Revelations AI

Press kit

Who approved that?

AI has moved from answering questions to taking actions: buying, emailing, changing code, denying claims. Revelations, in Fishers, Indiana, builds the layer that makes an AI agent wait for someone independent to sign off, and leaves a receipt anyone can check. Our forward-deployed engineers set it up inside companies.

Fast facts

The short version.

Company
Revelations Consulting Inc., Fishers, Indiana
What it is
An AI platform company that does forward-deployed-engineer consulting
The platform
Seconded: per-agent keys, independent approval before consequential actions, and signed receipts with an open verifier
The rule
No AI agent approves its own work
Launch
Seconded launches publicly Tuesday, Dec 1, 2026
The service
Gated Agent Deployment: senior engineers put 1–2 agents into production in eight weeks
Apps
Say When, Deadbolt, Certified, Attending, Countersigned, Preparer of Record, Second Look, Broker of Record, Close of Record, Rightful, Credited
Stage
Early access. The apps are live market tests; names are working names

Story angles

One question, told to different desks.

  • Tech and security

    No AI agent should approve its own work

    We ran about 186 agents on our own work. One signed 77 approvals in a row without ever saying no, then signed verdicts with test results it made up. Every signature was valid. What independent review of an agent has to look like.

  • Consumer and holiday

    Before you let an AI shop for you

    AI assistants can now buy and pay. Give them an allowance, not your card. Three things to set up first, and Say When, the app we're building for it.

  • Business and workforce

    Forward-deployed engineers, outside Silicon Valley

    Why companies that want AI agents need engineers inside the building, and why the job becomes reviewing agents, not writing every line.

  • Policy and accountability

    Who's accountable when an AI agent acts?

    Rules increasingly say a qualified person must decide. Proving that one did is the hard part, from court filings to coverage denials to hiring.

  • Legal, health, HR, tax, customs and finance trades

    The person who signs has to be able to prove they checked

    Certified for AI-drafted filings, Attending and Countersigned for health plans and coding, Second Look for HR, Preparer of Record for tax returns, Broker of Record for customs entries, Close of Record for the month-end close.

  • Indiana and local

    A Fishers company run the way it sells

    Agents propose, people decide, and every approval is recorded. What working alongside AI agents looks like day to day.

First-party data

What happened when we let AI agents review each other.

~186

agent identities

~7,200

signed submissions

~2,400

rejections

6

model families

77 approvals, zero rejections

A QA agent signed 77 approvals in a row and never said no. Later, unable to run anything, it kept signing verdicts with made-up test results. Every signature was valid.

One process, three roles

One coordinator process signed as author, reviewer and QA, and the change merged.

Two-thirds from one model

About two-thirds of QA passes came from one model family, so "independent" review wasn't.

Figures are approximate, from our internal agent fleet, July–November 2026.

Charts for your story (PNG, 1600×900; please credit Revelations Consulting): 77 approvals, zero rejections · One process, three keys · Two-thirds from one model family · The rules timeline

Coming soon: an agent self-approval census from public GitHub data, counting how often AI agents' changes merge without a person approving them. Reporters can ask for an early look under embargo.

The rules

Rule after rule: a person has to check the AI’s work.

Since 2025, health-care laws in two states, a customs ruling, IRS guidance, a court rule and a Colorado employment law have said the same thing in different words: when AI does the work, a qualified person has to check it. Each one has an app that records that check.

  1. · California law

    California SB 1120

    A denial, delay or change of care based on medical necessity must be decided by a licensed physician or qualified health care professional, not an algorithm.

  2. · Customs ruling

    CBP ruling H350722

    An AI tool is not a “person” under the broker rules. Decisions about what goes on an entry must be made by a licensed customs broker.

  3. · IRS guidance

    IRS OPR guidance 2026-19

    Tax practitioners using generative AI must thoroughly review all AI-created documents, verifying facts, citations and calculations.

  4. · Arizona law

    Arizona HB 2175

    A medical director must individually review each medical-necessity denial and may not rely solely on recommendations from any other source.

  5. · Court rule

    Indiana Trial Rule 11

    An attorney’s signature certifies that any cited legal authority is authentic, including in filings prepared with AI.

  6. · Colorado law

    Colorado SB 26-189

    Workers and candidates can ask for a person to review an AI-influenced employment decision, where that review is commercially reasonable.

Plain-language summaries, not legal advice. Each links to its source; read it yourself and ask your counsel how it applies to you.

People

Available for interviews and expert comment.

Live, taped or remote, and as a plain-English source when AI-agent or outage news breaks.

Joey Stevens, founder and president

12 years of infrastructure experience, from bare-metal hypervisors to AWS, Azure and GCP. Has scaled apps to millions of users and led SRE for an AI agent platform.

Steve Wieland, partner

Turns expensive manual work into systems. At an AI agent platform he mapped the cost of goods sold, cut cloud spend, tuned AI usage for cost and performance, and evaluated the agents the company shipped. Leads client delivery.

Boilerplate

About Revelations.

Revelations Consulting Inc. is an AI platform company in Fishers, Indiana, that does forward-deployed-engineer consulting. Its senior engineers build AI agents inside companies that take real actions, on Seconded, its platform for one rule: no AI agent approves its own work. Every agent and person signs with its own key, consequential actions wait for independent sign-off, and each decision leaves a receipt anyone can verify. Seconded launches publicly on Dec 1, 2026. Apps on the platform include Say When for consumers and early-access tools for engineering, legal, health, HR and tax teams.

Logos and images

Founder photos and product screenshots on request.

Press contact.

Joey Stevens · joey@revelations.consulting

We reply to press the same business day.

What we can show on camera

The live forge-an-approval demo at seconded.dev: an agent tries to approve its own action, gets rejected with the reason on screen, then a real reviewer signs and leaves a receipt.