- AI is dismantling the internet’s old attention-based model by making high-quality content cheap and abundant, shifting value toward trust, authority and unique data, Deloitte’s U.S. Chief Innovation Officer Deborah Golden tells The AI Innovator.
- Algorithms are the new gatekeepers, forcing companies to optimize for machine understanding first while still cultivating human trust and connection.
- Legacy systems are the biggest barrier to progress. Businesses must modernize infrastructure, embed trust and accountability, and think beyond efficiency to unlock AI’s full potential.
AI is doing more than powering new apps or automating office work. It is fundamentally rewriting the way the internet works – changing how value is created, how trust is earned and how companies must compete. That’s the view of Deborah Golden, Deloitte’s U.S. chief innovation officer, who said society will have to rethink decades-old assumptions about how the digital world operates.
“It is fundamentally changing the economics of the internet,” Golden said in an interview with The AI Innovator. “If you look at what the principles were that governed the value of the internet for the last two decades, it was scarcity, attention, click rates – and that’s being systemically dismantled.”
For the past 20 years, the internet’s business model relied on capturing attention through content. Back then, content was costly and relatively scarce – it takes time and money to create – and companies made money by using the content to grab clicks and sell ads. That world is vanishing fast.
“The very concept of content and scarcity has evaporated,” she said. “High-quality text, image, code – right now, that could be created at near zero marginal costs.”
AI has made it so cheap and easy to create high-quality text, images, and even software that content is no longer scarce – it’s infinite. That collapse of scarcity undermines the attention-based economy the web was built on.
“The model itself has been completely inverted,” Golden said. “We’re moving from what I like to call ‘attention economy’ to ‘authority economy.’ The strategic questions are no longer, ‘How do we get seen,’ but ‘Who should we be,’ and both ‘Who should we believe and how do we get that attention?’”
The ad click is losing power
The core metric of the web – the click – is on the wane.
“The central monetary event of the web is losing its power – the ad-driven model that predicated it forever is how to actually monetize that journey,” Golden said. “AI, as we think about it, is delivering the destination instantly. … It’s collapsing that journey. We used to take hours searching the web, looking for things, and now we can get there in a matter of seconds.”
“So the challenge isn’t just about finding a new model. It’s rethinking that value creation. It’s rethinking that journey. It’s rethinking that path of discovery,” Golden continued.
As content becomes a commodity, “how do we think about data differently?” she posited. “The ultimate asset is now data, which historically has been a commodity as well. Also, what is that unique insight or that unique access that we’re now providing?”
With AI able to flood the internet with content and at nearly no cost, the new scarce resource becomes data and insight – not just raw data, but unique, high-quality information that can train AI models or power valuable outputs.
This shift means companies must change their approach in the age of AI. Instead of focusing solely on getting clicks, they now need to focus on being trustworthy and credible. AI is forcing the internet to evolve from a place where attention was everything to a place where trust, credibility and unique data are the most valuable things.
Winning over algorithms before people
The way companies market themselves online is also changing. For decades, they focused on persuading human audiences. That still matters, Golden said, but now they must first convince AI systems – the algorithms and agents that increasingly act as the gatekeepers of information.
“Online interactions are changing,” she said. “For a century we’ve marketed to the human psyche. I think that still doesn’t change. … (But) first we’ve got to win over the algorithm – and our new customer is an agent.”
Before a company’s message can reach people, it must first pass the machine’s test. Trust is no longer just an emotional response – it’s also a technical standard. That means a company’s information must meet the algorithms’ criteria for being accurate, relevant and high quality. Clever marketing slogans won’t sway an AI agent. Instead, you need to make sure your content is structured, machine-readable and technically credible.
“Trust is being deconstructed into a technical proof,” Golden said. “The agent doesn’t care about your clever branding, so you’ve got to figure out how to win the API part of it. How do you get past the API? Loyalty is perhaps no longer just an emotion to be won. There’s a specification across that API.”
People remain critical to the process, not just for sentimental reasons. When everyone uses AI to become more efficient – to do things faster and cheaper – humans become the differentiator, providing the competitive edge.
“The human aspect is, ‘how do I replicate that belonging part, that human, deep, collaborative, connective intelligence of community?’” she said.
This means businesses now need to win on two fronts at once – first, by making sure AI agents recognize and choose them, and second, by nurturing the deeper human connection that builds brand loyalty.
How to get out of ‘pilot’ AI stage
Many organizations are learning how to capture the AI agent’s attention by making their digital presence more understandable to machines. Golden calls this “machine legibility” – structuring data and information so AI systems can recognize, interpret and use it effectively.
“How do you structure your digital DNA so that the agent doesn’t just learn and find what it needs to find, but is fundamentally understanding and validating the market, the product, the algorithm, the fingerprint?” she said.
But the biggest barrier isn’t the AI itself – it’s the old technology companies are trying to run it on. Even when companies start to adopt AI, they get stuck because their legacy systems can’t adapt to its unpredictable, dynamic nature. Unlike legacy systems, AI learns, adapts and makes decisions based on patterns in data, not fixed rules.
“Original code is built off of zeros and ones … and historical code (consists of) if-then statements; it’s predictable. AI in and of itself is unpredictable and (legacy) systems aren’t built for unpredictability,” she said. Businesses are “able to navigate the agent world but where things get stuck in the supply chain is when their existing systems aren’t able to pivot to unpredictability.”
That’s why many organizations are stuck in the ‘pilot’ stage of AI. They can’t scale those projects across the business because the existing infrastructure was built for a rule-based world.
The solution isn’t necessarily throwing everything away and starting from scratch – but it does require hard work to rebuild key parts of their systems so they can handle AI at scale.
“A lot of it is how much and how far they’re willing to go to actually reconstruct those systems,” Golden said. “This is not about a complete rip and replace, but it is about what hard work are they willing to do and take on to actually reconstruct certain systems.”
To make real progress, companies must be willing to rethink how their systems are built – from the tech architecture to partnerships – not just layering AI on top of old systems.
“Are we willing to recreate net new infrastructure, net new architecture, net new suppliers?” she said. “Where they’re gaining ground is looking at building new ecosystems, looking at new public-private partnerships, looking at relying on things that aren’t built within their four walls.”
The human + AI internet
Golden rejects the idea that the internet will split into separate worlds for humans and AI. “For me, it’s an ‘and,’” she said. “There is always a human element to trust.” While AI can perform work that is faster, cheaper and more reliable than humans, people bring capabilities that AI cannot copy. “AI automates the complexity, but it can’t replicate the artistry and empathy of true connection,” she added. “At least not today.”
The future isn’t about choosing between humans and AI; it’s about integrating the two – letting AI take care of the complex, repetitive work while people focus on the deeply human parts that technology can’t replace. “Even with agentic AI, I think we’ll have spaces where humanity is our advantage,” she said.
Beyond clicks, AI will also blur the lines among online activities that used to be separate. “AI is a solvent dissolving the traditional boundaries between online activities,” Golden said. “The ideas of shopping, entertainment, socializing are separate – they are becoming obsolete.”
Golden cited three areas that are changing:
- Online shopping will become a passive activity. “We won’t go shopping; we’ll simply have,” she said. Instead of people having to actively search and choose what to buy, AI will predict what they need and order it.
- Entertainment will become a market of one. The content viewers get will become more adaptive – for example, ordering a comedy won’t just deluge them with more comedic films. AI could tailor stories, music and experiences to each person in real time – even changing the content based on one’s reactions or the environment.
- Social interaction becomes “profoundly and visibly” augmented by AI. Examples include tools that enable real-time translation or social “coaching” could help people communicate better.
These changes are advancing quickly. “I think it’ll happen faster than five years, candidly,” she said. “The more people start to live in that world, the quicker we’ll see that adoption happen.”
Security, compliance as competitive edge
Securing systems in an AI-driven world requires a complete rethink as well. “We can’t simply play defense,” Golden said. “We have to re-architect our digital environment so that truth has a natural advantage.”
Golden laid out three big changes organizations need to make:
- Build authenticity into the data itself. Instead of just adding warning labels, companies should embed provenance – proof of authenticity – inside every file, piece of content or dataset. “Authenticity for me cannot be an afterthought,” she said. “It must be a feature of the file itself. And we’ve not really seen that yet.”
- Defend across all types of media, not just text. Most cybersecurity tools today focus on text – scanning documents, emails or code. But disinformation and attacks now come through text, audio, video and images all at once. That means security systems need to check and cross-verify all those types of data simultaneously and in real time – within milliseconds.
- Design for accountability, not perfection. No system will ever be 100% secure, and trying to make it so would cost billions. What matters more is how a company responds when something goes wrong. Being open about failures and fixing them quickly and transparently builds more trust than pretending everything is perfect. “How a platform responds when it fails … is going to be more important than the failure itself,” she said.
As for complying with a hodgepodge of regulations globally, Golden urged companies not to view rules as a burden but as an opportunity to lead.
“The savviest leaders see regulation not as a constraint to be managed, although often a challenge, but as a market to lead,” she said. “You can look at it as something that is a negative, or you can look at it as something that is a competitive weapon to win the trust of customers and their agents.”
Golden gives three key pieces of advice in this area:
- Build around universal principles, not specific rules. Laws around AI are still evolving – many haven’t even been written yet – and they differ by country, state and even industry. Instead of chasing every new rule as it appears, companies should design their AI systems around a core set of universal principles such as fairness, accountability, and transparency. If those are built into the foundation, they’ll be much more adaptable as specific regulations shift.
- Treat AI risk as a fundamental enterprise risk. Too many companies see AI risk as a siloed legal risk, but it’s much bigger: it carries capital, brand and strategic risk. That means it belongs at the highest level of decision-making – in the boardroom – and should be treated as part of the company’s core strategy, not a standalone project.
- Use compliance to build trust. Compliance shouldn’t just be a shield to avoid fines – it can be a tool to win trust from customers and from AI agents themselves. A company that proves its systems are fair, reliable and transparent will stand out in the market.
Business leaders’ biggest blind spot
Golden said the most common mistake leaders make is focusing too much on the AI tools themselves and not enough on the systems that support them. “The strategic blind spot is mistaking the tool for the terrain,” she said. “The real bottleneck to scaling AI isn’t the algorithm. It’s the unglamorous foundational work to fix the data governance, to fix the platform modernization, to fix the process re-engineering.”
Even the most powerful AI models won’t deliver results if the company’s data is siloed, its infrastructure outdated, or its processes broken. “It’s like trying to install a supercomputer in a house that has faulty wiring and leaky pipes,” she said.
Her advice to business leaders: Do the hard work of fixing foundational systems, and think beyond just efficiency. “Operational efficiency for AI is table stakes,” she said. “I think being able to look beyond the operational efficiency and leveraging AI for net new business models … – they need to be thinking about that.”
Golden believes AI is more than a tool – it’s a way to reimagine the future. “Technology at the end of the day, it’s a mirror,” she said. “It reflects our values and our priorities, but the profound challenge is really how can we look to reimagine truly complex problems with net new answers?”
She encouraged business leaders to be bold. “It’s not a moment for incrementalism,” the CIO said. “This is a moment for bold, future-forward vision. Certainly, we’ve got a lot of challenges in front of us. We’re obviously navigating a market shift, but we have an opportunity to rewrite the next chapter of the economy.”
