Every business runs on relationships. The systems that track them were built for a world where people were the only agents.
That world is over. Customers already carry personal AI assistants. Vendors already ship agents against data models drawn twenty years before agents existed. The gap between the interface a person uses and the surface an agent reads has become a structural problem.
Actionary is the CRM built on the other side of that gap.
The CMR thesis
Customer Managed Relationships is the frame. The centre of control moves from enterprise systems to the customer. Customer intent, permissions, and preferences become the orchestration layer. Enterprise systems become execution endpoints for customer-controlled agents.
The three-year horizon: winning a relationship becomes becoming the preferred machine-resolved option. Campaign-centric competition weakens against interoperability and execution quality. Support, sales, and service journeys become API outcomes before they become UI experiences.
The full argument, with the CRM and CMR shapes laid out side by side, lives at /about/cmr-vs-crm.
What agent-native means at the substrate
Every record type in Actionary publishes one description of itself — fields, types, relationships, validations, capabilities. The product app reads that description to render every page. The embedded agent reads the same description to reason. External agent clients fetch it through the same interface. One schema. The human view and the agent view share a single source of truth.
Every agent turn ships the tenant’s schema inside the system prompt, cached at the provider. The agent knows the schema without spending tool calls to discover it. Anthropic, OpenAI, Bedrock, Azure OpenAI, and Gemini all ship as full streaming adapters with per-step failover. The provider is portable. The loop is ours.
Workflows are first-class agent objects. A workflow becomes agent-accessible the moment it is defined, with no per-workflow integration work.
What governance-first means at the substrate
Tenant isolation is enforced by the database itself. Every tenant-scoped table carries a row-level security policy that pins every query to the tenant on the current request, enforced by a database role that cannot bypass it. A missed application-layer filter returns zero rows.
Cryptography reduces to one root. A single platform master key derives one key per subsystem — multi-factor authentication, single sign-on, provider keys, integration tokens — through a labelled key-derivation step. Rotation is one ceremony, one endpoint.
Platform staff and tenant users are separated by session type. Bearer tokens are stripped of the staff signal at the request boundary, so an external token cannot claim elevation regardless of payload. Details at /for-ctos.
Where the incumbents sit
The incumbents are racing to bolt agent APIs onto data models drawn twenty years before agents existed. They have recognised that agents matter. They have yet to recognise what comes after.
Bolting agent behaviour onto a pre-agent schema inherits the shape of the underlying system: permission drift between environments, partial coverage of what the agent can actually reach, per-record-type engineering work every time a new object type needs to talk to the agent, a second monitoring pipeline for the AI layer, and a separate audit trail for what the agent does.
Starting agent-native inverts every one of those costs. Any new record type a tenant admin defines is instantly readable, writeable, usable inside workflows, and audited — because the platform treats every record type uniformly.
The takeaway
Actionary is a substrate first, a CRM second, and a demonstration of the CMR thesis in code. The human work of curating relationships and the agent work of acting on them are the same work, on the same data, through different interfaces.
If the next decade of enterprise software is agents talking to agents on behalf of people, the surface those agents talk to has to be built for them from the first migration. It is.
If you are choosing a CRM for the next three years, we should talk.