The 19th century banking problem that AI hasn’t solved yet

Key point: There is no existing trust architecture for agent-to-agent (A2A) AI; Salesforce researchers argue this gap must be closed now. They propose four foundational elements—registered identity and reputation, boundaries (standards) instead of scripts, structured accountability, and calibrated escalation—to enable safe, auditable autonomous negotiations. Immediate implications affect healthcare billing, financial services (treasury/FX/credit negotiations), and supply chain coordination; firms that help define standards and build reputation infrastructure may gain a strategic advantage.

Analysis

The commercial value of a credible A2A trust layer will accrue to firms that own both the customer relationship and the metadata/identity plumbing — not merely the best base models. Salesforce-style vendors that can productize Agent Cards, reputation scoring, and audit-grade escalation will convert negotiation automation from a cost center into a SaaS monetization lever; early adopters in finance and healthcare could cut human negotiation hours and exception handling by a material margin (we estimate 20–40% in pilot programs) within 12–24 months, concentrating value in incumbents with large installed bases.

The primary barrier is institutional, not technical: standardization, cross-jurisdictional legal clarity, and audit frameworks. Expect a two-speed adoption curve — pilots and vendor coalitions over the next 12 months, then regulatory and standards consolidation over 2–5 years. Tail risks that could reverse adoption include a high-impact agent failure or collusive outcome that triggers a regulatory pause, and the “wriggling” variance problem that makes reproducible audits difficult; either can produce multi-quarter adoption stalls and accelerated demand for vendor liability and insurance products.

Second-order effects favor identity/reputation providers, enterprise security vendors, and specialist consultancies that bridge legal, ops, and ML (they become gatekeepers to trust). For supply chains, dependable agent negotiation will shift bargain power: routine replenishment becomes automated (reducing working-capital friction) while strategic supplier relationships grow more valuable and concentrated. The tactical advantage will go to organizations that help write standards — control of the specification will be a durable moat once reputation metrics matter at scale.

Key Decisions for Investors

• Overweight CRM (Salesforce) — add a tactical 3–5% position within 2–6 weeks or on a 5–10% pullback. Thesis: leader in enterprise trust architecture monetization; target +30% upside over 12–24 months if Agent Cards/reputation products begin pilot monetization. Risk: adoption delays or regulatory setbacks; use a 15% stop-loss or hedge with puts.
• Pair trade: long CRM / short GOOG (class C) equal notional, 12-month horizon. Rationale: CRM captures enterprise governance value; GOOG has scale but weaker enterprise positioning for managed trust stacks. Target relative outperformance of 15–25%; risk that Google accelerates GCP/AI integration or partners with incumbents — keep position size modest (2–3% net exposure).
• Event-driven GOOGL exposure via small bullish call spread (9–12 month tenor) sized 1–2% of portfolio, keyed to standards announcements or large banking/health pilots. Low-cost upside if Google’s A2A specification adoption increases platform demand; capped loss if standards move slowly.
• Buy hedges if exposure concentrated in enterprise AI: purchase 12-month protective puts on CRM (or a put spread) sized to cover 30–50% of position notional to guard against a regulatory or reputational shock. This preserves upside while limiting tail loss from a systemic agent failure.