Home > Blog > The Real Threat to SaaS: How AI Agents Are Shifting Pricing Power

The Real Threat to SaaS: How AI Agents Are Shifting Pricing Power

Mar 03, 2026UP!

• Blog
• AI agents
• enterprise software
• EU AI Act
• MCP
• pricing power
• SaaS

“SaaS is dead — AI agents will replace everything.” This claim is spreading fast. But the question itself is wrong. This article examines where pricing power actually moves in the age of AI, and why the answer matters for legal and enterprise risk strategy.

1. Reframing the Question

The right question is this:

SaaS will not die. But the source of pricing power is quietly shifting layers.

Markets are compressing SaaS multiples — not simply because of substitution risk. The real problem is that it has become unclear which layer retains economic rent (excess returns). This opacity is inseparable from legal frameworks, regulation, and national technology strategy.
The core issue is not “will SaaS disappear” but “who controls explanation costs and accountability” — and where network exclusion effects take hold.

2. What SaaS Was Actually Selling

SaaS was never really about the software itself. What enterprises were buying was a “package of trust.” That package had three components:

① Data lock-in
② Workflow dependency
③ Externalized accountability — SOC2 compliance, audit logs, regulatory update management, SLA guarantees

But beneath these sat a less visible competitive advantage: low explanation cost (network effect).
“We use Salesforce” — that single sentence clears internal approvals, satisfies counterparty due diligence, and passes an audit. This “no explanation required” state was the true moat, beyond any feature or price point. SaaS was not just a tool. It was an entity that made explanation unnecessary.

3. What AI Agents Disrupt: The Disappearance of Explanation Cost

In a world where agents execute across APIs autonomously, the structure of humans operating UIs collapses. The consequences:

UI differentiation loses meaning
SaaS brand identity becomes abstracted
Switching costs fall

In short: “which SaaS platform you use” as a category of explanation disappears entirely.
This resembles the Microsoft Word vs. Ichitaro dynamic in Japan. Technical superiority did not determine the winner — “please send it as a Word file” fixed the standard. In the agent era, the “no explanation required” layer simply moves up the stack.
This transition means rising price elasticity: pricing power weakens, net revenue retention (NRR) faces pressure, and multiple compression is the market’s advance pricing of this structural shift.

4. What Remains: Accountability Exists in Two Layers

Enterprises are not buying functionality. They are buying risk externalization infrastructure. The key is to separate accountability into two layers:

In high-regulation verticals — financial services, healthcare, defense — the institutional design capability to interpret regulatory requirements and translate them into agent-inclusive control architectures is not easily substituted.

The minimum accountability floor is a table-stakes entry condition. The sustained capacity to update control design is the long-term moat.

5. Three Conditions for Pricing Power

Explanation cost cannot be fixed by brand alone. Three conditions must be satisfied simultaneously.

① Identity Network

The foundation for uniquely attributing actions to actors. Large-scale ID infrastructure — Microsoft Entra, Google Identity — is powerful, but scale alone is not the point. In zero-trust architectures, the critical question is: “can consistent policy be applied across any cloud, any SaaS, any agent?”
Two diverging models emerge:

Integrated ID (Microsoft, Google)
Federated ID (Okta and similar cross-platform approaches)

Whether “agent identity” is treated as an independent market changes the entire structure. Viewing agents as mere tools versus treating them as quasi-principals (Agent-as-Principal) produces opposite outcomes for ID dominance.

② Unified Audit Log (Most Critical)

A unified log is not simply observability infrastructure. It is a “accountability adjudication device” and a “risk pricing device.” The EU AI Act (Regulation (EU) 2024/1689), SOC2, and FedRAMP are all being designed with log infrastructure as a foundational assumption. When incidents occur, the extent to which causality can be traced backward determines fault allocation, liability exposure, and insurance premiums.
Whoever controls the primary source of the unified log effectively controls:

Regulatory compliance cost
Accountability boundaries
Financial valuation of risk

This is where pricing power is directly anchored.

③ Protocol Standards

Open standards like MCP (Model Context Protocol) reduce connection costs. But this alone cannot establish dominance. Standardized connectivity is a necessary condition, not a sufficient one.

6. The Governing Formula

The number of players capable of satisfying all three conditions simultaneously is extremely small — limited to a handful of integrated enterprises.

7. Timeline for Standard Formation

Structural lock-in takes time. The current state sits at Phase 01–02:

8. The Division of Labor: Open Standards vs. Integrated Enterprises

The most realistic outcome is a structured division of labor:

Connectivity: Open standards
ID + Log: Integrated enterprises
Execution: Multiple competing agents

Pure open-standard-only dominance is unlikely. In enterprise and public-sector contexts, integrated companies capable of explaining “whose identity acted, and what they did” in a single sentence hold structural advantage.

9. U.S. Strategy: Claiming the Standard First

Looking at recent moves by the U.S. government and major technology companies, a coherent pattern emerges: define the AI standards and explanation frameworks first, then use that definitional position to lead the market.

NIST and Federal Procurement: Standard Pre-Emption + Procurement Leverage

NIST’s AI Risk Management Framework (AI RMF) is shaping federal AI procurement requirements. The result is a structure where de facto standard formation is being driven by the public sector from the top down.

Anthropic’s MCP: Open Protocol as Ecosystem Strategy

Anthropic’s Model Context Protocol (MCP) is designed not to lock users into a specific platform, but to form an ecosystem. It applies the principle that “an open protocol becomes the foundation of the ecosystem” to the world of AI agents.

Microsoft’s Three-Part Approach

Microsoft combines its identity infrastructure (Microsoft Entra), cloud platform, and compliance framework to propose “ID network × open protocol × certified cloud” as the architecture for minimizing explanation cost.

Competition Policy and Standard Dominance: Tensions in U.S. Policy

FTC antitrust scrutiny of concentration and the national security imperative of technological dominance exist in open tension. Under the current administration, export of a U.S. AI sovereignty strategy with military dimensions is expected to intensify.

10. Strategic Implications for Japan

The U.S. is combining standard pre-emption with procurement leverage to claim the definitional authority over “zero explanation cost.” If Japan remains entirely on the receiving end, log specifications and audit formats will become fixed to U.S. standards.

This is where competitive positioning in the agent era diverges. Where time and capital are invested now determines the outcome.

Conclusion

SaaS is not ending. What is ending is the old model of pricing power.
Open standards create “a world where any agent can run anything.” But the ability to answer “whose accountability governed that action” in a single sentence will belong to integrated enterprises holding massive ID networks and unified audit logs.

The next three years are a contest over who defines “zero explanation cost for agent governance.” The structure has not yet locked in. This is the fork in the road.

At Akasaka International Law & Accounting Office, we view the shift in AI agent governance not as a distant technology question, but as an immediate compliance architecture challenge for enterprises operating in regulated industries. In particular, how your organization allocates accountability between AI agents, platform providers, and internal control systems varies significantly depending on your industry, jurisdictional exposure, and existing vendor contracts — which is why we recommend an early review before your control architecture becomes fixed by default.

When companies assume that existing SaaS compliance contracts automatically cover AI agent actions, they often discover the gap only after an incident — at which point audit logs are incomplete, accountability is contested between vendors, and remediation costs dwarf what a proactive review would have required.

Early-stage consultation is significantly less costly than post-violation remediation. Most AI governance compliance questions — vendor contract gaps, log architecture requirements, EU AI Act applicability — can be scoped in a single session.

If any of the following applies to your organization, we recommend a compliance architecture review:

□ You have deployed or are planning to deploy AI agents in workflows that touch regulated data (financial, medical, HR) without a defined accountability framework □ Your vendor contracts do not specify which party holds liability when an AI agent causes a compliance breach □ You are unsure whether your current audit log infrastructure satisfies EU AI Act or SOC2 requirements for agent-driven actions □ Your legal or compliance team has not reviewed your SaaS agreements since AI agent capabilities were added by vendors □ You cannot currently answer: “If an AI agent in our stack made an unauthorized decision, whose log would prove it and who would be liable?”

If one or more apply, please contact us for an initial consultation.

You are welcome to contact us via the Contact Form to discuss and for more information.