By Jacques BughinÂ
Agentic AI is transforming the digital economy, replacing traditional search with intelligent execution. In this article, Dr. J Bughin presents a five-step framework that challenges binary narratives and reveals how businesses can adapt strategically. The future of monetization depends on navigating this shift with clarity, precision, and economic insight.
Summary
In the age of agentic AI, where artificial intelligences no longer simply respond but execute actions, traditional business models – such as Google’s – are being profoundly challenged. Managers need to find a clear, unambiguous answer to this question. This paper proposes a five-step analytical framework for understanding this rupture and deriving well-founded strategic decisions from it. Applied to the case of Google, this process reveals that:
- The current model is based on monetizing traffic via the SERP; however, it is structurally fragile. If agents bypass the SERP, disintermediate search, and, above all, reduce the value of the click, it could undermine the system.
- On the demand side, agents promise a growing search market by improving conversion rates and making previously ignored queries monetizable. This attracts new entrants and allows Google to cannibalize itself.
- Competition is evolving: according to game theory, a new equilibrium should quickly emerge between Gemini (Google) and the integrated advertising of LLMs, and at a pace faster than that driven by the adoption of agentic AI.
- The value will shift to execution. Google must therefore become an orchestrator of agents, not just a search engine.
- An interesting game balance is not an all-out battle, but a differentiation model where agents focus on industries (verticalization) while Google becomes more integrated, from Google Cloud, and Chrome to Google workplace and Gmail.
This framework makes it possible to move beyond binary reactions and approach transformation in a structured, rigorous, and economically sound way.
Introduction
The rise of large language models (LLMs) and agentic AI has catalyzed a wave of speculation about the end of search as we know it.
While popular discourse is dominated by two opposing conjectures (“Google will be wiped out” versus “LLMs are not profitable”), the future is more complex and requires a structured analysis of how search has been monetized, as well as a theoretical assessment of the evolution of search and monetization in the context of the evolution of AI.
Using models based on the microeconomics of search as well as the type of strategic interaction (static and repeated games) between Google and attackers such as Open AI, Perplexity and others, we try to offer a more powerful framework that not only explains the transformation underway, but debunks simplistic narratives (Table 1). Managers may find this framework important when they’re looking for more solid answers about what to do in AI transformation.
The Five-Step Framework
Table 1: Navigating AI confusion
| Step | Action | Objective |
| 1. Understanding the business model | Analyze the current revenue model of the dominant incumbent operators | Establish an economic base and structural dependencies. |
| 2. Evaluate actual disturbances | Identify how attackers modify monetization channels. | Determine the depth and extent of disturbance. |
| 3. Understanding the economics of demand | Understanding how new games change demand. | Assessing the market’s future – up or down |
| 4. Add supply-side economics | Understand the logc of the new equilibrium, dynamically . | Assess the intensity, stability and type of new competition |
| 5. Rebuild with aggregates | Analyze supply and demand . | Find new results and deduce actions/key assets/playing fields |
1. Understanding the business model
Google’s sponsored links, which manifest themselves primarily through search ads, are the cornerstone of its revenue model. By 2024, Google’s advertising revenues will reach around $240 billion, with search ads contributing around $175 billion, or 57% of the company’s total revenues.
While these figures underline the significant value of sponsored links within the Google ecosystem, Google has other revenue streams, such as Google Cloud, which will benefit from the deployment of AI. In addition, advertising revenue is driven by three fundamental levers: the immense volume of global search queries, the subset of high-intent queries that trigger paid ad auctions, and the Google platform’s control over the search engine results page (SERP). By dictating page structure and bidding rules, Google effectively monetizes attention and intent on a massive scale.
However, this dominance comes with inherent vulnerabilities. Firstly, the vast majority of queries – around 80% – are not commercially monetizable. They respond to needs for information, navigation, or exploration. Secondly, SERPs themselves are saturated and increasingly commoditized, with search engine manipulation diluting the value. Thirdly, the user must always act outside the Google interface to accomplish tasks, creating friction in the user experience. These limitations constitute the structural exposure of Google’s traditional model.
2. Assessing real disturbances
-
The impact of AI: GenAI, but above all, agentic AI
LLMs clearly change the structure of search by reducing the need for links (direct answers) and reducing navigation (multi-click paths become a single prompt). With LLMs, over 60% of queries are now informative or intent-driven, which is ideal for AI-generated answers. Users interact with summaries and don’t click on links, reducing volume for Google
The other danger is the collapse of traditional ranking logic, as the concept of “#1 ranking” is replaced by being quoted, summarized, or cited by LLMs. The implication is that the ranking value that increased cost-per-click disappears and pricing power is reduced.
Although initially limited to synthesis and dialogue, the integration of agentic AI considerably broadens the scope of disruption. With the emergence of single-agent systems, a single AI entity can autonomously perform discrete tasks – for example, booking a restaurant, sending an e-mail, or initiating the drafting of a document – without human intervention. Multi-agent systems go further: they break down complex workflows into sub-tasks, coordinate APIs, and execute a sequence of decisions on the user’s behalf. In both cases, the agent not only interprets the user’s intention but acts on it, transforming traditional requests into executable commands.
On a large scale, this transition is transforming the very nature of digital search. It replaces the advertising-funded discovery layer with agent-based orchestration, increasing the potential economic value of each query, but also reshaping who controls that value and how it is monetized.
-
Advertising value chain
This evolution is turning the structural microeconomics of search on its head, by orienting it towards the delivery of results. This shift replaces the monetization of navigation (selling advertising space along the way) with the monetization of execution (capturing value at the result level).
But the rise of agentic AI isn’t limited to disrupting search. It’s putting systematic pressure on Google’s broader monetization engine – including display advertising, YouTube content monetization, and even, eventually, a large number of B2B SaaS intermediaries. In display advertising, AI agents bypass banner placement logic by performing tasks directly from the user’s prompt or workflow. In enterprise contexts, agentic AI increasingly disintermediates SaaS categories for which Google (via Workspace, Ads Manager, or Analytics) has monetized coordination or knowledge. When agents plan campaigns, manage CRM entries, or optimize user journeys, they bypass several layers of existing SaaS infrastructure. This creates downward pressure on margins and squeezes the space for traditional marketing and advertising technology.
Ultimately, Google and its AI competitors are converging on a new high-value node: the orchestration layer. This is where decisions are made, actions are initiated, and margins can be captured. Whether powered by Gemini, OpenAI, or specific vertical agents, this layer holds the key to monetization in the age of agentic AI. What search was for information, orchestration is becoming for execution: the critical control point in digital value chains.
3. Understanding the “demand” side of change
An important unknown is how agentic AI will affect the profit reserve. However, microeconomics tells us that the profit pool will be larger due to three factors. Firstly, agentic execution improves the quality and relevance of interactions. Unlike the current model, where most ads are shown to users who are not yet ready to convert, agentic ads can be integrated directly into high-intent workflows. Secondly, agents reduce transaction friction. By reducing the funnel, they accelerate the passage to action. This reduces waste in sales channels and increases results attributable to advertising. Supply-side efficiency encourages brands to bid higher for access to agent-driven engagement.
Thirdly, the long tail of non-monetized queries – previously low-intent, informative searches – can now be captured and transformed into valuable transactions.
These effects are, in principle, multiplicative on the level of return on (search) advertising spent – so it’s not necessary to have a major impact on a single effect – a smaller, but combined, impact is the real crux of whether the market will grow. As these three effects are likely to combine with agentic AI, it’s reasonable to think that the market will be bigger, not smaller, as the technology evolves towards agentic AI.Â
4. Add the supply side of change
-
Why LLM will establish advertising as an additional source of revenue
If agentic AI increases the value per query, it threatens to cannibalize the very mechanisms that fund today’s search giants. For Google, the main concern is that agentic systems will bypass the SERP entirely, cutting off its advertising supply chain. Gemini, Google’s counter-offensive, seeks to preserve monetization while adapting the interface to a query-driven future.
On the other hand, players like OpenAI and Perplexity face an entirely different challenge: most of their users are free. OpenAI, for example, is said to have over 100 million weekly active users, but less than 5% pay for ChatGPT Plus. To maintain the high costs of LLM inference and GPU-intensive infrastructure, these platforms need to monetize the remaining 95% of users.
The strategic logic behind LLM advertising monetization is therefore simple but unavoidable. First, inference costs at scale require offsetting cash flows. Secondly, user payment models are reaching a ceiling – most users won’t pay for general-purpose chat. Thirdly, verticals such as procurement, local services, and SaaS recommendations are rich in intent and ripe for monetized orchestration.
-
Game-theoretic perspectives: Modeling competition between LLM and Google advertising
- Pure strategy equilibrium (Nash solution)
When several suppliers compete, it is important to know whether it is possible to categorize the type of competition that is likely to occur. Here, the tools of game theory, which examine the payoffs to each player based on the movements of the others, are uniquely valuable in assessing possible behavior, now and in the future, based on repeated interactions.
Suppose we model the interaction between Google and LLM challengers as a static, repeated game, and the values of the game (including LLM subscription) in the static model are as follows (in billions of dollars by 2030):
Table 2: Game theory payoff matrix (illustrative)[i]
| LLM : No ad monetization | LLM : Monetizing advertising | |
| Google : Do nothing | (144, 30) | (108, 75) |
| Google: Reinvention by Gemini | (150, 60) | (161, 44,5) |
The payoff matrix (Table 2) shows that LLMs have an incentive to engage in advertising for some choice of Google. Thus, the main idea of game theory is the prevalence of a stable equilibrium, where dominant strategies converge with LLM-mediated advertising (Nash stragie) -and… the total market has grown.
-
Pure strategy equilibrium (Nash solution)
These results apply only to the one-shot game. Let’s assume a more realistic game, where there is uncertainty about the profitability and development of AI by agents, and that the game of interactions between Google and LLMs is repeated over 2024-2030. At this level, the dynamic changes: initially, LLMs stay away from advertising monetization, even though they experiment and gain the trust of users. Gemini is also partially deployed, but not in a head-on situation. As the capabilities of LLMs improve, advertising enters their ecosystems. Google, faced with strong erosion, accelerates Gemini deployment and integrates the new advertising logic into AI agent flows. In the end, both parties compete in the field of agent-based monetization.
This type of game is known as a mixed-strategy game, in which the different players combine several strategies at random to test their best position, and of course, hide their first intentions (Table 3). But this uncertainty disappears and converges towards the equilibrium shown in Table 2.
Table 3: Game frame evolution
- Mixed strategy phase (2024-2026) Dominant play for Google: 60-80% deploy Gemini (to reinvent itself, but avoid total cannibalization of margins); 20-40% delay Gemini (observe user habits, avoid overreaction Dominant play LLM: 40 -70% monetize advertising (Capture initial value in verticals like travel ) 30-60% – increase their footprint (Build trust,)
- Iteration and feedback (2026-2028): Updating beliefs (Bayesian learning on earnings structures) and refining strategies
- Convergence towards a pure strategy (2028-2030) Players commit to pure strategies, with Google fully integrating Gemini into search.
This evolutionary path, derived from game theory, is not innocent:
- Firstly, it means that rational logic should lead to an equilibrium where the new business model becomes dominant for each player.
- This model is evolutionary, not because Google has difficulty executing it, but because it’s more strategically optimal to embark on a mixed strategy. This mixed phase creates a space for experimentation without open conflict. Each party sends strategic signals (e.g., Gemini integration in Android but not in the search home page; OpenAI testing of sponsored suggestions in Pro mode only).
- Even if the game is evolutionary, — it’s fast: initially, there’s already more than a 50% chance that Google will launch into LLM, – this is marginally lower for LLM to launch into advertising, but the probability is far from zero. In 3-4 years, the strategy will lead to a reversal of the dominant business model, while the agentic penetration of AI in advertising and search is not yet dominant – 30 to 40% of customers use it.
This dynamic is the result of a positive loop effect. Increased usage leads to better feedback on the user interface and improved agent quality. Better agent quality reinforces trust and leads to more commercial requests. And if more resources are available, LLMs invest more in model optimization.
This loop has other implications: it favors the first to have a closed-loop infrastructure – so we can expect Google to integrate Gemini into Android, Chrome, Maps and Gmail. New LLMS attackers such as OpenAI or Perplexity could then choose to secure their position as agents in the key workflows of other players competing with Google such as Salesforce’s Slack, Microsoft Teams, or Zoom), thus creating multiple different ecosystems, without aggressive competition favoring the extraction of ROI from customers.
5 Bringing together all the elements of microeconomics
5.1. The metamorphosis of online research
From this perspective, the future of online search is not one of extinction or a struggle for survival. It’s about a metamorphosis where the revenue model will evolve from advertising around discovery to monetization around execution.
Google’s dominance depends on its ability to maintain trust, share relevance, and user flow. LLMs, meanwhile, are set to evolve from high-cost, low-revenue utilities to sustainable platforms. This will require a diversification of revenue sources from subscription to advertising, but advertising that is integrated, not imposed.
5.2. News of Google’s death is greatly exaggerated – But Google needs a boost
Google’s destiny is not binary: death or survival, but it is clear that the business model is set to shift towards agent-based execution, — and that this dynamic will force Google to reinvent itself. The success of this reinvention will depend on several interdependent factors.
The demand effect shows that the transition can be profitable. The loop effect clearly shows that Google must also remain a major player if it is to make a successful transition. The loss of more than 25% of classic search users, who are turning instead to LLMs (outside Gemini) for their searches, means that it may be difficult for Google to maintain its price levels, CPC. Gemini’s reinvention path is also about achieving a leadership position, but primarily in the search agent (not LLM) arena. So, Google’s current platforms will be Google’s best assets moving forward, while Gemini becomes the journey to execute well for Google’s rosy future.
Final Thought
Ultimately, the application of the above approach can be summarized in a Tabelau (Table 4)
Table 4: Summary of results
| Step | Applied to search and Google |
| 1. Understanding business income | – Google earns around $175 billion a year from search ads (57% of total revenue) – Monetization = Query volume Ă— CTR Ă— CPC – Only 10-20% of queries are monetized – Power lies in the platform’s control over SERPs and bidding rules. |
| 2. Evaluate actual disturbances | – LLMs respond directly, bypassing links and SERPs – Agentic AI performs tasks, eliminating navigation steps – Traditional CPC logic is weakened; ranking power is eroded – Platforms like OpenAI/Perplexity intercept high-intent queries. |
| 3. Understanding the economics of demand | – Agentic AI improves performance through better targeting and task integration.- Long-tail queries become monetizable – Funnel friction is reduced→ higher intent capture. – Result: Market expands through improved advertising results. |
| 4. Add supply-side economics | – LLMs must monetize to cover inference costs (subscription ceiling reached) –                                                                                                       Game theory shows that LLMs adopt advertising, Google launches Gemini:                                                                                                 – Competition shifts to agent orchestration (Gemini, Copilot, etc.) – Result: Coexistence in multi-agent ecosystems, no monopoly. |
| 5. Aggregate reconstruction | – Execution becomes the new monetization layer – Google needs to integrate deeply (Gemini in Android, Chrome, Gmail) – The new value lies in agent control, task execution and orchestration infrastructure.
-The speed of model changeover is rapid, and faster than customer adoption – because competition takes place at the margins, to ensure growth. |
Although this synthesis may seem simple, its “tour de force” lies in the fact that it is the result of a comprehensive and detailed micoreconomic analysis. In fact, in times of disruptive technological transformati, – such as the rise of agentic AI – success doesn’t depend on intuition alone, and even less on fear. In times of disruption, the first thing to do is to make sense of change, and develop knowledge for a clear and persistent path of change. The time has come to establish a discipline aimed at building a solid foundation of strategic data. Business leaders and policy-makers need to rigorously model technological trajectories, changes in user behavior, and competitive dynamics. This five-step framework should enable more decisive and credible action to be taken.
Jacques Bughin
