Onboarding Digital Labor
Capable agents are arriving faster than the structure to run them. This series is about that structure: why onboarding, not capability, is the bottleneck; the framework for doing it (Ground, Govern, Graduate); and what it looks like in practice and at scale.
Lattice is the worker layer of a larger governance stack. How it fits with Concord: The worker and the action.
The problem
Capability is no longer the bottleneck. Context, control, and accountability are. Why the obvious tools fall short.
Part 2The framework
Ground, Govern, Graduate: give an agent meaning, bound what it can do, and let it earn trust over time.
Part 3In practice
The framework on one ticket, where to start tomorrow, and what changes as agents become a fleet.
Prefer one sitting? The single-page version covers the whole argument.
What does it mean to onboard an AI agent?
Onboarding digital labor means giving an AI agent what a company gives a new employee: a role, the context to do the job, limits on what it can touch, someone accountable for the outcome, and a path to more trust. A prompt configures a model. It does not onboard a worker.
Common questions
Why isn't a prompt enough to deploy an agent?
A prompt sets behavior. It does not grant a role, scope permissions, name an owner, or define what good looks like — the standing structure every other kind of worker gets before touching real systems. Without that structure, capable agents stall at the pilot stage.
What does a properly onboarded agent inherit?
Three layers — Ground, Govern, Graduate: grounding gives the agent meaning (role, context, ontology); governance bounds it (permissions, policy, a control plane); graduation lets it earn autonomy through evaluation and trust levels.
How does an agent earn more autonomy?
By climbing a trust ladder: observe, recommend, draft, act with approval, act within guardrails, autonomous in bounds. Promotion changes how often a human stands in the path, not the accountability — the audit trail never thins.
How does Lattice relate to Concord?
Lattice governs the worker; Concord governs the action. Lattice sets the standing structure — role, permissions, trust level — and Concord enforces it per action, feeding the evidence back into the agent's scorecard. Policy flows down; proof flows up.