Valantor is an AI company focused on visual intelligence for complex enterprise documents. Its GroundX platform reads tables, forms, diagrams, and layouts and returns structured data and source-linked answers, while Outcome Plug-ins such as FraudX and ExtractX apply the technology to specific business workflows. Pulse 2.0 interviewed Valantor CEO and Co-Founder Pat McCamley to learn more.
Pat McCamley’s Background
When asked about his background, McCamley shared:
I am currently the CEO and Co-founder of Valantor. I have spent most of my career as a successful operator, specializing in GTM and Corporate Development functions. From large companies such as Siemens to smaller venture-backed startups such as Arcadia.
While at Crius Energy, I served as a public officer driving strategic partnerships, M+A and Strategy for the firm, eventually selling the company to Vistra Energy, a large Independent Power Producer.
How Valantor Started
When asked how the idea for the company came together, McCamley explained:
While President of Arcadia, one of the largest utility data and bill payment companies in the world, we were working with AI before it became the buzz term it is today. I saw the pitfalls of leveraging AI in a productive and cost-effective way. Outcomes were what I wanted, not features.
I saw the opportunity of an outcome-driven model with clearly defined solutions and real ROI’s. In 2025, I left Arcadia to begin working on Valantor, where our journey as an OaaS (Outcome-as-Agentic-Solution) company began.
Favorite Memory
When asked about his favorite memory working for Valantor so far, McCamley said:
The day we closed the acquisition of EyeLevel. It was a year in the making and the foundational technology we are building Valantor on.
Core Products And Features
When asked about Valantor’s core products and features, McCamley detailed:
Valantor’s platform is GroundX. It reads the tables, forms, diagrams, and layouts inside complex enterprise documents. Give it a contract, a claim file, or a policy manual and it returns structured data in the fields you define, plus answers to questions that link back to the exact page they came from.
We designed GroundX with AI agents as a first-class user. GroundX Studio installs as a plugin of skills that gives an agent expert knowledge of the platform, and our MCP server lets it run live operations.
An AI agent can then stand up advanced document use cases like extraction, grounded search, and document classification without a team having to learn the platform first. Engineers can still build directly on our REST API and Python SDK, of course.
GroundX runs hosted or self-hosted, including on-premises and air-gapped sites, so you can keep document processing entirely inside your own environment.
On top of the platform we build Outcome Plug-ins, packaged products for a single job. FraudX reads a whole insurance claim file, runs more than 20 investigator-defined checks, and returns a fraud score, a chat interface over the claim, a source-linked evidence package, and a network view of connected parties.
ExtractX replaces legacy OCR and template workflows, delivered by our team or a partner.
Solving Complex Document Challenges
When asked about recent challenges in the sector and how Valantor has addressed them, McCamley explained:
The central challenge in our sector is that most enterprise data lives in complex visual documents that general-purpose AI cannot read reliably, and generic retrieval stacks lose accuracy on tables, forms, and diagrams, especially at scale.
Regulated buyers face a second problem: they cannot send sensitive documents to streaming model providers such as OpenAI and Anthropic.
We addressed both by building GroundX to read page structure before extracting, up to 99% accuracy, and superhuman performance on the public DocBench benchmark, combining document reading, storage, and search in one system, and shipping a self-hosted and air-gapped deployment path so regulated customers stay in control of their data.
This challenge is what led us to our first Fortune 500 client. They are very interested in the on-prem deployment due to the sensitive nature of their client data. The early results have been very encouraging and have led to multiple use cases identified across the organization in the first few months of working with them.
How FraudX Works
When asked about FraudX and how the product works, McCamley said:
Insurance investigators have a fundamental problem: too much evidence and not enough time. A complex claim can contain thousands of pages, medical records, depositions, bills, photographs, reports. Somewhere in that mountain of evidence may be the inconsistency that changes the entire investigation.
FraudX reads all of it. It uses GroundX to understand the complete claim file. FraudX Score surfaces the signals that deserve attention. Chat with Claims lets investigators interrogate thousands of pages and get answers tied directly to the evidence.
And then there’s FraudX Network. For the first time, carriers can see patterns and relationships across a massive pool of claims, not just their own. Recurring providers, attorneys, claims, and behaviors- connections that are almost impossible to see when every carrier and every claim is investigated in isolation. We’ll be publicly launching the FraudX Network later this year.
Importantly, human investigators are always in control. FraudX gives them the ability to review every page, follow every signal, and see the bigger picture, all at machine speed.
In our internal benchmarks, that can make complex claim-file review up to 40x faster, which will create substantial ROI for our clients.
In addition, the financial impact can be enormous. Identifying an issue in a single high-value claim before a payment is made can potentially save millions of dollars.
Key Company Milestones
When asked about some of Valantor’s most significant milestones, McCamley highlighted:
The founding team spun out of IBM in 2019 specifically to solve the documents-for-AI problem, drawing on roots in IBM Research and IBM Watson and consumer-scale AI at The Weather Company, and built GroundX from first principles before the current AI boom.
GroundX achieved superhuman performance on the DocBench comprehension benchmark.
GroundX On-Prem is generally available, including air-gapped deployment, and Red Hat shipped an official OpenShift AI quickstart for it.
We put the platform into production with enterprises such as one of the world’s largest human capital management and payroll companies, serving more than a million businesses worldwide, as well as other customers such as AskVet, and shipped FraudX and ExtractX as our first Outcome Plug-ins.
Customer Success Stories
When asked about specific customer success stories, McCamley shared:
Air France/KLM built a customer-service assistant on thousands of policy documents, fare tables, baggage rules, refund decision trees, and hit 96.2% accuracy against a 60% target, clearing human-level accuracy on questions that take new agents nine months of training.
Their Knowledge Management Manager, Karin Oskam, said: “The Valantor platform delivered truly impressive results, and the bot’s ability to improve over such a short period was amazing.”
AskVet built its virtual veterinarian on GroundX; founder and CEO Cal Lai said they “couldn’t have done it without the underlying technology.”
On FraudX, we have customers on the record including Willis Law Group, Andromeda Advantage, and Tradesman.
Funding And Revenue
When asked about funding and revenue metrics, McCamley shared:
Valantor has raised capital from both institutional and family offices.
Over the last two quarters, we have more than tripled revenue with continued acceleration for the balance of 2026 and into 2027.
Total Addressable Market
When discussing the total addressable market Valantor is pursuing, McCamley explained:
Valantor is focused on a $12.4B global intelligent-document-processing (IDP) market by 2030, growing at ~33% CAGR (per Grand View Research, up from $2.3B in 2024). The strategy is to target verticals where documents are complex, high-volume, and heavily regulated, where the difficulty of the problem is itself the moat.
Five target verticals, each with an estimated 2030 serviceable market:
- Insurance ($2.0B): product FraudX, already GA. Claim files run thousands of pages and fraud investigation (SIU) headcount can’t scale against $308B in annual US fraud.
- Financial services ($2.8B, the largest slice): product ExtractX, GA. Covers KYC, AML, regulatory reporting, and lending ops, with large institutions on multi-year buying horizons.
- Healthcare ($2.0B): a stated vertical fit. Medical records, claim forms, and lab results under HIPAA, with the “AskVet” story cited as proof.
- Public sector ($1.8B): vertical fit driven by document-heavy, air-gapped environments where on-premises deployability is the deciding factor.
- Energy / utilities ($1.7B): product GridX, still on the roadmap. Asset-heavy, technical-diagram-intensive workflows plus regulated reporting.
Competitive Differentiation
When asked what differentiates Valantor from its competition, McCamley identified four areas:
Four things. First, customer-controlled deployment: on-premises, air-gapped, or managed cloud, which most alternatives cannot match.
Second, high accuracy on complex documents, because GroundX reads page structure before extracting, up to 99% accuracy, superhuman on DocBench.
Third, one integrated system for document reading, storage, and search, so customers run fewer tools.
Fourth, the ability to drive 25 to 50% cost reductions with faster implementation through the GroundX Studio, which gives AI coding agents like Claude Code portable GroundX knowledge and workflows.
We are also model-agnostic, and we integrate document understanding, reasoning, and execution into one Visual Intelligence layer rather than selling a single point tool.
Future Goals
When discussing Valantor’s future goals, McCamley said:
Extend the Outcome Plug-in pattern across more regulated, document-heavy verticals, bring GroundX Studio to business users, and grow the outcomes-as-a-service model, where a customer buys a governed business outcome with AI plus human accountability rather than assembling and operating the toolchain themselves.
Strategically, the goal is to establish Visual Intelligence as the control plane enterprises standardize on for their unstructured visual data.
AI Plus Human Accountability
When invited to discuss another topic, McCamley concluded:
One point worth adding is the “AI plus humans” accountability model: enterprises increasingly buy accountable outcomes, not raw AI capability, and our approach keeps human experts in the loop and accountable for the work.
