Plain-language definitions for the words r.Team uses on this site — the Platform, the agent loop, and the fiduciary work they carry. This is the same map as the Platform page.
r.Team is a Fiduciary Reasoning Platform with agents at the core. Fiduciary duty, ethics, compliance, cybersecurity, and advisory regulation are built into how the agents reason — not wrapped around them. The map is one list, used the same way on every page: Reasoning Engine, Agent Studio, the Bridge, Trust Intelligence, the Record, and a conversational layer. See the Platform.
The judgment work of trust administration: reading what a governing instrument actually says, applying distribution standards to real circumstances, and producing a documented rationale an officer, counsel, or examiner can review. Distinct from trust accounting, which defends itself through reconciliation. See What is fiduciary reasoning?.
The shared context every agent reasons from: the governing instrument, the firm's written policies, the applicable regulation, and the firm's prior decisions. Agents do not carry a private view of the rules. The same inputs, policy version, and evidence yield the same proposal, and the record shows which version was in force. See the Reasoning Engine on the Platform.
Known in engineering contexts as inference; on this site we name the capability by what operators experience.
Where governed agents are composed. An agent is a sequence of actions drawn from the Action Catalog. It inherits every constraint those actions carry: what it may read, what it may propose, what it must cite, and which review tier its output requires. See Agent Studio.
The firm's inventory of permitted agent actions. Each action declares its data scope, what it may propose, what it must cite, and where a human must stop it. Agents are composed from the catalog; they do not invent unconstrained steps.
Four gates that decide who must approve an agent proposal before anything writes. The agent cannot skip, batch past, or time out a gate.
| Tier | Example terms | Who approves |
|---|---|---|
| Clerical | Trust name, execution date, taxpayer identification, situs | One reviewer, sampled for quality |
| Structural | Trustee chain, beneficiary classes, revocability, fee basis | Trust officer, single approval, full audit record |
| Interpretive | Distribution standard, discretionary language, powers of appointment, investment direction, retention clauses | Trust officer plus fiduciary counsel — two people, never one |
| Judgment | Any term that will drive a distribution decision, a tax election, or a concentration position | Committee action with written rationale on the record |
Clerical work can be sampled. Structural changes go to a trust officer. Interpretive terms require officer plus counsel — two people, never one. Judgment that will drive a distribution, a tax election, or a concentration position requires committee action with a written rationale on the record.
One auditable doorway to the systems you already run — trust accounting, custody, CRM, document management. Agents read through the Bridge, propose through the Bridge, and, only after human approval, write through the Bridge. The platform overlays those systems. It does not replace them. See the Bridge on the Platform.
Agents that continuously review reports and operational data through the Bridge — statements, reconciliations, fee runs, outages, exceptions — and route every proposed action to an operator for approval. Discovery proposes. The operator decides. Nothing executes until a human decides. See Discovery through the Bridge.
Document intake and cited abstracts of governing terms. Abstracts cite. They do not conclude. A named officer, counsel, or committee reviews every field before anything reaches the system of record. See Trust Document Intelligence.
Firm-defined, versioned policies that decide what routes where — which findings are clerical, which are structural, which are interpretive, and which require judgment-level approval. On the platform this lives in Trust Intelligence and the Reasoning Engine, not as a separate product. See Trust Intelligence.
A per-account hash chain of fiduciary events. Only a fingerprint is published; no client file leaves the firm boundary. Years later, what was decided and why is arithmetic, not recollection. See the Record on the Platform.
The conversational layer on the Platform. Administrators ask questions of accounts and activity in plain language. Questions stay inside the same gates as every other agent — scoped, cited, and stopped for a human where policy requires it. See the Platform map.
Intelligence that works alongside your team at eye level: proposing and explaining, never overriding, with the human decision recorded at every step. The operator gains reach. The Engine gains the operator's judgment, captured in the Record. See how the home page states it.
Two applications of the same rule: run the new thing beside the old until evidence authorizes the change.
On the platform. A prove-before-write gate. Agents run against your live system of record and earn write access only after they match it. The platform overlays your core systems; it does not replace them. See Deploy beside what you have.
As a consulting engagement. The Parallel-Run Independence Playbook is an optional conversion for firms that have decided to leave a per-account vendor — one signed block at a time, until the vendor relationship is retired. See the playbook.
The arithmetic and control side of the trust department: reconciliation, fee calculation, exception surveillance, tax-lot tie-outs, custodial interfaces, and the controls that keep statements and books examiner-ready. Distinct from — and complementary to — trust administration's document and judgment work.
A governed responsibility in the trust relationship — trustee, co-trustee, investment manager, custodian, protector, or similar — with authorities, duties, and approval thresholds that operations and agents must respect.
The contracted rules for how fiduciary and administrative fees are calculated and applied to an account or relationship — tiers, minimums, waivers, and exceptions. Agent Studio agents can monitor calculated fees against the governing schedule.
A person or entity entitled to present or future benefits under a trust or related structure. Current, contingent, and remainder interests often differ — and those distinctions matter for distributions, reporting, and Reasoning Engine proposals an officer can review.
The institution that holds assets and produces position, transaction, and cash files consumed by trust accounting and operations. The Bridge connects to custodian feeds rather than replacing them.
Name, Intelligence, Likeness — keep, download, and publish. Likeness is the portable, inspectable shape of a practitioner's judgment: memory, actions, and skills, written to a markdown document the practitioner or Trust Professional Group holds title to. They can download that document and publish it under the individual's name or the group's name. Inside a tenant, an agent applies that likeness to a specific customer. Across tenants, a published likeness may be bought or rented. Proprietary data is not published and is not taken. The host may run an agent against that likeness while the owner uses the host. That is use of a tool, not a claim on the asset.
You are not being replaced. You are being published. Published means you keep the markdown document and can take it with you. Proprietary data stays put. See the NIL Intelligence Likeness concept.
Combining language understanding with explicit, verifiable rules. A trust instrument is both a document to be read and a set of obligations to be enforced. Language understanding alone guesses. Rules alone cannot read the instrument. Together, governed inside the agent loop, they can prepare work an officer can check.
The engineering term for running a trained model over data to produce a result. On this site we call this capability the Reasoning Engine, because that describes what it does for your operators rather than how it works under the hood.
