The narrative has a clean ring to it. A two-person agency walks into a pitch against a fifty-person firm and walks out with the contract. Not because they were faster. Not because they were cheaper. Because they were more legible. The client could see exactly what they were buying, how it would be delivered, and who would be doing the work. The giant, by contrast, was a black box with a logo.
It is a satisfying story. It is also, in the publicly available record of 2022 through 2025, a story with almost no documented evidence. The three cases that supposedly anchor the pattern do not appear in the data that exists. That absence is itself worth examining. If the phenomenon is real but undocumented, what would it take to prove it? And if it is not real, why does it feel so true?
The gap between intuition and evidence
Start with what we actually know. The annual eDiscovery Business Confidence Survey, published by ComplexDiscovery, tracks what professionals in that industry worry about. In the second half of 2025, the top concern was “Increasing Volumes of Data,” cited by 23.44% of respondents, according to the survey’s authors. “Budgetary Constraints” and “Increasing Types of Data” were virtually tied at 21.88% each. “Lack of Personnel,” which had been the perennial number-one challenge in post-pandemic surveys, fell to fourth place at just 12.50%.
That shift is interesting. If personnel is no longer the bottleneck, something changed. The survey’s authors suggest that the integration of Generative AI may be acting as a force multiplier, alleviating pressure on human teams. That is a plausible inference. But it is not a case study of a small agency winning a big deal. It is a macro trend that creates favorable conditions.
The problem is that no publicly documented cases of a two-to-three-person agency displacing a fifty-plus-person competitor on a specific engagement appear in the sources available. Not one. The hypothesis is compelling. The evidence is absent. That does not mean the phenomenon does not exist. It means the conversation about it has been running ahead of the facts.
What legibility actually means
The concept of legibility in services procurement is straightforward. A client wants to know what they are paying for, who will deliver it, and how success will be measured. Large agencies often obscure these answers behind layers of account management, scoping documents, and departmental handoffs. A smaller agency can lay out the same information in a single conversation.
The eDiscovery survey data hints at why this might matter more now than it did five years ago. With data volumes and budget constraints tied as the top concerns, clients are under pressure to justify every dollar spent. A legible proposal, one where the client can trace inputs to outputs, is easier to defend internally. A proposal that requires faith in a large team’s internal processes is harder to sell to a CFO.
The drop in personnel concern reinforces this. If clients were primarily worried about having enough people to do the work, they would favor large agencies with deep benches. But as that concern recedes, possibly because AI tools are absorbing some of the workload, the advantage of headcount diminishes. The advantage of clarity rises.
The giant was a black box with a logo. The small agency could show the client exactly what they were buying.
The AI multiplier, documented elsewhere
The most concrete evidence of AI as a force multiplier comes from large enterprises, not small agencies. In a compilation of real-world generative AI use cases published by Google Cloud, the authors detail specific deployments. LUXGEN, a Taiwanese electric vehicle brand, uses Vertex AI to power an AI agent on its LINE account, reducing the workload of human customer service agents by 30%. Valeo, the automotive supplier, is deploying Gemini for Workspace to its entire 100,000-person workforce, and more than 35% of Valeo’s code is now generated by AI.
These are not small agencies. They are multinational corporations with thousands of employees. The AI multiplier is real, but its documented impact is in contexts where scale already exists. The question for a two-person agency is whether the same tools can produce the same leverage without the supporting infrastructure of a large organization.
The answer is probably yes, but the data does not prove it. A two-person agency using Gemini to generate code or Vertex AI to automate customer responses could theoretically match the output of a much larger team. But no case study in the available sources confirms that this has happened in a competitive displacement scenario. The inference is reasonable. The proof is missing.
What the data does and does not tell us
The eDiscovery survey gives us a snapshot of what professionals in one industry worry about. Data volumes and budget constraints are the top concerns. Personnel is no longer the primary issue. Those conditions, in theory, favor smaller, more efficient teams that can handle large data loads without large headcounts.
But the survey does not tell us whether those smaller teams are actually winning business. It does not name a single agency, large or small, that gained or lost a client. It is a survey of concerns, not a log of outcomes. The gap between a favorable environment and a documented win is where the speculation lives.
The same is true for the Google Cloud use cases. They show that AI can reduce workloads and generate code at scale. They do not show a two-person agency using those same tools to underbid or outperform a fifty-person competitor. The tools exist. The conditions are right. The case studies are absent.
This is not a criticism of the sources. It is a description of the evidence available. The hypothesis that small agencies can displace large competitors through legibility and AI leverage is plausible. It is not proven.
A speculative framework for the agency that could win
If we take the available data seriously, we can sketch the profile of an agency that would be well-positioned to compete against larger firms. It would be an agency that addresses the top client concerns: data volume and budget. It would use AI tools, like those documented by Google Cloud, to handle data-intensive work without adding headcount. It would present a legible proposal, one where the client can see exactly how the work gets done and by whom.
The agency would not compete on speed or price. It would compete on transparency. It would say: here is the tool we use, here is the output it produces, here is the person reviewing that output, and here is what it costs. The fifty-person firm would say: we have a team of experts, a proven methodology, and a track record of success. The small agency’s pitch is easier to evaluate. The large firm’s pitch requires trust.
This is the framework that emerges from the data. It is speculative because no documented case confirms it. But it is grounded in real trends: declining concern about personnel, rising concern about data volumes and budgets, and documented AI tools that reduce the need for human labor.
The closing observation is this. The story of the two-person agency that beats the fifty-person competitor is not yet a documented fact. It is a hypothesis that the available evidence supports but does not prove. That makes it a useful test case for how we talk about competitive advantage in services. The most compelling narratives are often the ones with the thinnest evidence. The work of proving them is still ahead.