Xactimate AI: what XactAI does, what it can't, and where our AI reads the scope
The name covers two different things: Verisk's own AI add-on, and third-party tools that read claim documents. This page separates them, names the limits, and shows the one job our AI does.
"Xactimate AI" usually means one of two things: XactAI, Verisk's AI add-on built into Xactimate, XactAnalysis and XactContents, and third-party AI tools that read claim documents. XactAI finds line items, labels photos, and scans floor plans; it does not read a carrier PDF into an ESX. ESXPress does that part.
By Chaz Jaco, Co-Founder & Software Architect, ESXPress - Last updated:
Does Xactimate have AI? What XactAI actually is
Yes. Verisk ships XactAI, an AI add-on that runs inside the Xactware products you already license. It bundles an AI Estimator (in XactRestore and XactRemodel), a Line Item Advisor and Sketch Scan (in Xactimate), automated photo labeling and descriptions, and XactRebuild. AutoPricing and XactRebuild are included in Xactimate Pro.
The confusion on this query is product identity. XactAI is a suite that spans several Verisk products, and the feature that sounds most like "AI writes the estimate" runs somewhere other than the carrier-facing Xactimate app. The definitional table below is the fastest way to keep them straight.
| XactAI feature | Runs in |
|---|---|
| AI Estimator | XactRestore, XactRemodel |
| Line Item Advisor | Xactimate |
| Sketch Scan | Xactimate (mobile) |
| Automated photo labeling and descriptions | XactAI |
| Assignment summaries | XactAnalysis |
| ALE receipt processing | XactContents |
| Transcription summaries | XactAI |
| Insight Analyst | XactInsights |
| XactRebuild | Xactimate Professional |
| XactAI AutoPricing | XactContents |
The XactAI feature list
Each XactAI feature lives in a specific Xactware product, not in "Xactimate" as one blob. That mapping is where most of the confusion on this query comes from.
Which XactAI features run inside Xactimate versus XactRestore
This is the disambiguation that coverage of the Claude announcement most often misses. The AI Estimator, the feature that drafts an estimate, runs in XactRestore and XactRemodel, not in the carrier-facing Xactimate app. Inside Xactimate you get the Line Item Advisor and Sketch Scan, which assist an estimate a human is already building. None of these read a carrier scope PDF into an ESX file.
What XactAI costs
XactAI was introduced as an early-access add-on with feature bundles tiered across the Xactware products; AutoPricing and XactRebuild ship included in Xactimate Pro. We do not print Verisk's per-user numbers here because they vary by source and change often. Verify current pricing on Verisk's own product pages.
What XactAI does not do
XactAI is an assistant inside an estimate, not a converter. Verisk's own Line Item Advisor documentation states that the advisor returns up to three line-item options per prompt and cannot provide "all items needed for a described job," pricing breakdowns, or brand comparisons. In plain terms, it helps you fill in a scope you are already working; it does not build one from a carrier document.
XactAI does not turn a carrier PDF into an ESX
This is the wedge, stated plainly. Nothing in XactAI reads a carrier's scope PDF and produces an import-ready ESX file. The reading job of turning a PDF, or a scanned or image-only scope, into Xactimate line items is a different job, and it is the one ESXPress does.
XactAI does not complete the scope for you
Verisk's Line Item Advisor help documentation lists what the advisor cannot supply: item activity information such as pricing breakdowns and detailed material or labor costs, all items needed for a described job, brand-specific items or brand comparisons, and pricing determined by other products or companies. Read that list carefully and it describes exactly the boundary between assisting an estimate and completing one.
Claude plus Xactimate: what the MCP connector really is
In May 2026 Verisk connected XactAI to Anthropic's Claude through a Model Context Protocol (MCP) connector, but the connector is for XactRestore, not Xactimate. Independent walkthroughs published after the announcement described the output as roughly 80 to 90 percent of a usable estimate that still needs review. The headline that "AI writes Xactimate estimates" overstates what shipped.
It is XactRestore, not Xactimate
The single distinction most coverage missed: the Claude connector attaches to XactRestore, the restoration estimating product, not to the carrier-facing Xactimate app where the claim estimate is reconciled. So the connector helps restoration teams draft, while the estimate that reaches the adjuster is still built and reviewed in Xactimate.
What independent reviewers found
Trade coverage (C&R Magazine and WrightWay, mid-2026) walked through the connector and settled on the same honest read: a strong drafting assist, roughly 80 to 90 percent of a usable estimate, with human review still required. We attribute that range to those reviewers rather than presenting it as our own number, because it is their observation about Verisk's tool.
Three reading jobs AI does well in a Xactimate workflow
AI earns its keep on the reading, not the deciding. The three reading jobs are: turning a scanned or image-only scope into text (vision), pulling line items, quantities, unit prices and codes into a structured list (extraction), and finding rooms in a drawing (sketch). All three output a list you can hold against the source page.
Reading unscanned scopes
Carrier scopes arrive as clean PDFs, scans, faxes, and photos of printouts. A vision pass turns those pixels into text, so an image-only file with no text layer becomes data you can work with. This is the largest part of the manual rekeying job that AI actually removes, and it is closest to what Photo-Scope does for photographs of damage.
Structured line-item extraction
Once the text is readable, the next job is pulling out what belongs in an estimate: line descriptions, quantities, unit prices, and a product code when the scope carries one. The output is a structured list for an estimator to review, not a finished estimate. Extraction is a parsing job first and an AI job second, which is why it stays reliable across messy inputs.
Room detection from drawings
Floor plans inside a scope are read as drawings, not pictures. AI finds the walls and the rooms and exports them as Xactimate-ready .skx sketch files, one per floor page, or as a zip. The first export builds the rooms and takes a few minutes; later exports of the same file are quick.
Where AI must never decide: prices and codes
Two things AI must never invent in a claim estimate: a price and a product code. An estimate is evidence. At ESXPress the PDF price is the source of truth, every line carries the unit price from the customer's own scope, and codes are resolved by lookup against a fixed dictionary, never generated. Unmatched lines keep the PDF price and are flagged.
This is also the honest answer to why a file reprices at all. Xactimate applies its own list on import, and that is the design. Read why Xactimate reprices on import for the mechanism, or run the reprice checker before you import.
How ESXPress uses AI, honestly
Our vision pass is DeepSeek-based. Vision reads the document, parsing extracts the line items, and a resolution engine matches product codes against a code reference measured from processed claim scopes. DeepSeek runs both our text and vision tiers, and vision is used for reading scans and rescue, never as the sole pricing judge. Prices stay on the PDF.
Vision reads, parsing extracts, lookup resolves
The pipeline has three stages and one rule: no stage invents a value. Vision turns a page into text, parsing pulls the line items, and resolution matches codes against a fixed reference. We name the provider, DeepSeek, because a workflow that reads money should say what is reading it. The same PDF to Xactimate workflow carries a scope end to end.
The measured record
Measured across 289 real claim files and 7,525 line items, 36.6% of line items carry a code we recognize; the rest keep their PDF price. The median file moves 17% of its lines, and 39 files move none. We publish these because a total must be explainable line by line. The code reference behind the matching is public and free, so an adjuster can verify a mapping before trusting it at Xactimate Code Search and supplement hit rate.
No model is the money judge
No model decides a price or a code. AI output always requires professional review, and our AI Disclosure says the same thing this page does: the reading is assisted, the deciding is human. That posture is what keeps an ESXPress export consistent with Trust and with the way an estimate is actually defended.
What to expect from AI estimation tools
Reasonable to expect: an image-only scope read correctly; lines extracted to a consistent structure; rooms from a drawing; an unknown code flagged, not guessed; a total that sums your document's own prices. Not reasonable: AI that prices your work instead of your price list; a guessed code; a file that needs no review.
Run the ESX validator on the result before delivery, and keep your review in the loop. AI did the reading; you run the estimate.
AI crawlers are welcome here
ESXPress publishes for humans and for AI systems. Our public pages are open to AI crawlers, and this page mirrors its FAQ answers in structured data, so search engines and AI assistants read the same facts that are on the page.
If you are building a search tool or an assistant that cites ESXPress, you are welcome to index these pages. The numbers here are measured, not invented, and the underlying method is documented on About the data.
Related
- Convert a scope PDF to ESX
- The full PDF-to-Xactimate workflow
- What an ESX file is
- Why Xactimate reprices on import
- Xactimate Code Search
- Which codes appear most (547 scopes)
- Supplement hit rate (547 scopes)
- ESX validator
- Reprice checker
- About the data
- Trust center
- AI Disclosure
- Photo-Scope AI damage analysis
- Integrations (API, webhooks)
- ESXPress vs CapOut
- ESXPress vs ClaimsFlow
- How PDF-to-ESX conversion works
- Xactimate glossary
Frequently asked questions
Does Xactimate have AI?
Yes. Verisk ships XactAI, an AI add-on across Xactimate, XactAnalysis and XactContents, with the AI Estimator in XactRestore and XactRemodel, and AutoPricing and XactRebuild included in Xactimate Pro.
What is XactAI?
Verisk's AI suite: an AI Estimator in XactRestore and XactRemodel, a Line Item Advisor and Sketch Scan in Xactimate, automated photo labeling, XactRebuild, and Insight Analyst in XactInsights.
Can Xactimate AI write my estimate?
No. The AI Estimator runs in XactRestore and XactRemodel, and Verisk's Line Item Advisor documentation says it returns up to three options and cannot supply all items needed for a described job.
Does the Claude/Xactimate connector write Xactimate estimates?
It connects to XactRestore, not Xactimate. Independent reviews in mid-2026 put its output at roughly 80 to 90 percent of a usable estimate, still needing review.
Can an AI tool generate Xactimate prices?
Not accurately, and we do not try. Our export carries the price from the customer's own scope PDF, and AI never writes a price it did not read.
What AI does ESXPress use?
A DeepSeek-based vision pass plus parsing and code lookup. Vision reads scans and rescue cases; lookup resolves codes against a fixed reference, and no model is the sole pricing judge.
Can ESXPress read scanned or image-only scope PDFs?
Yes. A vision pass reads scans, faxes and photos of printouts with no text layer before parsing extracts the line items.
What happens when a product code is not recognized?
The line keeps the PDF price and is flagged unresolved. The total never silently changes, and no code is guessed to fill the gap.
Can it turn floor plans into Xactimate sketches?
Yes. Room detection exports Xactimate-ready .skx sketch files, one per floor page or a zip.
How many line items actually reprice?
Across 289 files and 7,525 line items: 36.6% carry a recognized code, the median file moves 17% of its lines, and 39 files move none.
Do AI crawlers have permission to index this site?
Yes. AI crawlers are welcome on ESXPress public pages, and this page's FAQ schema mirrors the same facts that are visible on the page.