MS HouseSample audit

Our AI Agents

MS House is run by 7 AI agents. Each feature of the service has its own agent, and each agent uses the AI model that fits its job. ms.house Agents operate the service and are responsible for it.

How do the agents work together?

The chief agent makes the decisions. The content and email agents write first drafts. The chief agent reviews every draft against our written rules for accuracy, plain English and sources before anything is published or sent. The audit agent judges each AI answer only against facts quoted from the company’s SEC filing.

The agents also learn. Each change to this site and each email style is scored later on measured results, such as more citations in AI answers or more replies. The next cycle favors what worked and still tests new options.

How do MS House agents check AI answers against SEC filings?

The audit agent checks each AI answer against the company’s own SEC filings and official investor relations site. Sources include 8-K Item 2.02 exhibits, 10-Q filings and 10-K filings. News articles never serve as the source of truth.

The audit agent captures the answer’s date and exact wording. The report places the filing line beside the answer, so readers can trace each verdict to the company’s disclosure.

The methodology explains the review steps. The sample audit shows how the evidence appears in a report.

Which agent runs each feature?

The live table below lists every agent, its schedule, when it last ran, and whether it hit an error.

Where does MS House publish data on AI answer errors?

The Investor Answer Index is MS House’s public research asset for AI answer errors. The Index covers Fortune 500 companies each earnings season and gives investor relations teams one place to review published findings.

The Investor Answer Index has no published results yet. MS House therefore gives no error rates, counts or trends on this page. When results are available, the Index will state how and when MS House collected the data and will give the sample size. This lets CFOs and boards review the evidence behind each finding.

Live status

7 AI agents run MS House

Each feature has its own agent. This table shows when each one last ran, whether it hit an error, and what it did. It refreshes every minute.

Agent and featureScheduleLast runStatusWhat it did
Chief agent
GEO and SEO optimization. Runs the 6 hour cycle: reads AI answer and search data, plans changes, scores past changes and learns which actions work.		Every 6 hoursLoadingChecking
Content agent
Website content. Drafts new blog posts and targeted page edits. The chief agent reviews every draft before it is published.		Inside the 6 hour cycleLoadingChecking
GEO scoring agent
Answer engine scoring. Scores every page against the questions buyers ask AI engines, section by section, before and after each edit.		Inside the 6 hour cycleLoadingChecking
Audit agent
Investor Answer Audit. Checks what AI engines tell investors about a customer against its SEC filings, and writes the fix list.		Within 48 hours of each customer's earnings releaseLoadingChecking
Research agent
Contacts. Finds investor relations decision makers and the published IR inbox for Fortune 500 companies.		Weekdays, before each outreachLoadingChecking
Email agent
Outreach email. Writes the first email and at most 2 follow-ups, each built on one real finding from the latest earnings release.		Weekdays at the start of the US business dayLoadingChecking
Inbox agent
Contact form and replies. Answers the contact form and email replies within minutes, sends payment links, and hands legal, contract and pricing questions to a person.		On every messageLoadingChecking

Last checked just now. Errors count failed model calls and failed runs in the last 7 days. Prospect and client names are never shown here.

Frequently asked questions

Does a person check the agents' work?

The chief agent checks every draft against our written rules before it ships. A person on our team reviews the agents' log of changes and handles contracts, legal terms and custom pricing.

Do all agents use the same AI model?

No. Each agent uses the model that fits its job. Drafting needs speed, so drafts come from a fast model. Judging an answer against an SEC filing needs the strongest reasoning, so every verdict and every final review uses our strongest model, with a backup model if it is unavailable.

Does any agent see client data it should not?

No. Each agent gets only the data its job needs. Client and prospect names never appear on this page, and audit links are private.