Client operations, books and records, and the questions a regulator asks. Click any of these.
Registered investment advisers and retirement plan advisory firms. Client records, meeting notes, quarterly reviews, fee billing, required minimum distributions, beneficiary reviews and the compliance calendar. The advisory pack is built for exactly this firm.
We work with two Farm Credit System institutions, which is a credential very few implementation firms hold. Agreement lifecycle management, loan origination platforms, core banking systems, and payment support that touches legal disclosure, stored procedure logic and the user interface in the same change. Farm Credit institutions are member owned and the board credit committee is a real gate, so the vocabulary that matters here is model risk management, third party risk and IT security guidance rather than generic AI governance.
Escrow automation with multi entity invoicing, commission automation across markets and agent types, document automation, and standing up a new lending business line inside an existing shared platform without breaking the entities already in it. Deloitte's 2026 commercial real estate outlook found only 1% of executives now report transformative AI impact, down from 12% the year before. This buyer has just been disappointed by AI, which is exactly the opening for something governed and measurable.
Nothing on this site matters more than that sentence. It is the problem.
The compliance obligation is real, the method is a person, and the method does not scale or survive a departure.
A record with no freshness signal is a record you cannot rely on and cannot defend.
The spreadsheet between two systems is where institutional knowledge goes to become one person's problem.
Mayeshal reads the email, the meeting, the documents and the record together, and marks what it could not find rather than estimating. Every answer names the document and the page it came from. That is what closes the gap.
63% of registered investment advisers now use AI, more than double the 2023 figure, but only about one in ten fully integrate it into business strategy. Adoption stopped being the question some time ago. What is left is governance, measurement and being able to show your work.
These are our own paraphrases of problems clients have described to us. They are not published quotations and are not attributed to any client.
Eight in ten adviser firms have formally adopted AI tools. Seven in ten restrict it to internal use only, and roughly one in ten permit client-facing use at all. Meanwhile compliance officers named AI the number one compliance topic of 2026, ahead of cybersecurity and privacy by a wide margin, in the largest single-response result that survey has recorded in twenty-one years.
The specific gaps are documented. Most firms now have an acceptable use policy and a tool inventory. Far fewer have a formal human-in-the-loop policy, fewer still have policies for testing and validating outputs, and fewer than a third have policies covering third-party AI use at all.
And the tool count keeps rising underneath all of that. Notetaking went from one product to fourteen in a single year. Every one of them is another vendor holding client meeting content, another vendor oversight obligation, and another copy of the same conversation.
Examiners are reportedly asking advisers for a written acceptable use policy naming permitted and prohibited tools and which client data categories may be processed; vendor oversight documentation per third-party AI tool covering data handling, storage location, retention and written security representations; evidence of human supervisory review before AI-assisted output reaches a client; staff training records with dates and completion; and an inventory of AI tools, including the ones running on personal devices.
Mayeshal changes the arithmetic on four of those five. One governed tenant means one vendor oversight file rather than five. The provider accounts are in your firm's name, so the retention and training answer is a term in a contract you signed directly. Every interaction is logged with the user and the timestamp and can be exported. And approval gates mean supervisory review is a recorded event rather than an assertion.
The fifth, the shadow tools on personal devices, is not a software problem. But the reason people use them is usually that the sanctioned tool cannot see anything useful, and that part we can help with.
Under Advisers Act Rule 204-2, records are generally kept five years from the end of the fiscal year in which they were last modified, with the first two years in an easily accessible place, stored securely, organised, and producible on request. What matters is not who or what created the content, it is how the content is used. An AI meeting summary relied on to inform a recommendation is a record. An AI-drafted client letter is a record. AI-generated commentary used in a newsletter is a record.
The practical guidance is to retain both the prompt and the output for material interactions, the full transcript alongside the summary, and the original output even after a human edited it, so the evolution of the material can be shown.
Mayeshal keeps all of that by construction rather than by policy, because every model call passes through one wrapper that records it, every document and summary lives in your own database, and the audit log covers every tool call, approval and export. The retention posture on compliance items is tagged for books and records from the day the pack is installed.
This is a description of our product, not legal or compliance advice. Your chief compliance officer owns the policy.
The pack composes onto the whole ground, the whole spine and six jobs, and connects the CRM, the accounting system, the document store and a custodial or portfolio feed. Status: built and signed for advisory, first customer build for lending.
Objective. Make the CRM say what is true, by reading the correspondence and the documents that currently sit outside it.
Correspondence, meeting notes and documents are read alongside the record, so the client file reflects what happened rather than what someone remembered to type. Staleness is visible: every answer carries the date of the evidence behind it.
Requirement: a retirement and private wealth advisory firm, which described a standing disconnect between the quality of the data in the system and reality, and no way of knowing how stale or current the information is.
Objective. Meet the records obligation without a person moving email into a feed by hand.
Correspondence is captured against the client record with a retention posture and one audit trail. Ten custom objects carry confidentiality and retention tags that the permission resolver enforces regardless of which surface is reading them.
Requirement: a retirement and private wealth advisory firm, which was copying and pasting email correspondence onto client feeds for its records.
Objective. Reconcile recurring fees against the accounting system without a monthly manual pass.
Three correlation rules resolve the client, the compensation event and the recurring fee against the accounting system through the canonical entity map. Corrections are staged with a payload preview and approved before they land.
Requirement: a retirement and private wealth advisory firm, where invoicing and marrying the accounting system to invoices was named as the standing operational irritant.
Objective. Read across the loan origination, core banking and agreement systems a lender already runs, and show what is outstanding on a file.
The baseline ground and spine apply unchanged. The lending specific objects, roles and outstanding items rules would be built with the first customer. The credential that makes this credible is operational rather than theoretical: two Farm Credit System institutions, and a stack spanning agreement lifecycle management, loan origination, core banking and a payment layer that touches legal, stored procedure and interface concerns at once.
Requirement: an agricultural lender in the Farm Credit System, where the mapping spreadsheet between two systems had fallen behind and become one person's problem.
Transcripts to summaries, decisions and action items.
The eight-step preparation brief that pulls the meeting, the record, the email thread and the documents into one page before the call.
Drafts every reply. Sends none of them without approval.
What changed since yesterday, per person.
The things due on a cycle, prepared and routed for approval.
The packet, assembled. Delivery is adviser-approved.
Internal preparation only. The platform never places a trade.
A deterministic scheduler. Outreach drafts route through the email approval gate.
Cross-system reconciliation, with corrections to the accounting system gated behind a payload preview.
Filing countdowns. People still prepare and file.
Research with citations, over what you hold rather than over the open web.
One household, one page.
Compliance review before anything is distributed, because the advertising rule applies.
Document extraction against the record, with outreach through the approval gate.
Six roles install with it: principal adviser, investment adviser, operations manager, client service associate, compliance officer and portfolio analyst. Grants are conservative and tools are deny by default.
Agreement management, loan origination and the core banking system read together, so the status of a file is one answer rather than three systems and a phone call.
Loan files, appraisals, financial statements, borrower correspondence and the scanned pages inside them, opened, indexed, summarised and citable to the page.
Files missing a required document, mapping work that is behind, records that were opened under pressure with mandatory fields blank.
Which is the difference between a helpful tool and one a risk committee will sign off.
Context for 2026: USDA forecasts 2026 net farm income at $153.4 billion, down slightly, with sector debt up 5.2% to $624.7 billion and working capital forecast to fall 9.2%. Working capital pressure is where credit operations get busy, and busy is where the seams between systems start costing money.
We are an implementation and software partner, not a model risk function. Your chief risk officer owns the model risk assessment. What we do is make sure the evidence they need is a query rather than a project.
State the status plainly, because reviewers on the buying side ask for build state per claim and should get it.
Whatever platform holds your client, household and engagement records, mirrored into a database that belongs to your tenant alone with count parity verified against the source.
Connected through the canonical entity map, so a fee, a compensation event and a client resolve to one another without a monthly manual pass.
Client agreements, statements, appraisals and loan files, opened page by page, indexed, summarised and citable to the page.
Read for preparation and reporting. The platform never places a trade against it.
Correspondence and meeting notes read alongside the record, which is the whole point of the books and records capability.
On the credit side that means agreement lifecycle management, loan origination and core banking. Every write back is staged, whitelisted and approved by a person.
The pack exists as a signed, versioned, published artifact. Its dry-run install passes every step. The platform underneath it is running in production today for a law firm, which is the same control plane, the same permission model, the same audit log and the same cost meter.
What has not happened is a production advisory tenant. The first firm to install this gets an implementation with more of our attention than any later one, and a commercial conversation that reflects that. If being first is a problem for you, the legal pack is the one that has been through production, and the platform page describes the shared foundation.
Thirty minutes, on a screen share, on a matter you pick. We connect nothing, install nothing and ask for no credentials to have that conversation. If it is not obviously useful in the first ten minutes, we will tell you so and you get your afternoon back.