For decades, business strategy has implicitly assumed a simple fact: customers are human. They search, compare, decide, and buy. Agentic AI quietly breaks that assumption. As software agents increasingly act on behalf of individuals and organizations, a growing share of economic decisions is no longer made by people, but by systems operating within their constraints. This shift does not just change how firms execute; it changes how markets work. In this post, I explore five strategic implications of a world where firms are no longer selling primarily to humans, but to agents that decide for them.
From human customers to agent customers
A growing share of economic transactions will no longer be initiated, evaluated, or executed by humans, but by software agents acting on their behalf. This is not a distant future scenario; it is already visible at the edges of the economy. Procurement bots compare suppliers, travel agents assemble itineraries, pricing algorithms negotiate with pricing algorithms, and personal assistants increasingly act as intermediaries between intent and execution. In these interactions, firms are no longer selling directly to people. They are selling to decision proxies.
One early but telling signal comes from journalism. In mid-2024, the Associated Press launched AP Intelligence, a new unit designed explicitly to serve the machine-readable internet. The message is subtle but important. Journalism is no longer only about producing stories for human readers; it is also about producing structured, reliable facts for systems that will retrieve, recombine, and act on information. AP now takes roughly 5,000 pieces of daily content and delivers them with comprehensive metadata, organized in a consistent, machine-readable schema. The value is not just the text, but its structure.
This reflects the emergence of two parallel internets. The first is the one we know: narrative, persuasion, context, writing for humans. The second is navigated by machines: APIs, schemas, metadata, retrieval pipelines. In this second internet, the act of writing is no longer the scarce resource. What matters is accuracy, consistency, and formalization. AP’s existing partnerships with OpenAI, Google, and Microsoft suggest where this is heading: journalism not only as content, but as infrastructure powering retrieval-augmented generation systems, model training pipelines, and the tools people will actually use.
The nuance matters. Humans do not disappear from the picture. They still define goals, preferences, and constraints. But agents increasingly execute choices within those boundaries, at speed and at scale.
Strategically, this is a profound shift. It is not automation inside firms — optimizing internal workflows — but automation at market interfaces.
As more transactions are mediated by agents, the central strategic question quietly changes: not “how do we persuade customers?”, but “how do we get chosen by machines acting on their behalf?”
| Strategic dimension | What changes with agentic AI | What breaks | Strategic question for leaders |
|---|---|---|---|
| 1. Who is the customer? | Customers split into principals (humans) and agents (decision proxies) | The alignment between user, buyer, and decision-maker | Are we optimizing our value proposition for humans, for agents, or for both — and where do we accept trade-offs? |
| 2. How firms compete | Competition shifts from persuasion to compatibility | Brand- and narrative-led differentiation | Are we legible, comparable, and selectable by machines? |
| 3. Pricing & contracts | Transactions become continuous, granular, and dynamic | Opacity-based pricing, complexity-driven bundling | Where does our pricing power come from when agents can constantly renegotiate? |
| 4. Distribution & intermediation | Control moves to those who design agent decision paths | Direct access to customers | Who controls how agents search, rank, and choose in our market — and are we visible there? |
| 5. Organization & governance | Firms must become agent-ready internally and externally | Informal processes and discretionary execution | What are agents allowed to decide on our behalf, and where do humans intervene? |
Strategy implication #1: Who is the customer?
Classic strategy rests on a simple alignment. The customer is a human decision-maker, and the value proposition is a bundle of perceived benefits weighed against price. Branding, storytelling, trust, and differentiation are all tools to influence that human judgment. Agentic AI quietly breaks this alignment. When software agents search, compare, and transact on behalf of humans, the entity making the choice is no longer the one experiencing the product.
This creates a dual-customer problem. Firms increasingly serve two audiences at once: the principal (a human or an organization that defines goals and constraints) and the agent (the system that evaluates options and executes the decision). These two do not optimize for the same things. Humans respond to brand, narrative, and social trust signals. Agents respond to structured data, measurable performance, verifiable guarantees, latency, price, and reliability. What reassures a person may be irrelevant to a machine, and what satisfies a machine may be invisible to a person.
The consequence is a bifurcation of value propositions. BlackRock’s internal use of agentic systems illustrates this tension well. Portfolio managers still set strategy, risk appetite, and objectives, but execution increasingly flows through platforms like Aladdin, where machine-readable constraints and performance metrics dominate. Strategy is human; selection and execution are progressively automated.
Salesforce’s launch of Agentforce makes the shift explicit. The platform is not primarily designed to persuade human buyers, but to enable autonomous agents to transact, resolve issues, and trigger workflows across enterprise systems. In this world, “who is the customer?” is no longer a philosophical question but a design constraint. Firms that continue to define their customer exclusively as a human decision-maker risk optimizing the wrong interface. Strategy now begins one step earlier: deciding whether you are building primarily for people, for agents, or for the increasingly uncomfortable space in between.
Strategy implication #2: Competition shifts from persuasion to compatibility
When agents transact, persuasion loses ground to interoperability. Marketing, branding, and narrative do not disappear, but they cease to be the primary competitive surface. Agents do not “like” products; they select them. And selection is driven less by appeal than by fit. In agent-mediated markets, firms don’t compete for attention — they compete for compatibility.
This shifts the basis of advantage from messaging to machinability. Increasingly, what matters is API quality, machine-readable contracts, clear service-level commitments, and predictable behavior under automation. These are not traditionally considered sources of differentiation, yet they become decisive when agents compare options at scale and at speed. The logic is familiar from search engine optimization and marketplace ranking algorithms, but generalized: not just content discovery, but pricing, procurement, logistics, customer service, and enterprise workflows.
Standardization therefore becomes strategic again. Agentic transactions favor clear specifications, low ambiguity, and well-defined interfaces. This weakens differentiation based on storytelling or bespoke complexity and strengthens differentiation based on system-level fit. Firms that integrate cleanly into agent workflows gain an advantage, even if their offering appears undifferentiated to humans. This is where infrastructure quietly captures value: not by being visible, but by being indispensable.
The pattern is already visible. Fluency’s agentic optimization platform differentiates not through brand or creative positioning, but through deep technical integration across advertising networks, enabling autonomous campaign optimization. At the other end of the stack, players like OpenAI’s Codex and cloud AI providers compete primarily on performance, reliability, and API accessibility — attributes that matter far more to machines than to people. In agent-first markets, the winners are not the most persuasive firms, but the most legible ones.
Strategy implication #3: Pricing, contracting, and negotiation are redesigned
Agents do not just change who executes transactions; they change how transactions happen. When autonomous systems search, compare, negotiate, and switch on behalf of humans, markets become more continuous and more granular. Agents can compare thousands of offers in real time, renegotiate conditions dynamically, and move between suppliers with near-zero friction. Transactions stop being episodic events and start looking like ongoing optimization processes.
This undermines several long-standing sources of pricing power. Opacity-based pricing becomes fragile when agents can parse and compare fine-grained terms. Bundling based on complexity loses its protective effect when components can be unbundled and recombined automatically. Cross-subsidies hidden inside contracts become visible when agents optimize line by line. The strategic implication is stark: unless pricing power is structurally protected (through scale, switching costs, or control over standards) it tends to erode.
At the same time, contracts themselves begin to change form. Instead of legal text interpreted ex post by humans, we move toward rule-based agreements that can be executed and monitored by machines. Clauses become conditional logic. Guarantees become measurable thresholds. Performance adjustments happen in real time rather than through periodic renegotiation. Firms that can formalize their value proposition and operationalize it reliably gain an advantage over those that rely on ambiguity or discretion.
Salesforce’s Agentic Enterprise License Agreement is an early signal of this shift. Rather than pricing agents as seats or features, Salesforce experiments with usage-based and relationship-driven models that reflect how autonomous workloads actually behave. The bet is that in an agent-mediated world, value is not captured by locking customers into static contracts, but by remaining continuously relevant inside automated decision loops. Pricing, like strategy itself, becomes less about setting terms once and more about staying selected over time.
Strategy implication #4: Distribution and intermediation are reconfigured
When agents choose on behalf of users, distribution quietly changes hands. If autonomous systems decide which platforms to query, which protocols to use, and which options to surface, then power shifts upstream to those who design the decision environment itself. Agent builders, operating systems, and recommendation layers become the new gatekeepers. Whoever defines how agents search, rank, and decide increasingly acts as the distributor — even if they never touch the underlying product.
This dynamic extends a familiar pattern from digital markets. Search engines, app stores, and marketplaces captured value not by producing goods, but by controlling discovery and access. Agentic AI generalizes this logic across far more transactions. In September 2025, OpenAI enabled direct purchases from select websites inside ChatGPT and integrated services like Spotify and Figma, allowing users to act without switching applications. The interface disappears; the intermediary becomes ambient.
Incumbents are already reacting defensively. Amazon sued Perplexity, alleging that the startup’s browsing behavior violated its terms of service by acting as a shopping agent rather than a human user. Airbnb chose not to integrate with ChatGPT, arguing that the feature was not “quite ready.” These moves are less about technical readiness than about distribution risk. Once agents become a dominant access layer, opting out can preserve control — but at the cost of visibility.
This introduces a new strategic hazard: becoming agent-invisible. A firm can be perfectly competitive for humans and yet never surfaced by agents that mediate demand. The analogy is familiar — being absent from Google Search or excluded from Amazon Marketplace — but broader in scope. Salesforce’s expanding Agentforce partnerships point to one response: embedding deeply in agent ecosystems, controlling orchestration and data governance rather than fighting intermediaries outright. In agent-mediated markets, distribution is no longer a channel to optimize. It is a position to defend.
Strategy implication #5: Organizations must be redesigned to serve agents
All of the previous implications ultimately collapse into an organizational question. Serving agent customers is not just a product or go-to-market issue; it requires firms to be legible, predictable, and governable by machines. That, in turn, demands internal redesign. Firms that want to transact with agents must reorganize around data readiness and interface clarity, not just around teams, functions, or channels.
At a minimum, this means treating data and interfaces as first-class organizational assets. Product catalogs need to be structured rather than descriptive. Availability must be real-time rather than assumed. Performance needs to be measured continuously and exposed transparently. What used to live in slide decks, PDFs, or tacit knowledge increasingly needs to be formalized into schemas, APIs, and metrics. For agents, opacity is not strategic ambiguity; it is simply unusable.
Just as important is governance over autonomy. Firms must decide what agents are allowed to commit to on their behalf, which decisions require human escalation, and where discretion stops. This is not only an external question about customer-facing agents; it mirrors an internal one. The same issues arise when deploying agents inside the firm to manage pricing, procurement, or operations. In both cases, strategy becomes a matter of defining boundaries, guardrails, and exception-handling — not issuing instructions.
This is why some observers describe the emergence of an “agentic organization.” In such organizations, workflows are designed with agents as default actors, and humans intervene primarily to set objectives, resolve conflicts, and handle non-standard situations. Strategy is no longer expressed only through plans and incentives, but through the architecture of systems that act continuously on the firm’s behalf. In a world of agent customers, the firms that win will be those whose organizations are not just AI-enabled, but agent-ready.
Conclusion
Taken together, these shifts reframe what it means to win in agentic markets.
Competitive advantage no longer comes primarily from being compelling to human judgment, but from being legible to machines, selectable by algorithms, reliable under automation, and embedded in the ecosystems where agents operate.
These are not tactical concerns; they are strategic levers. They determine whether a firm is visible, comparable, and trusted inside the decision loops that increasingly govern economic activity. In agent-mediated markets, firms win less by being loved, and more by being trusted by code.
Photo de Annie Spratt sur Unsplash
