The AI Agent War, Part II – Enterprise Implications for 2025

TL;DR for Executives

AI agents are no longer just developer tools. They are becoming sovereignty assets, industry disruptors, and governance flashpoints.

As introduced in Part I, CB Insights applies two benchmarks to separate hype from durability: the Mosaic Score (0–1000) and Commercial Maturity (1–5). These ratings allow executives to benchmark which players are durable, which are scaling, and which remain speculative.

Geopolitics: AI as State Power

Governments now see AI agents as sovereignty infrastructure — on par with semiconductors and cloud. This shifts vendor selection from a procurement decision to a geopolitical calculation.

• Anthropic ranks 955/1000 on Mosaic (99th percentile, Headcount +16 in the past year), with revenue forecast at $35B by 2027. Its federal defense contracts, including work with the National Nuclear Security Administration (NNSA), embed it directly in the US defense-industrial base. Durable, but tightly tied to US policy.
• Code Metal (Headcount +38 in the past year) has been granted “Awardable” status in the DoD’s CDAO Tradewinds Marketplace, validating its AI-optimized low-level code for defense, IoT, and medical devices. This approval signals that AI is entering mission-critical systems faster than most enterprises expected.
• Mistral AI (Mosaic 909/1000, Headcount +11 in the past year) is Europe’s flagship sovereignty play. Poolside (Mosaic 803/1000, Commercial Maturity 3/5, Headcount +50 in the past year) with 120x year-over-year revenue growth ($250K → $30M) stands as another EU-backed counterweight. Both highlight Europe’s push for AI independence.

Implication: In the next 12–24 months, procurement rules, trade negotiations, and M&A reviews will increasingly favor domestic or “trusted” providers. Enterprises that depend too heavily on one geography’s vendors may face regulatory or political shocks.

Beyond Coding: Cross-Sector Transformations

Adoption is spreading fastest in industries where compliance, payments, and customer engagement drive both costs and risks. Most players remain early-stage, but disruption potential is significant.

• Skyfire (Commercial Maturity 3/5, Headcount +70 in the past year) is expanding beyond pilots but not yet entrenched. Its agent-driven payment rails point to a future where autonomous agents transact directly, bypassing banks and card networks.
• Runner H (H Company) (Mosaic 866/1000, maturity not yet rated, Headcount +164 in the past year) reflects strong execution. Its “self-healing” automation adapts to UI changes without manual intervention, lowering maintenance costs. Runner H is credible as a workflow partner or acquisition candidate, though not yet mature enough to anchor enterprise-wide platforms.
• Bland AI (Commercial Maturity 3/5, $3.8M 2024 revenue, Headcount –84 in the past year) automates enterprise phone calls with hyper-realistic AI voice agents. Rapid growth but not yet enterprise-proven.
• Kubiya (Headcount –33 in the past year) is seeing early adoption in Fortune 500 IT operations, automating incident response and routine DevOps tasks.
• Strata Identity (Mosaic 655/1000, Headcount –90 in the past year) shows traction in regulated industries. Its Maverics platform treats AI agents as “first-class identities,” embedding compliance at runtime.
• BASE44 Maturity not yet rated, headcount not disclosed) is still experimental, while Lovable (Commercial Maturity 3/5, Headcount +150 in the past year) signals the democratization of app creation from natural language.

Implication: Finance, healthcare, logistics, and other compliance-heavy industries will be the first to experience disruption. Leaders should pilot agents in high-cost, high-risk workflows, while preparing for consolidation among early-stage vendors.

Risks: Why Governance Cannot Lag Adoption

AI agents are scaling faster than governance frameworks can keep up. This mismatch is the biggest danger for enterprises: early wins in productivity can quickly turn into systemic vulnerabilities.

• Operational fragility: Most startups rank only 2–3/5 in Commercial Maturity — expanding but unproven. Enterprises risk getting stuck in endless pilots or building workflows on brittle vendors.
• Compliance landmines: Privacy, auditability, and identity orchestration remain afterthoughts in many agent platforms. Retrofits will be expensive — and regulators are moving faster than expected.
• Strategic overexposure: Heavy dependence on US incumbents concentrates supply-chain and policy risk. By contrast, Baseten (898/1000 Mosaic, 98th percentile) and Mistral (909/1000) show that alternatives exist — but few enterprises are hedging properly.
• Talent gaps: Roles are shifting from coders to orchestrators, evaluators, and compliance integrators. Without re-skilling, enterprises will lack the capability to govern agent ecosystems at scale.

Executive signal: Boards should treat AI governance maturity like a credit rating — a measurable indicator of enterprise resilience. By 2026, investors and regulators will expect companies to demonstrate not just AI adoption, but AI risk governance benchmarks as part of disclosures.

Final Word for Executives

The AI agent war will not be won by those who pilot the most tools. It will be won by those who institutionalize defensibility, resilience, and sovereignty.

Executives must move beyond experiments and adopt a portfolio strategy:

1. Diversify: Hedge geopolitical and vendor risks by balancing US incumbents with EU sovereignty players and infrastructure hedges like Baseten.
2. Embed governance: Bake privacy, identity, and audit controls into every layer from day one — waiting will multiply costs and risks.
3. Buy before the wave: Target acquisitions with high Mosaic scores but low maturity — companies with momentum but not yet entrenched, ripe for consolidation.
4. Re-skill now: Build teams of orchestrators, evaluators, and compliance leads. Talent is the bottleneck, not the models.

2025 marks the inflection: agents are moving from pilot experiments to operating infrastructure. Enterprises that act decisively — balancing innovation with governance — will not just deploy agents, but control them as levers of competitive advantage for the next decade.

Sources

Insights are based on CB Insights’ AI Agent Tech Stack Report with supporting data from the Coding AI Agent & Copilot Report.
References include sovereignty plays (pp. 5–7), sector adoption signals (pp. 10–14), and governance risks (pp. 15–17).

Contact [email protected]  if you have trouble accessing.