AI for water damage restoration: from first notice of loss to a ready file.

This page is not about answering services or call bots. It is about the part of a water damage job that decides whether the invoice holds up: whether the file shows what was found, what was done, and that the structure was monitored every day until it was dry. Building and checking that file is where AI earns its place in a restoration business.

Grok Bot provides named Bots that work on a persistent cloud computer, follow written procedures (skills), run on schedules (routines) and prepare drafts that a person sends. It has no telephony: it does not answer calls, take voicemail or send text messages. First notice of loss therefore stays with the owner, the on-call technician or the answering service the company already uses, and the assistants work on what those people capture.

What follows is an Aevrion operating pattern for a water mitigation company: which documents the assistants prepare at each stage from loss to payment, which decisions never leave the owner, estimator or technician, and where a custom integration is needed because no documented connector exists for restoration software.

Based on the Grok Bot documentation at docs.x.ai as checked on 13 October 2026. Its statements on Bot rules, Bot-to-Bot messaging, routine triggers, connected apps, approval expiry and payments were rechecked against docs.x.ai and Cursor's help pages on 26 October 2026. The trade examples are Aevrion's illustrations, not documented Grok Bot use cases. Aevrion Ops is an independent consultancy. It is not affiliated with, endorsed by or a partner of xAI, SpaceXSI or Cursor. Grok Bot is named here only to describe it.

This page works from one premise: a mitigation invoice stands or falls on the file behind it, meaning loss details, photos, moisture map, daily readings, equipment log, and the estimate lines those records support. A missing day of readings, or equipment billed that the log does not show was placed, is a gap nobody can close once the structure is dry and the equipment has gone.

The industry standard frames the work the same way. The IICRC's page for ANSI/IICRC S500, the water damage restoration standard (5th edition, 2021, as checked on 13 October 2026), lists administrative procedures and project documentation, inspections and preliminary determinations, and psychrometry and drying technology among the components it covers. One IICRC-approved instructor, writing on the blog of Accuserve, a property-claims and repair company, says that when drying is disputed it usually comes down to how well it was documented.

Documentation is repetitive, rule-based and easy to check against a written list, which is why it suits an assistant, and it is the first thing to slip when the owner has been on a loss since two in the morning. Seen from the adjuster's side, the file is built in seven stages, and an assistant has a part in each without deciding any of them.

1. 01

   First notice of loss

   Decides

   Is the loss record complete enough for the owner to decide on dispatch?

   Output

   A loss record in the company's format, with every missing field listed as a question

2. 02

   Arrival and authorization

   Decides

   Is the signed work authorization on file, with who signed it and in what capacity?

   Output

   The authorization filed against the job, and a note for the owner when the signer is a tenant or property manager

3. 03

   Day one

   Decides

   Do the first photos, readings and moisture map cover every affected room?

   Output

   Day-one evidence indexed by room, with the technician's category and class determinations recorded as the technician's

4. 04

   Daily monitoring

   Decides

   Was every room and reading point visited yesterday, with the equipment the log says was running?

   Output

   A morning gap list for the technician, closed on the next visit

5. 05

   Drying goal met

   Decides

   Has the technician signed off that the goal was reached, and on what date?

   Output

   The sign-off, the final readings and each unit's pull date in one place

6. 06

   Estimate and submission

   Decides

   Does every estimate line trace to a log entry, a photo or a note?

   Output

   A mismatch list for the estimator, then a submission packet that the owner sends

7. 07

   Payment and close

   Decides

   What has the carrier paid, queried or left open?

   Output

   A receivables status line and drafted replies to each query, sent by a person

When first notice of loss arrives by phone, Grok Bot cannot take it. The owner, the on-call technician or the answering service handles the call and the owner's scripted safety guidance, and the owner decides whether a crew goes out. An assistant works on what the call produced: the answering-service email, the owner's typed notes, the customer's photos.

An intake Bot turns those notes into a loss record in a fixed format and lists what is missing as questions for the owner, not guesses. Getting the notes to it is a custom integration, or supervised browser use where the vendor permits it, subject to testing; the documentation names no connector for restoration or answering-service software.

Source repair is a separate job. The S500 page states that finding and correcting the source of the water is the property owner's responsibility, not the restorer's, though the owner may contract the restorer or another specialist to do it. A burst supply line is plumbing work, and the plumbing articles in this series hand dry-out to restoration at exactly that point.

Fields an intake Bot checks for in the loss record (an Aevrion checklist; adapt it to your own intake script):

• SOURCE AND STATUS

  What failed, whether water is still flowing, whether the supply has been shut off and by whom.

• TIME

  When the water was first noticed and when it probably started. Elapsed time matters later, when category is assessed.

• EXTENT

  Rooms and floors affected, whether ceilings are involved, whether water reached a lower level.

• CONTAMINATION FLAGS

  Any sewage, toilet overflow or water from outside, recorded as the caller reported it for the technician to assess on site.

• PEOPLE AND SAFETY

  Occupants, anyone vulnerable, pets, power status, and anything the caller said about sagging ceilings or wet electrics.

• INSURANCE AND AUTHORITY

  Carrier and policy or claim number if known, and whether the caller is the owner, a tenant or a property manager.

Two determinations shape a mitigation scope. Category describes how contaminated the water is, Categories 1 to 3 in the standard's terms. Class reflects the evaporation load: roughly, how much water the drying has to remove. A Cleanfax explainer on the S500 terms (published April 2016, checked on 13 October 2026) makes two points that matter for documentation: time and temperature can change the category, and the 2015 edition dropped fixed time references for when that happens. Class depends on the wet surface area and on how much water sits in porous or low-evaporation materials at the start of the job.

The EPA's guide to mold and moisture in the home (checked on 13 October 2026) adds the reason speed matters: it advises drying water-damaged areas and items within 24 to 48 hours to prevent mold growth.

None of that is an assistant's call. An assistant makes the timeline visible: when the loss was reported, when the crew arrived, when equipment went in, when each day's readings were taken. If day one recorded one category and the technician's notes on day three mention odor or discoloration, the documentation Bot raises it as a question for the technician. It does not reclassify the job.

Grok Bot's documentation recommends starting with the smallest useful roster and creating a separate Bot only when the work has its own goal, sources, approval boundary or schedule. In mitigation three such jobs are easy to tell apart. The roster below is Aevrion's pattern; the role names are illustrative, not documented Grok Bot roles.

Each Bot keeps its standing rules in memory, given to it in chat (never send to a customer or carrier, never edit a reading, report a gap rather than fill it), which the documentation treats as guidance, not a technical control. Shared procedures such as the loss-record format are written once as a skill: a reusable set of steps, decision rules and expected output.

On a busy day the three can share a group chat, which holds two to six Bots, for the morning queue. Anything needing the owner's yes still arrives in the owning Bot's own chat, since approvals and email drafts are never shown in a group.

The documentation Bot's main job is a routine, a written procedure run on a schedule, which every morning reads the previous day's drying records for each open job and lists what is missing or does not match. If the drying-software export it needs is absent, it reports that it could not check rather than repeating yesterday's list. The drying-log article in this series sets out the full check.

Carrier and third-party administrator programs are one route to mitigation work. Crawford & Company's Contractor Connection, for example, describes itself as a managed repair network, lists water mitigation and board-up among its residential services, and mentions a quality assurance program and a five-year workmanship warranty (as checked on 13 October 2026). On a program job, the program's own requirements govern the file, and the billing Bot works from the checklist the owner keeps for it in the business's files, not from memory, which the documentation says is not an authoritative source.

Texas adds a hard boundary. Under Texas Insurance Code §4102.163, as published on 13 October 2026, a contractor that works, or may work, on a property cannot adjust that property's claim or advertise that it will, and neither an adjuster license nor a power of attorney from the insured changes that. That is one state's rule; other states need their own check. The design consequence: the billing Bot drafts status updates and document submissions, and never argues coverage or negotiates the claim.

Drafts reach a carrier only when a person sends them. Grok Bot can show an editable email draft with its recipients and body, and the user chooses to send it or discard it.

At stage six the billing Bot sets the estimate's equipment, room and demolition lines against the logs and photos and hands the estimator a list of mismatches, each tied to its record, working from an exported file or a custom integration with the estimating software where the vendor's API or terms allow it, after testing. The estimator decides what each mismatch means, and every line and supplement stays the estimator's; the drying-log article in this series covers that comparison in detail.

Approval gates are configurable, not universal: the tool, the risk of the action and the owner's rules decide what pauses, and approving reverses nothing already done. A request raised by a routine at six in the morning lapses after about ten minutes (Cursor's help pages say thirty) if nobody answers, and its action never runs. That is why the morning routine only prepares.

The Bots on one account all use a single cloud computer, and with it the same files and saved browser logins, so an insurer or program portal signed in there is open to every one of them. Separate Bots are not a security boundary; a workload that needs its own credentials gets a separate user account, as the Grok Bot documentation recommends.

• ON SITE

  Safety stops for electrical hazards, structural sag, contaminated water or suspect older materials; category and class; demolition and antimicrobial use.

• DRYING

  Equipment placement and changes, the drying goal, and the sign-off that it has been met.

• MONEY

  Work authorization and payment terms, every estimate line and supplement, and any payment. Payment confirmations go through a human takeover of the computer, not through a Bot. Cursor's help pages add that a card saved in the Bot's Vault can be used at checkout without asking each time, so this design saves no card there.

• CARRIERS AND CUSTOMERS

  Every message to a carrier, an adjuster or a homeowner, and any promise about a timeline.

This page covers the water side: mitigation, drying and the documentation that pays for them. Mold remediation and fire and smoke restoration are related jobs with their own licensing, protocol and coordination questions; the fire and smoke restoration guide is linked below. When a water loss turns into either, the paperwork changes with it.

Upstream sits the source repair: the plumbing owner's overnight article in this series ends where this one begins, with dry-out handed to a restoration company.

Grok Bot's guidance is to begin with read-and-prepare work, review the results, and only then add approved actions or a routine. For a mitigation company the documentation check is the natural first step: it reads, it lists and it sends nothing.

Aevrion Ops can build this sequence with a restoration owner: the loss-record format, the documentation checklist, the routine and its no-data rule, the approval boundaries, and the custom integration that brings drying-software exports and answering-service notes to the Bots, where those vendors' APIs or terms allow it and after testing. To scope a restoration back office, contact us; the readiness assessment is a shorter first step.

1. The assistants' value is a complete, adjuster-ready file, not answering the phone; Grok Bot has no telephony.
2. First notice of loss stays with the owner, technician or answering service; an intake Bot structures the notes and flags missing fields.
3. Category, class, demolition and the drying goal stay with the technician; the assistant keeps the timeline.
4. A morning routine checks yesterday's drying logs for missing readings, unvisited rooms and equipment mismatches, and only prepares.
5. In Texas, §4102.163 of the Insurance Code stops a contractor adjusting the claim on a property it works on, so the billing Bot drafts and a person sends.
6. Restoration software and carrier portals need a custom integration or supervised browser use; portal sign-ins are shared by all of an account's Bots.