SI and AI services · Service
SI and AI compliance and supervision controls
Approval gates, audit trails, human review and recordkeeping built into AI-assisted workflows, mapped to the rules your firm actually answers to.
Short answer
SI and AI compliance controls make sure AI output is reviewed, approved and kept before it reaches a client or an official record. We write the policy and data-flow maps, map them to FINRA, SEC, bank and insurance expectations that apply to you, then build the controls into Salesforce, HubSpot, Claude and ChatGPT so supervision happens inside the workflow.
This service sits in Step 4, human-in-the-loop design of our SI and AI deployment sequence. See all six steps.
Regulators have not written a separate rulebook for AI. They expect existing rules to apply: communications with the public are still communications, supervision still has to be reasonable, books and records still have to be kept, and models used in decisions still need governance. FINRA said as much in Regulatory Notice 24-09.
That makes the work practical. We find where AI touches a regulated activity, put a person and a record at that point, and make the record easy to produce. Your compliance team approves the controls; we configure and document them.
How we map SI and AI use to your obligations
| Obligation | Where AI touches it | Control we build |
|---|---|---|
| FINRA Rule 2210 (communications) | AI-drafted emails, posts and client letters | Draft-only output, principal approval step for retail communications, approved-content library |
| FINRA Rule 3110 (supervision) | Reps using AI in client work | Usage policy, review queues, sampling and exception reports |
| SEC Rule 17a-4 and FINRA 4511 (books and records) | Prompts, outputs and AI-written notes | Retention settings, Compliance API or log export to your archive, CRM activity records |
| Regulation S-P | Client data sent to AI tools | Data classification, approved tools list, vendor terms review |
| Federal Reserve SR 26-2 (model risk) | AI used in decisions or scoring | AI register, validation evidence, change log per model or prompt |
| NAIC AI model bulletin (insurers) | AI in underwriting, claims or marketing | Written AI program, testing for unfair outcomes, vendor oversight |
| GDPR and UK GDPR | Personal data of EU or UK clients | Lawful basis, data-flow maps, retention limits, processor terms |
We configure and document controls. Legal interpretation and regulatory sign-off stay with your compliance and legal teams.
What you get
- Acceptable-use policy and data classification for AI use
- Data-flow maps for up to 3 AI tools
- Supervision mapping to the rules that apply to your firm
- An AI register (model and prompt inventory) with owners and review dates
- Human-review and approval design, then the build in Salesforce or HubSpot
- Logging and retention configured to feed your archive
- One-page staff guidance and a walkthrough with compliance
How it comes together
- Find the touchpoints. Every place AI drafts, decides, or reads client data.
- Map obligations. Which rule applies at each point and what evidence it needs.
- Design controls. Draft fields, approval steps, sampling, retention.
- Build and test. Approval processes, flows, audit fields and log exports.
- Hand over. Your compliance team approves; we document how to produce records on request.
Is this the right call?
Good fit when
- Broker-dealers, RIAs, banks, insurers and lenders putting AI near client communications
- Firms that allowed AI use and now need to show supervision
- Exams or audits coming up with AI on the agenda
Think twice when
- You need a legal opinion; we work alongside your counsel
- Vendor due diligence questionnaires only; we can support, but your vendor management team owns them
Mistakes we help you avoid
Writing policy with no control behind it
A policy that says "review AI output" means little if the CRM lets AI write straight to the record. Build the review step into the system.
Keeping outputs but not prompts
To show what happened, you need the input, the output, the model and who approved it.
One-time approval
Models and prompts change. Re-run your evaluation set and log the result each time.
In practice
Our founder was COO of a wealth management firm before starting Vantage Point, and our practice serves financial services and other regulated industries. Supervision and recordkeeping are part of the design from day one.
Questions we get
Can a broker-dealer or RIA use Claude or ChatGPT with client data?
Yes, with enterprise plans, approved data paths, human review of client-facing output and records kept under your retention rules. FINRA Regulatory Notice 24-09 reminds firms that existing rules apply to AI use, so the controls follow the same rules you already supervise.
Do AI prompts and outputs count as books and records?
They can, depending on what they contain and how they are used. We set up retention and export so you can keep what your compliance team decides must be kept, and keep AI-written notes in the CRM as normal activity records.
What is SR 26-2?
SR 26-2 is the Federal Reserve's April 2026 supervisory guidance on model risk management, which updates the expectations banks followed under SR 11-7. Banks using AI in decisions should keep an inventory, validation evidence and change records for those models.
Do you give legal or regulatory sign-off?
No. We design, build and document controls. Your compliance and legal teams approve them.
Free, about 3 minutes: take the SI and AI readiness quiz. Answer 10 questions and get an instant readiness score across 10 areas. No email needed to see your results.
Updated October 7, 2026. Vendor features and terms change; we confirm them for your org before scoping.
