PWOS in motion
Watch the operating system govern one decision
PWOS is the in-house system Protocol Wealth's advisory work runs on. Below is one AI-assisted decision moving through it — from the question, to a governed draft, to a human fiduciary's decision, to a record that can be reconstructed. The point is not speed. The point is that each layer exists to make the layer above it trustworthy.
What this page is. A plain-language walkthrough of the governance flow that sits under the firm's AI-assisted work. It illustrates the process; it is not an interactive product, a performance claim, or a description of any specific client's situation.
What this page is not. It does not depict autonomous advice. AI never makes a fiduciary decision, never contacts a client, and never moves money. Where a capability is still being built, it is labelled in development.
The question, and the inputs
LiveAn advisor poses a real question — for example, "what does the current environment mean for this client, and is anything worth raising with them?" PWOS assembles the inputs: the framework's systematic read of market conditions (regime and asset-durability scoring from public data) and the client's own context held in the firm's records.
The market read is rules-based and reproducible — the same inputs produce the same reading. See what the framework is →
AI drafts — under constraints it cannot bypass
LiveAI assists by surfacing the relevant observations and drafting a plain-language read for the advisor. The constraints are structural, not promises:
- Client identifiers are kept out of the models. They are removed before anything reaches a model; the egress check fails closed, so a call that would carry them simply does not happen.
- The model drafts; it does not act. It holds no keys, signs nothing, moves no funds, and has no path to send anything to a client.
- US-region inference, no training on the firm's data under the firm's vendor terms.
The output is a draft for a human — nothing more. Our AI principles →
A human fiduciary decides
LiveThe draft surfaces to a human advisor who knows the client — risk tolerance, time horizon, tax position, liquidity. The advisor accepts, edits, or rejects it. Anything client-facing additionally clears the firm's Marketing-Rule review. There is no automatic path to a client; a human approves and sends.
In development: an automated compliance gate that holds every AI-generated client-facing deliverable for explicit sign-off before an advisor can even open it. Today that gate is the advisor's manual review plus the firm's existing compliance review.
A record that can be reconstructed
LiveEach step — who asked, what the model was shown, what it drafted, and what the advisor decided — is recorded in an append-only, tamper-evident trail that is mirrored to write-once storage. The path from question to decision can be reconstructed after the fact, which is what a regulatory examiner needs to see.
This is what "compliance as architecture" means in practice — the record is a property of the system, not a document assembled later. Trust Center →
Where this sits today
The advisory practice is live. PWOS runs the practice.
Live
The advisory firm, and the governance layer above — identifiers kept out of the models, human-in-the-loop with no auto-send, and the reconstructable audit trail.
In development
The multi-agent deliverable workflow and the automated compliance emission gate described above — designed, not yet a product.
A trajectory, not an offer
Licensing PWOS to other firms is a direction we're building toward — it is not currently a purchasable product.
Protocol Wealth, LLC is an SEC-registered investment adviser (CRD #335298). Registration does not imply a particular level of skill or training. This page is for informational purposes only and describes the firm's internal process; it is not investment, legal, or tax advice, a recommendation, or an offer of any product or service. AI is used to assist the firm's professionals; a human fiduciary reviews and is responsible for every client-facing decision. Capabilities marked "in development" are not currently offered.