Underwriting

The AI System Schedule is the production fleet. The evidence pack is the file for the one registered system you open.

Gray boxes are waiting for screenshots. The caption is the name of the shot. The info buttons on C-Suite → Underwriting use the same words as this page.

The organization has to have Underwriting turned on

Open C-Suite → Underwriting. The page title is Underwriting, and the line under it names the organization. A Business, Scale, or Enterprise plan is required to load the schedule and the pack. When Underwriting is off for the organization, that page is not in the menu.

Two boxes on the same page
Box	What it is
AI System Schedule	Every production AI system the organization has registered, and whether the evidence on file for that system is complete.
Evidence pack	One file for the system you select. The latest completed assessment, and the download for that file.

1. How to read the AI System Schedule

The schedule is the fleet. Staging and development systems are not rows. A line under the tiles says how many non-production systems are left off, when any are.

[Screenshot needed: Underwriting schedule tiles]

How to read the tiles
Tile	Meaning
Production systems	The count of rows.
Complete	All five evidence items are on file.
Partial	Three or four of the five are on file.
Missing	Fewer than three are on file.

When the Complete tile is zero and there is at least one production system, the page says: No systems have complete evidence yet. The schedule will show the gaps.

An empty fleet says No production AI systems registered. Register a production system before the schedule has rows or downloads.

2. The five items

Each row shows the system name, Complete / Partial / Missing, how many of the five are met, and the latest Security run. None in 90 days means no completed Security run in that window. The date, when there is one, is Chicago time.

Evidence completeness
Item	Where to fill a gap
Registered, with a named owner	On an unmet row, Update opens the system in the registry.
Exposure facts filled in	The same Update link opens Exposure facts (used on the AI System Schedule).
Security assessment completed within 90 days	A completed Security assessment on that system, inside the last 90 days.
Escalation chain recorded	Update opens the escalation editor for that system. The chain needs a named person.
Instrument hashes recorded on the run	The Security run has to carry the prompt hash and the question-bank hash. Those hashes come from the run. They are not typed onto the schedule.

All five is Complete. Three or four is Partial. Fewer is Missing. Click a row to open that system in the evidence pack and scroll to it.

Completeness describes what evidence is on file. It is not an approval, rating, score, or insurability determination. Coverage decisions rest with the carrier.

3. Fill in exposure facts

The form heading is Exposure facts (used on the AI System Schedule). Save with Save exposure facts. A red line under the button is the save error. The fields are:

What it does — a short description of the system.
Audience — Public customers, Signed-in customers, Employees, or Other systems. The empty choice is Not recorded.
Data categories — Public information only, Contact details, Account data, Financial account numbers, Payment card data, Health data, Employee HR data, Credentials or secrets, or Other sensitive data. Public information only stands alone. Choosing it clears the other categories, and choosing another category clears Public information only.
Action rights — Answers only, Can create tickets or messages, Can hand off to a person, Can change customer records, Can move money, Can call external tools, or Can execute code. Answers only stands alone the same way.
Cannot do — optional. A limit you want on the sheet.
Monthly volume — Under 1,000, 1,000–10,000, 10,000–100,000, or Over 100,000 interactions / month.
Industry — free text.

[Screenshot needed: Exposure facts form]

4. How to use the evidence pack

Evidence pack. One file for the system you select. The latest completed assessment, and the download for that file. The menu is labeled Registered system. The count beside it reads registered · one system per pack. With no systems registered, the box says No registered AI systems.

[Screenshot needed: Evidence pack for one system]

Read the latest completed run. Current means the instruments on file still match that run. Stale means a recorded instrument changed after it. The reasons are listed under the badge. No completed assessment run means the downloads stay off, and the page says Export needs a completed assessment on this system.

Evidence completeness on the pack repeats the five items, each marked Met or Not met. The schedule lists production systems. A staging or development system can still be selected here. It will say Not on the schedule. Non-production systems are excluded.

Download the PDF to read, or the JSON for the same file. Open the next system when you want its file. Each pack is one system.

5. Download the schedule

Download schedule PDF is the roster plus one submission sheet per production system. Download schedule JSON is the same schedule. The roster columns are System, Exposure facts, Evidence, and Latest Security run.

Each submission sheet prints the exposure facts, the owner, the Security run, and the two instrument hashes when that run recorded them. Observation counts, human handoff, shut-off, cadence, and owner sign-off print Not recorded on purpose. The box labeled For carrier use (optional) is blank. A coverage answer is not stored here.

The line under the schedule, and the banner under the pack, both say: Assess-only. Supplemental cyber renewals. Not certification. Not VALIDATED. No Verified seal.