Design explainer — not a claim that Forward Deployed Agents ships as a named feature on r.team.
Recommends · does not decide
rAgent Team
Forward Deployed Agents × Agent Studio
A front-door peer for advisors and trust teams: it recommends governed workflows from Agent Studio — with citations — then stops for human approval. No auto-purchase. No silent writes.
Pairing a named person with an rAgent Team agent helps
fiduciary duty
work stay clearer, more consistent, and easier to reconstruct later.
For RIAs, dual-registrants, and trust/ops buyers inside a Fiduciary Reasoning Platform — r.Team / Got-Data.
Forward Deployed Agents sit inside the firm’s real environment — documents, ledgers, approvals,
journals,
and
activity logs
— not in a side chatbot.
Think of an on-site specialist. They read, compare, notice patterns, and keep a clean record. They do not take the keys or take custody.
Someone still owes fiduciary duty. People decide. The agent helps with clarity, consistency, and the paper trail.
What you can do better with an agent beside you
You stay in the chair. The agent reads beside you, flags what it finds, and keeps the notes.
Read long governing documents without skipping clauses.
Compare thousands of journal lines to a fee schedule or a distribution limit.
Remember every exception.
Keep a time-stamped record of what was seen and asked.
Show the standard that was applied so another officer, an auditor, or a court can follow.
People stay better at judgment, relationships, and hard calls. This is a partnership, not a replacement.
The gap
The wealth stack is built to transact. Governing documents and fiduciary constraints are not. A recommendation that never reads the instrument is not one another officer can follow.
Start as an observer — observatory examination
The safe first assignment is an observatory examination. The agent starts by watching the platform as it really runs.
Permission card
What the observer may and may not do:
You may look.
You may take notes.
You may compare what you see with the governing documents and the rules.
You may raise your hand if something looks off.
You may not move money, change an account, or give a client advice on your own.
What it watches
The trust or advisory agreement — who the beneficiaries are, and what the
trustee
or
adviser / RIA
may do.
Principal vs income
accounting.
Investment selection and review.
Fees.
Exceptions, overrides, and approvals.
Whether the paper trail holds if someone asks, “Why was this done?”
Journals and activity logs are the true story of the platform. The agent should read accounting journals, system activity logs, workflow logs, and the audit trail on trades, cash, document edits, and beneficiary changes.
Patterns that improve work
The same exception shows up every month.
A four-step review is skipped after step two.
Corrections cluster in one place.
Work piles up at month-end.
Two systems disagree, so someone posts plug entries.
Those findings can mean fewer steps, clearer handoffs, stronger controls, and less rework. The agent proposes. People decide.
Early warning on fraud or cyber abuse
Odd hours or locations.
Over-broad viewing.
Rapid exports.
Privileged changes without approval.
Failed logins, then a success.
Journals that reverse or disguise earlier entries.
Crime is a legal conclusion. The agent flags unusual patterns, preserves the trail, and alerts security, compliance, or operations. A person or counsel decides what the pattern is.
Recommends, does not decide
Does
Read the governing instrument and the facts in front of it
Propose a governed workflow from the Agent Studio catalog
Cite the standard and the source
Stop for a named human
Does not
Auto-purchase or silently write
Replace the trustee, adviser, or officer
Decide a discretionary act
Call itself investment advice
Place trades
Approve distributions
Change beneficiaries
Send advice to a client
Quietly fix the ledger
Lock or wipe a system, or accuse a person
Looking, in scope, is the work. Acting as police, portfolio manager, or trustee is not.
How the work usually unfolds
An engagement usually follows this arc. The agent does not close the case.
Embed — controlled access to the firm’s environment.
Learn the local language — names, codes, and how this desk actually works.
Watch real cases — live files, not a demo script.
Notice patterns — repeats, skips, clashes, and gaps in the trail.
Hand findings to humans — propose, cite, and stop. A person owns the next step.
On a single request, the product steps stay the same:
Ask — the front-door peer takes the question and the file.
Recommend + cite — an Agent Studio action, with the instrument and standard attached.
Human approve — review tiers from Clerical to Judgment. Nothing moves without a person.
Compose & log — write the rationale; log to The Record (hash-chain). Bridge writes only after approval.
The human owns the duty. The agent owns observation and the paper trail.
Illustrative recommendations
Examples of the kind of workflow a front-door peer might propose. Not a catalog, score, or performance claim.
Fee Schedule Reconciliation
Match billed fees to the governing schedule. Cite the clause. Stop for the officer before any adjustment.
IPS drift
Flag holdings against IPS bands. Cite the band. Do not rebalance.
Governing-doc verification
Read the governing instrument for a proposed distribution. Cite the power. Wait for approval.
Oversight Advisor Assets
Compare adviser-directed assets to the oversight mandate. Cite the mandate. Leave the trade to the human.
Trust controls
Human-in-the-loop
Bridge writes only after approval
Hash-chain Record
Not investment advice
For RIAs and trust/ops
Advisers Act duties, IPS language, dual-hatted roles, and trust-ops constraints are the input — not a slide. The peer has to show the standard it applied.
Advisers Act
Duty language is the standard to cite, not a slogan on the wall.
IPS
The policy statement is a constraint. Drift is a finding, not a trade.
Dual-hatted
Adviser and broker hats stay separate. The recommendation names which hat is on.
Trust/ops
The instrument governs. Ops can propose; the officer approves.
Agent Studio
An Action Catalog of governed workflows. Each action carries a review tier from Clerical to Judgment. rAgent Team recommends from that catalog using the Reasoning Engine; a person chooses.
Clerical → Judgment
Test fiduciary knowledge
The lab behind the menu is the apprenticeship: Study (Ask), take the test (have rAgent take it), take it beside them. Scores stand side by side.
An AI trained by trust professionals. Certified for trust administration. Deployed on r.team.
rAgent is apprenticed to named practitioners: it learns from what they know, then takes the same test they do — so you can see, item by item, where it reasons like they do and where it doesn't.
FAQ
Does rAgent Team decide?
No. It recommends and cites, then stops. A named human approves.
Is this investment advice?
No. Not investment advice, and not a trustee or adviser relationship.
Does Forward Deployed Agents ship on r.team today?
This page is a design explainer. It does not claim FDA ships as a named feature on r.team.
Where do I take the exam?
Use Test your Fiduciary Knowledge against rAgent Team. Ask, Practice, and Have rAgent take it live there — email required for Ask and practice.
Will the agent replace my trust officer or adviser?
No. It observes, compares, and proposes. A named person decides and signs.
Can it tell me if fraud is happening?
It flags unusual patterns and preserves the trail. Whether that is a mistake, a weak control, abuse, or a cyber incident is a human and legal determination.
Glossary
Short meanings used on this page. Hover or focus a dotted term above, or read the list here.
fiduciary duty
A legal duty to put the beneficiary or client first and to follow the standard that applies.
governing instrument
The trust, will, or agreement that states who gets what and what the trustee or adviser may do.
trustee
The person or firm that holds and administers trust property for beneficiaries.
adviser / RIA
An investment adviser, including a registered investment adviser, that gives advice about securities for compensation.
Advisers Act
The U.S. Investment Advisers Act of 1940, which sets duties for many advisers.
IPS
Investment Policy Statement: the written bands and rules for how money may be invested.
principal vs income
Trust accounting that splits corpus (principal) from earnings (income), because they often go to different people.
discretionary act
A choice the instrument leaves to human judgment, not a mechanical step.
dual-hatted
A person or firm that wears more than one regulatory hat, such as adviser and broker.
custody
Holding client or trust assets, or having the power to move them.
journal
An accounting entry that records a change to the books.
activity log
A system record of who viewed, changed, exported, or approved something, and when.
audit trail
The linked history that lets someone reconstruct why a step was taken.
human-in-the-loop
A named person must approve before anything writes or moves.
hash-chain record
A write-once log (The Record) where each entry links to the last, so later edits show.
Closing
Forward Deployed Agents are on-site observers. An observatory examination is their first assignment: watch the trust or advisory platform as it really operates, including its journals and activity logs; find patterns that improve the work and warn of possible abuse; and report what they find — without taking the wheel.
An AI trained by trust professionals. Certified for trust administration. Deployed on r.team.
rAgent is apprenticed to named practitioners: it learns from what they know, then takes the same test they do — so you can see, item by item, where it reasons like they do and where it doesn't.
Split topics — I’ll answer each one, not one blended search.
Help
The apprenticeship
Three activities. Scores stand side by side on the rail. Enter an email to Ask, Take a test, or Have rAgent take it. Help, About, and r.Team stay open without it.
Study
Use Ask. Type a question — rAgent answers from practitioner notes and cites the chunk it used. Definition questions get the working method first.
Comma-separated topics are answered one section each, not as one blended dump.
Take the test
Use Have rAgent take it. rAgent takes the same mixed drill a practitioner takes. Its score lands next to yours on the rail.
Show this test in Ask — after rAgent’s test (or after you finish), one click opens Ask with every stem, A–D choice, selected vs correct, and why. Per-item Ask about this miss still works.
Take it beside them
Use Take a test. You take the same exam. Choose A–D, submit, then review. Your score stands beside rAgent’s, miss by miss.
Recent tests stay on the rail. Click one to load its misses.
About
rAgent Team
Reasoning agents, apprenticed to named practitioners, certified for trust administration, deployed on r.team.
How the apprenticeship works
Study — rAgent reads a practitioner’s notes and answers from them, citing what it used.
Take the test — rAgent takes the same mixed drill a practitioner takes.
Take it beside them — the practitioner takes it too; scores stand side by side, miss by miss.
When rAgent scores beside a named practitioner, it is certified and deployed on r.team under that practitioner’s name.
The craft
Fiduciary reasoning: read the governing instrument, apply the standard, weigh these facts, write a rationale another officer can follow. Series 65 notes are in the corpus for exam form; they are not the whole product.