happen. As with any regulation, the challenge is that incumbent companies will do whatever they can to block it.”
He continued explaining that GDPR is a great example of that:
“The intent behind GDPR was to make it worthwhile to develop privacy technology, like an AI agent that could protect your privacy while providing services. But what happened? American companies made it as cumbersome as possible to reject cookies and crafted their privacy policies as American as possible. I remember in the first year of GDPR, a Finnish dairy company printed its privacy policy on a milk carton, in the Savo dialect, no less—just seven sentences. And it was totally valid, legally speaking.”
Tarmo shed light on the infamous cookie banners that many of us dread, explaining that regulators likely envisioned browsers handling these settings natively, allowing users to manage their preferences directly. “Interestingly, there are now browser plugins that automate this process, rejecting or accepting up to 90% of cookie banners for you,” he shared.
Manually-crafted digital content: a renaissance amid AI saturation
The digital media landscape is becoming increasingly saturated with AI-generated content, raising concerns about both quality and the growing challenge of distinguishing fact from fiction. Pasi shared his perspective:
“I think we’re heading into a world where it will be nearly impossible to tell whether pictures or video content are AI-generated. I don’t know what that will mean for society when you can’t trust anything you see on the internet anymore. It used to be a joke—‘it’s from the internet, you can’t trust that’—but now it’s becoming reality.”
There was a time when photos could serve as evidence—“pic or it didn’t happen.” However, today’s AI-generated images are so convincing that most people can’t discern the difference.
Pasi went on to highlight the societal risks this poses: “I think it’s only going to fuel the atmosphere of conspiracy theories, where people stick to their own versions of reality. The internet already enables individuals to build their own bubbles of truth, consuming only the news they choose to believe in.”
Sarita agreed, emphasizing how tight these bubbles have become: “Back in the day, even if you had certain beliefs or ideas, you still spent most of your time interacting with other people. But now, with the rise of remote work and people spending more time alone, often in front of their computers, it’s entirely possible to live almost exclusively within your bubble. Your world becomes shaped by the content you consume, with fewer external influences to challenge it.”
The increasing trend of people forming relationships with AI raises significant concerns for both individuals and society at large.
Sarita highlighted that some social media platforms have already started experimenting with tags to identify AI-generated content. Pasi shared his forward-looking perspective: “I foresee that, very soon, we’ll see either government-mandated or market-driven standards for labelling AI-generated content—or perhaps labelling the opposite, certifying content as non-AI-generated and authentic.”
Tarmo chimed in with a witty reflection: “Rare, handcrafted by a human. Or maybe something like digital, manually created media?”
Pasi agreed, imagining a future where most content will be AI-generated, leaving authentic human creations as the exception. “Yeah, 98% of stuff will likely be AI-generated, and the rare handcrafted pieces will carry labels certifying their authenticity. I wouldn’t be surprised if this shift happens in the next 24 months.”
Reflecting on the future of entertainment, Pasi posed an intriguing question: “News is a different question altogether, but will people care if music, games, or movies are AI-generated? And what will happen to the George Clooneys and Nicole Kidmans of Hollywood?”
Tarmo offered a contrasting perspective, drawing on his own experience introducing his kids to ‘90s TV shows like Star Trek: The Next Generation. He noted how his children appreciated seeing real actors performing on physical sets. “When we watch something from 2023, they expect it to be flashy and over-the-top, but it’s not really that impressive to them. On the other hand, when they see an old movie, and realize there were 500 people on set creating something tangible, they’re genuinely amazed.”
Pasi couldn’t help but smile as he said, “I just want to tip my hat to you, sir, for showing Star Trek: The Next Generation to your kids. That’s a big service to humanity. I love that show. I grew up on it.”
He continued, drawing a connection to the AI discussion: “And actually, that brings us back to AI agents. On these shows, the crew would talk to the ship’s AI-powered computer, and in that iconic female voice, it would always comply, handling complex tasks with ease. It feels like we’ve reached a point where the sci-fi dreams of the ’80s and ’90s are starting to feel outdated—because we’re living them now.”
With a wry grin, Tarmo added, “Although ChatGPT disappoints more often than the Enterprise’s computer.”
Running out of training data – more efficient models or licensing fees to creators?
Generative AI models are projected to exhaust human-generated data by 2026-2032, according to Epoch.ai. Relying on AI-generated data for training risks “model collapse,” where recursive training on outputs from earlier versions causes deviations from real-world data. Over time, errors compound, reducing accuracy and increasing bias and inequality in AI, ultimately rendering outputs less reliable and reflective of reality.
Tarmo highlighted a breakthrough with OpenAI’s o1 model: “It demonstrated that you don’t necessarily need to build the next bigger model. You just need to make it think harder, and you achieve the next level of productivity. Other AI labs are adopting the same approach.”
This shift is promising news for developers of smaller AI models, especially those tailored to serve niche purposes, like supporting less common languages such as Finnish or Swedish, to prevent a phenomenon called cultural flattening.
Authentic human-generated content could see a resurgence if AI developers and consumers alike begin valuing and paying for it.
He then pointed to another emerging trend: “There’s a growing race for the remaining untapped data, much of which resides in the archives of large media companies and national institutions. These companies are tightening their grip, negotiating deals, and ensuring they get paid. Increasingly, data is no longer freely available to AI model builders—it must be licensed. It’ll be interesting to see how that plays out.”
Tarmo also noted a concerning side effect of this trend: “A lot of material that used to be publicly available is now going behind paywalls or disappearing altogether. Companies are pulling their archives out of fear that the content might be used for AI training without proper compensation.”
“That’s a very interesting trend,” Pasi observed, adding: “I think the jury is still out on whether we’ll see a model where the creators or owners of original content actually get paid for their work being used to train AI models. I have good friends in the music industry who are actively lobbying for this. They’ve made it clear: they’re pro-AI, open to all of it—as long as they get paid. It’s possible that we’ll see mandated licensing fees for training AI on copyrighted content.”
He concluded on an optimistic note: “What really excites me is that we’re witnessing this monumental shift. It’s arguably the biggest leap in technological development since the dawn of the web—or perhaps, as I like to say, since the invention of electricity.”
“Or sliced bread,” Tarmo quipped, sparking a laugh.
You can find the podcast, M01 Rocketeers, here.
Interested in continuing the discussion? We’d love to hear your thoughts, so please leave a comment on our LinkedIn post.
You can also follow us on Spotify or subscribe to our newsletter to be among the first to find out about our next episodes!
Here’s the scenario – the story from the future discussed in the podcast, followed by the panellists’ company introductions:
FUTURE TALES: SHOW ME THE MONEY
What if you could get paid for watching ads? Not just any ads, but ones that are carefully tailored to your tastes, needs, and interests—ads that inspire, uplift, and even entertain!
If you’re one of the savvy consumers who opted to pay for social media to gain control over your data, you’re probably used to an ad-free experience. And in today’s world, where most apps are interoperable, you only need one platform to stay connected with your friends. You might even be paying for every piece of content you consume based on recommendations from your AI agent—be it news, entertainment, or education. But with fewer ads in your life, have you noticed you’re missing out on some of the latest trends and innovations?
What if you could change that and get paid in the process?
Here’s how it works: ask your AI agent to connect to the free TD2C app, a platform backed by the European Commission’s trade department. Its mission? Supporting small and medium-sized European businesses while keeping your personal data secure.
Once connected, your AI agent can start negotiating with sellers’ agents, letting them know you’re interested in earning money by engaging with their ads. You’ll receive only the most relevant and inspirational ads, carefully selected by your agent. The more time you spend watching, clicking, and engaging with these ads, the more you’ll earn. Even if you’re not a big spender, the cash flow keeps coming!
And the best part? You’re in control. Your agent will only show ads when you want—whether you’re shopping for robotaxi services or the latest beauty products. Instead of simple listings, you’ll enjoy immersive experiences: interactive videos, VR demonstrations, and creative advertorials.
Because your AI agent selects ads based on what’s most relevant to you, companies can only reach you by offering genuinely high-quality products and services. No gimmicks, no third parties, no manipulation—just true value. Plus, your personal data remains completely private and never shared with third parties.
Get inspired, stay in control, and start earning—on your terms.
Starcart is the AI Agent for Online Shopping, and the easiest way to shop online. Starcart allows you to buy goods from 1,000+ online stores with one shopping cart, saving you the trouble of finding and buying the things you need. As a Starcart customer, you automatically get the cheapest prices for 15+ million products, plus discounts from the store price, making Starcart cheaper than the cheapest store. Powered by intelligent AI agents, Starcart’s fully automated platform allows real-time purchases to any online store. With 5M€ in funding, Starcart currently operates in Finland and Sweden, and is about to expand across Europe. Try Starcart out at https://starcart.com
Sitra is Finland’s fund for the future. Our mission is to promote the well-being of Finland and accelerate economic growth within the limits of nature’s carrying capacity. We anticipate the future, innovate and renew society – and we take our vision seriously and work in any way we can to achieve it. We base what we do on expertise and foresight. As an agile, experienced, future-oriented organisation, we have a wide variety of tools at our disposal to identify the need for change and to make it happen. Our role as a future-oriented fund involves creating preconditions for reform, spurring everyone towards making a change and providing opportunities for co-operation. Sitra is an independent fund that reports directly only to the Finnish Parliament. Our current challenge areas are sustainable economic growth, data and AI, democracy innovations, and the sustainability transition. Learn more: https://www.sitra.fi/en