AI at Protocol Wealth
How we use AI, and why it matters that you know.
AI is a tool. Like every powerful tool before it, the question is not whether to use it but how to use it without losing something that matters more than the efficiency it provides.
The Concern
Technology is a tool — and tools have consequences
A car is a tool. It reshaped cities, created industries, and gave people freedom of movement that previous generations could not have imagined. It also kills tens of thousands of people every year. We don't refuse to drive, but we insist on seatbelts, speed limits, licensing, and liability. The tool is powerful precisely because it's dangerous. The governance exists because we decided the power was worth having — but only under constraints.
AI is the same kind of tool, with a characteristic that cars don't share: it can create positive feedback loops. A system that generates its own inputs, evaluates its own outputs, and acts on its own conclusions can compound errors faster than any human can catch them. A car that drifts left will hit a guardrail. A model that drifts left will generate data that confirms the drift, train on that data, and accelerate.
This is not a speculative risk. It is a structural property of systems that feed their outputs back into their inputs without an external check. The financial industry has seen this pattern before — in algorithmic trading, in credit-scoring feedback loops, in models that performed beautifully on backtests built from the same data they were trained on.
Our operating thesis argues that AI is the current layer shift — that it commoditizes analysis the way the printing press commoditized copying and Excel commoditized arithmetic. What survives the shift is the verification layer: judgment under ambiguity, fiduciary accountability, the ability to connect a recommendation to the reality of a specific person's life. We believe that. But believing it is not enough. You have to build systems that enforce it.
The Position
Human in the loop — not because of AI, but because of us
The phrase "human in the loop" gets used as an AI safety slogan. We mean something more specific. We mean that humans cannot afford to lose their role as verifiers — not because the machines are unreliable (though they sometimes are), but because the act of verifying is what builds and maintains the skill to verify.
A pilot who never hand-flies the aircraft loses the ability to hand-fly the aircraft. An analyst who never builds a model from assumptions loses the ability to spot a broken assumption. A fiduciary who never reads the primary source — the custodial statement, the tax return, the trust document — and instead reads only the AI summary of the primary source, will eventually lose the ability to catch the thing the summary got wrong.
We keep humans in the loop not as a safety theater gesture but because the loop is where competence lives. Outsource the verification and you don't just risk errors — you lose the capacity to find them.
This is our framework applied to technology: traditional goals through modern infrastructure, with the discipline to use new tools only where they serve and preserve the ability to revert when they don't.
Guiding Principles
Five principles that govern how we use AI
These are not aspirations. They are operating constraints — enforced in our workflows, our compliance program, and our technology architecture.
Humans decide, AI assists
The model indexes, retrieves, drafts, and surfaces. Humans judge, conclude, and own the outcome. Every output has a named human who can defend it without the model in the room. Accountability does not transfer to the tool.
In practice, this means every material AI-assisted output — a portfolio recommendation, a compliance review, a client communication — is reviewed by a registered investment adviser representative before it reaches you. We maintain audit trails of AI-assisted work so that the reasoning can be examined after the fact by our compliance team, by regulators, or by you.
Build the skill before adopting the tool
The capability comes first; the automation comes second. This applies to ourselves, our team, and the next generation. Skills you let atrophy before mastering them are skills you've permanently outsourced.
We don't use AI to skip steps we haven't learned how to do manually. Our investment framework — the regime detection, the durability scoring, the quality checks described on our investing page — was developed, tested, and refined by human analysts before any of it was augmented with AI tooling. The AI makes the process faster and more consistent. It did not replace the understanding that built the process.
AI use is a decision, not a default
Every prompt is a choice not to think. Match tools to problems, not problems to tools. When AI use feels frictionless, add friction back in. "Because we can" is not a reason.
We evaluate each use of AI against the question: does this serve the client's interest, or does it serve our convenience? A model that drafts a portfolio commentary saves time. A model that generates a portfolio recommendation without a human evaluating the assumptions behind it saves time at the client's expense. The first is appropriate. The second is not. The distinction matters more as the tools get better at producing plausible output.
Protect data and maintain reversibility
Sensitive data stays inside boundaries you control. Critical knowledge doesn't get trapped in prompts, agents, or vendor systems. Every AI-dependent process has a manual fallback. Lock-in is a risk, not a feature.
We operate under a Zero Data Retention agreement with Anthropic — inputs and outputs sent via the Claude API are not retained beyond request duration, our data is excluded from model training, and inference is restricted to US-based infrastructure. Client information is anonymized and tokenized before submission where feasible. We do not submit Social Security numbers, account numbers, or private keys to external AI tools. If our external AI provider's terms change, we retain the ability to migrate to self-hosted models.
These controls are detailed in our Privacy Policy and Terms of Service.
Disclose honestly and watch what you amplify
Be transparent about where AI is used. Don't manufacture volume you don't actually want. Bad inputs produce confidently bad outputs at scale, so source quality matters more, not less.
We disclose our use of AI on this page, in our Privacy Policy, Terms of Service, and Disclosures. Our security page names every vendor in the data path. Our open-source repositories let you verify how AI is integrated into our analytical tools rather than taking our word for it.
We do not use AI to generate high-volume marketing content, synthetic testimonials, or engagement-optimized communications. When we publish, a human wrote it or a human reviewed and edited what a model drafted — and the human's name is on it.
In Practice
What AI does and does not do at Protocol Wealth
What AI does
- ✓ Monitors portfolios for drift, regime changes, and anomalies around the clock
- ✓ Retrieves and organizes data from custodians, market feeds, and regulatory databases
- ✓ Drafts research summaries, compliance checks, and operational documentation for human review
- ✓ Flags inconsistencies, catches numerical errors, and provides a second pair of eyes on adviser work
- ✓ Scores assets against our 8-check quality framework with consistency a human team cannot match at scale
What AI does not do
- ✗ Make final investment decisions on your behalf
- ✗ Override the fiduciary judgment of your investment adviser
- ✗ Execute trades or transfer client assets
- ✗ Determine your fees, account access, or legal rights
- ✗ Produce client-facing recommendations without human review and sign-off
Your rights
As described in our Terms of Service and Privacy Policy, you have the right to:
- • Request an explanation of the human review that occurred for any AI-assisted output you received
- • Inquire about what AI tools were involved in any analysis, report, or recommendation
- • Opt out of external AI processing of your data
- • Opt out of all AI-assisted workflows related to your account, to the extent feasible
Next
Keep reading
Operating Thesis
Why AI is the current layer shift, and what the verification layer looks like.
Framework
How we think, and what we build around what matters to you.
Privacy Policy
How your data is handled, including AI-specific controls and your data rights.
Security
Every vendor in the data path, named publicly.
Last updated: June 5, 2026. Protocol Wealth LLC is a SEC-registered investment adviser (CRD #335298). See our Form ADV for authoritative regulatory disclosures.
Registration with the SEC does not imply a certain level of skill or training. Advisory services are provided only under a signed advisory agreement. This page describes our approach to AI; specific recommendations and workflows depend on individual client circumstances.
AI tools do not make autonomous investment decisions, do not execute trades, and do not provide final personalized advice without human review. AI-generated output may contain errors or omissions and is not relied on as the sole basis for advisory decisions. We maintain supervisory records of AI-assisted workflows as part of our compliance program.
For details on how your data is handled when AI tools are used, see our Privacy Policy. For the contractual terms governing AI-assisted services, see our Terms of Service. For regulatory and risk disclosures, see our Disclosures.
All investments involve risk, including the potential loss of principal. Past performance does not guarantee future results.