Section B — What each company publicly demonstrates

Company mappings

Each mapping contains an evidence register drawn from primary sources, a three-column comparison of publicly demonstrated capability against the proposed architecture, public patent direction, and prioritised patent evolution opportunities for consideration before the updated provisional is finalised.

Google

Agent interoperability, multimodal models and creator ecosystem

Publicly describes Agent2Agent as enabling autonomous agents built on different frameworks to discover one another, communicate, collaborate and hand off tasks; and documents Agent Engine Memory Bank for long-term agent memory.

81.0
weighted / 100

Salesforce

Enterprise agentic commercial orchestration and unified customer profile

Publicly describes Agentforce as an environment for building and orchestrating enterprise AI agents, with Data Cloud unified profiles, autonomous marketing orchestration and open agent interoperability.

78.6
weighted / 100

Adobe

Enterprise content supply chain and creative production

Publicly describes an agentic content supply chain connecting enterprise context, brand intelligence and AI agents across planning, creation, activation, delivery, reporting and insights.

78.2
weighted / 100

Microsoft

Enterprise agent platform and knowledge grounding

Publicly documents agents with knowledge, tools, connected agents and persistent memory, together with multi-agent orchestration across enterprise data and systems.

76.5
weighted / 100

Meta

Creator and viewer behavioural intelligence at consumer scale

Publicly describes a creator assistant that understands a creator's audience, engagement trends and performance, learns the creator's goals over time and provides increasingly personalised recommendations.

69.6
weighted / 100

Amazon Web Services

Agent infrastructure and platform primitives (not a creator ecosystem)

Publicly documents AgentCore Memory maintaining short- and long-term memory so agents retain historical interactions, preferences and knowledge across sessions, and documents multi-agent collaboration under a supervising agent.

66.1
weighted / 100

NVIDIA

AI infrastructure, agent tooling and simulation-grade Digital Twins

Publicly documents agent toolkits, microservices and Omniverse Digital Twins; the twin work is principally physical and industrial and should not be treated as equivalent to behavioural creator/viewer twins.

51.8
weighted / 100