Reg BI compliance on every retail recommendation, every night — the agents bring the evidence, a principal decides
Every recommendation reviewed overnight; every flag decided by a principal, with evidence.
See one case, screen by screen ↓Reg BI compliance review is how a broker-dealer's supervisors check retail recommendations against the Reg BI compliance requirements: disclose, weigh costs and reasonably available alternatives against the client's investment profile, manage conflicts, and keep records that show it. Reg BI Compliance Review gives that work to six agents that test every retail recommendation and its written rationale overnight, so a supervising principal decides each flag with the evidence in front of them.
AI in asset management: a buyer's guide to agentic AI for investment managers →
Updated
Reg BI compliance asks a broker-dealer to weigh costs and reasonably available alternatives and to fit every retail recommendation to the client's investment profile — and to show that it did. The evidence lives in five places: the profile and meeting notes in the CRM, the share classes and breakpoints in the prospectus, the shelf in product cost data, the payout grid in the compensation system and the advisor's reason on a rationale form. So supervising principals sample. At thousands of recommendations a day, a 2% sample leaves most of them unread.
A principal reading one ticket at a time cannot see that 37 tickets from one branch carry the same five words. FINRA's 2026 oversight report again finds firms not documenting their comparison of costs and alternatives and not following up on generic or duplicated rationales. And time matters: an L-share annuity sold to a client with a ten-year horizon can be exchanged at no cost inside the free-look window. Find it a month later and the fix becomes restitution.
Estimated: about 38 minutes per flagged review by hand against about 6 with the evidence assembled (the pilot measure in the working solution's sample data), for 61 items, in 8-hour days.
Six agents test costs, profile fit, rationale and conflicts on every recommendation; a principal decides each flag.
Quillfield Securities, the broker-dealer of Quillfield Capital's wealth business, has 1,184 financial professionals in 62 branches and 14 OSJs. It is Wednesday, October 14, 2026. The overnight review of Tuesday's recommendations finished at 6:40 a.m. Here is how Renee Castillo works with the agents, screen by screen, in the working solution.
Renee's Today screen puts the agents' conversation beside what needs her, not an alert list. The agents reviewed all 4,812 recommendations from Tuesday: 4,519 cleared with their evidence filed, 293 need a principal and 61 are hers, 37 due today. Two advisor replies are waiting, already pre-checked. One next step sits at the top: 37 identical rationales from the Scottsdale branch.
Review agents: “The biggest pattern: 37 copies of one rationale from Scottsdale, traced to a branch file. FINRA lists generic and duplicated rationales as a finding.”
On the constellation each of the 4,812 recommendations is a dot, and dots sit close together when their written rationales say the same thing. Firm-approved templates show green and are cleared — 1,206 on the rebalance template RB-2, 140 on tax-loss harvesting, 64 on CD ladders. Copied reasons stand out as tight orange clusters: 37 from Scottsdale, 26 from Coral Gables, 22 from Plano. She can color the map by outcome, test, product or region, or narrow it to her OSJ and clients 65+.
“Closer together = more alike in wording. Tight clusters are copied reasons; green clusters are firm-approved templates.”
Four advisors wrote exactly “Client seeks growth and income.” on 37 recommendations: Kyle Barrett 22, Dana Whitlock 7, Sean Murphy 5, Lisa Tran 3. They cover 21 variable annuities, 9 structured notes and 7 A-share funds, and 26 of the clients are 65 or older. The agent traced the phrase to Scottsdale_RBI_quick_notes.docx on the branch shared drive, last changed June 30, 2026 — not a firm template — and looked back on its own: 186 earlier copies since July 1, 6 of which may need restitution, $48,210 estimated.
The Remediation & Records agent suggests one branch follow-up with Greg Halvorsen covering all 37, so she doesn't open 37 tickets.
Kyle recommended $410,000 from a maturing CD ladder into the Harborline Horizon Variable Annuity, L-share. Her profile reads moderately conservative, income and preservation, a 10+ year horizon, low liquidity need. The Care agent sets out the costs and reasonably available alternatives side by side: 4.03% a year all-in for the L-share against 2.58% for the A-share of the same contract, with the same rider and sub-accounts — $5,945 a year, about $59,450 over ten years. The Profile agent finds annuities would be 34% of her liquid net worth, over the firm's 30% guideline at 70+. The Conflicts agent shows a 6.00% concession against 2.00% plus a 0.25% trail. The 30-day free-look period for an owner 65 or older ends November 12.
Every figure is cited: a click on [2] opens the Harborline prospectus, page 15, the A-share fee table.
Renee asks the agents beside the file. The Care agent answers with its sources: SecureRate 5 pays 4.85% guaranteed for 5 years with no annual fee, which fits preservation, but it has no lifetime income rider; if she wants lifetime income, the like-for-like alternative is the A-share. The Profile agent finds nothing in her file that supports the L-share: no note about needing the money within 4 years, and $36,000 cash plus a balanced fund outside the annuity.
Review agents: “The quickest fix for Mrs. Whitcomb is an exchange to the A-share before November 12, which costs her nothing.”
The Remediation & Records agent has drafted the follow-up from the four findings; Renee edits it. The agents suggest reason code FU-COST — costs or share class not justified — and she can change it. The reply is due Friday, October 16, copied to the branch manager, with an A-share exchange request and client letter prepared for her approval. Then she opens one branch follow-up for the 37 and sends the 6 look-back cases to the CCO.
“Nothing is sent until you confirm.” · “Closure needs a second person: a compliance analyst confirms the advisor's fix before this case closes.”
Same L-share, $180,000, from a Tucson advisor. The Care agent passes it (0.81): a September 22 CRM note says Harold will buy a home in about 3 years, which suits a 4-year surrender. The Profile agent flags it (0.86): his profile still says 10+ years and was last updated in February 2023, 44 months ago, past the firm's 36-month limit. Renee sides with Care and asks the advisor to update the profile, reason code OVR-DATA-STALE.
“Your choice and its reason code train neither agent automatically; the analyst reviews disagreements weekly.”
Kyle's screen lists it in plain words: her profile says she won't need the money for 10+ years, so the L-share charges 1.45% a year more for early access she doesn't need; his rationale is the same as 36 others and names no alternative; her free-look ends November 12. He writes his reply, the agents pre-check it — alternatives named, their costs, the fit with her profile, the share class explained, 4 of 4 — and he proposes the A-share exchange.
“The agents check your reply before Renee sees it. They never write it for you.”
The agent checks Kyle's reply against the original findings: it names the A-share at 2.58%, SecureRate 5 at 4.85% and the CD ladder at 4.30%, ties the choice to her 10+ year horizon and lifetime income goal, and proposes the exchange inside free-look. The client letter tells Mrs. Whitcomb her yearly charges fall from 4.03% to 2.58% at no cost to her. Renee approves the letter and proposes the closure; Priyanka Desai confirms it on October 16 — two names on the record.
The letter is drafted by Kyle Barrett from the firm template CL-7; the drafter never approves it.
The look-back sends 6 cases, $48,210 estimated in all. For Dorothy Kessler, 78, $265,000 into the same L-share on July 21, past her free-look: $895 of excess cost to date, $10,600 of commission difference refunded and $485 of interest — $11,980, with an exchange to the A-share and the surrender charge waived by the issuer. Priyanka prepares each calculation; Martin approves, rejects or asks for more facts.
“Estimates stay drafts until the CCO approves.”
The same night's recommendations, seen by five of the people who carry them — what their week looked like, and what it looks like now.
Six agents test costs, profile fit, rationale and conflicts on every recommendation; a principal decides each flag.
Gathers every retail recommendation of the day — solicited trades, switches, account-type changes, rollovers and recorded holds — and joins each to the investment profile, product costs and share classes, the shelf of alternatives, holdings and the payout grid.
Tests costs and reasonably available alternatives on the firm's shelf over the client's horizon, including share class, breakpoints, riders and liquidation costs.
Tests fit with the investment profile: age, objective, risk tolerance, horizon, liquidity, concentration, complex products for clients 65+ and profiles older than 36 months.
Places every written rationale by similarity, finds copied, generic and contradicting reasons, recognizes firm-approved templates, traces a copied phrase to its source file and looks back for earlier copies.
Checks compensation: the payout differential between share classes, proprietary products, sales contests, non-cash compensation and unusual patterns by representative.
Drafts follow-ups and client letters, checks advisor replies against the original findings, estimates restitution, seals each case file and builds exam packs.
Decide, confirm, remediate. The agents propose; a named person decides.
Ask about the night, a client, a cluster or a rule — or tell the agents what to do. Every answer cites its source.
What changed since yesterday?
Flags are 6.1% of recommendations, against 6.4% on Monday. The Scottsdale cluster is the largest one-night cluster since the overnight review went live on October 5. A look-back found 186 earlier copies from July 1 to October 2, when recommendations were only sampled. The compensation file arrived 48 minutes late; the Conflicts agent waited for it, so nothing was cleared without it.
Why is Eleanor Whitcomb's annuity flagged?
The same contract's A-share, with the same rider and sub-accounts, costs 1.45% a year less: 2.58% against 4.03%. Her horizon is 10+ years, so the L-share's short surrender period buys nothing. The Profile, Rationale and Conflicts agents flag it too.
Why not the guaranteed annuity?
SecureRate 5 pays 4.85% guaranteed for 5 years with no annual fee, which fits “preservation”. But it has no lifetime income rider. If she wants lifetime income, the like-for-like alternative is the A-share of the same contract. Kyle should say which she wants.
Assign my disagreements to me first
Done. Your 5 disagreements are now at the top of your queue, ordered by due time. Harold Brandt is first.
13 screens from the working solution, on its sample data. Pick one to see it large.
The Reg BI tests are settings the CCO approves. A change is proposed by one person, tested on last week's recommendations and approved by another, and it is on the audit trail.
| Setting | Default | Choose from |
|---|---|---|
| Care agent flags a cheaper same-feature alternative | ≥ 0.25% a year over the client's horizon | Proposed and approved as a change request — e.g. CR-26-034 to ≥ 0.20% a year |
| Annuities as a share of liquid net worth, clients 70+ | 30% | Set from the written supervisory procedures; changed by proposal and CCO approval |
| Complex product for a client 65+ | Always to a person | Same-day decision, by 5:00 p.m., then the OSJ manager and the CCO |
| Investment profile older than | 36 months | A stale profile is a question for the advisor |
| Cluster of near-identical rationales | Similarity ≥ 0.92, ≥ 5 members | Firm-approved templates are recognized and cleared |
| Rationale too short to be specific | Under 12 words | Or a rationale that contradicts the profile |
| Payout differential against the nearest alternative | Over 0.50 points | Also proprietary products and sales-contest windows |
| Auto-clear | Only when all four tests pass at confidence ≥ 0.85 | A client 75+ with a complex product is never auto-cleared |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample data · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures (SEC order, July 30, 2024 — $140,000 penalty with $9 million of remediation; FINRA's Reg BI exams of about 1,000 broker-dealers, issues in about half, via Mintz; FINRA, 2026 — $475,000 fine and about $634,000 restitution over variable annuity share classes; FINRA 2026 Annual Regulatory Oversight Report, December 9, 2025; SEC Regulation Best Interest, Release 34-86031; Rules 17a-3(a)(35) and 17a-4(e)(5)). People, firms, clients and products named on this page are fictional — characters and sample data in the working solution.
Reg BI compliance review is how a broker-dealer's supervisors check retail recommendations against the Reg BI compliance requirements: disclose, weigh costs and reasonably available alternatives against the client's investment profile, manage conflicts, and keep records that show it. Reg BI Compliance Review gives that work to six agents that test every retail recommendation and its written rationale overnight, so a supervising principal decides each flag with the evidence in front of them.
The care obligation asks the firm to understand a product's costs and risks, weigh reasonably available alternatives and fit the recommendation to the client's investment profile. The Care agent tests costs and same-feature alternatives on the firm's shelf over the client's horizon, and the Profile agent tests objective, risk tolerance, horizon, liquidity and concentration — each figure cited to its source.
For each recommendation it sets the recommended product beside the alternatives on the firm's shelf, with the all-in yearly cost, the cost over the client's horizon and what each means for the client. In the sample case that is the L-share at 4.03% a year, the same contract's A-share at 2.58%, a guaranteed annuity, a conservative portfolio and keeping the CD ladder, each with its prospectus page — the best interest documentation, written down with its sources.
The Rationale agent places every written rationale by similarity, so identical reasons form a tight cluster on the constellation, and flags rationales under 12 words or that contradict the profile. It recognizes firm-approved templates so they are not mistaken for copying, traces a copied phrase to its source file and looks back for earlier copies.
That is the design goal: 100% of retail recommendations reviewed every night. In the working solution all 4,812 of one day's recommendations are reviewed by 6:40 a.m.; 4,519 clear with their evidence filed and the rest go to a principal. Anything the agents cannot join to its data waits, and is never cleared.
Yes. The agents flag and draft; only the supervising principal decides a flagged item, always with a reason code. An analyst confirms every closure, the CCO decides restitution and approves every test, threshold and agent change, and nothing reaches a client or advisor without approval.
Every agent step and decision is written to an append-only, fingerprinted audit trail with the agent's version, and case files are sealed in a write-once archive for 6 years under Rule 17a-4(e)(5). An exam pack exports the recommendations in scope with their rationale, profile snapshot, cost and alternatives evidence, findings, decisions and both names, with client data masked by default.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your written supervisory procedures, product shelf, payout grids and a sample of past recommendations — and it goes live once every quality gate has passed. First-time setup pilots the agents on 2,000 past reviews so you go live on measured results.
Accounts open the same day, and not-in-good-order stops before the custodian sees it.
1-min demoSource of Wealth VerificationEvery dollar of a client's wealth traced to cited evidence before the BSA/AML Officer signs.
1-min demoIRA Rollover AnalysisEvery rollover file carries a specific, cited best-interest rationale, ready in minutes.
1-min demoWe'll run Reg BI Compliance Review on a sample of your own past recommendations and their rationales.