Perpetual Resource Partners LLC used to spend 2 to 4 hours cleaning up each new title lead sheet before its land team could work it. Claude now converts and imports a new sheet in 15 to 30 minutes, including human review of ambiguous mappings and possible duplicates. In six weeks, an Elios Forward Deployed Engineer, a software engineer who works inside the client's own systems, took PRP's early-stage custom CRM to production on Microsoft Azure and connected it to Claude Cowork through a Model Context Protocol (MCP) server.
Title work lived in dozens of separate spreadsheets
PRP runs a relationship-led acquisition strategy. Its land team works directly with landowners and working interest owners, so knowing who owns what, and who has already been contacted, drives every acquisition conversation.
Before LandHub, that knowledge lived in dozens of title lead spreadsheets on a network drive, organized by section, plus a Live Deals Tracker in Excel. Those files held owners, tracts, the relationships between them, document references, and hand-written research notes.
The records existed, but connecting them took work. Many owners hold interests in more than one tract, so seeing everywhere a single owner showed up meant opening several files. The same owner also appeared under slightly different names across title runs, so the team wrote notes like "appears to be associated with..." to track those links by hand. Tracking them by hand risked contacting the same seller twice.
PRP had already started building its own CRM, called LandHub, to fix this. It had stalled short of production.
What Elios deployed
One Elios Forward Deployed Engineer finished LandHub, tightened its security, and deployed it in PRP's own Azure environment. PRP owns the code, the Azure subscription, and the GitHub repository. Each team member signs in with the Microsoft account they already have, through Microsoft Entra ID. Only people in two groups that PRP manages in Azure, one read/write and one read-only, can get in.
The application follows how PRP's land team already works:
- Owners, tracts, and holdings. Every owner card shows every related tract, with working interest, royalty interest, net revenue interest, and acreage. Tracts support both Texas and New Mexico survey systems.
- Deal pipeline. Deals move through the team's own stages, from Prospecting through Evaluating, Making Offer, Negotiating, Approving Offer, and Closing. LandHub won't update a tract's recorded owner until every closing step is complete.
- Activity logging. Calls, texts, and emails are logged against the owner, tract, or deal, with the next follow-up date attached.
- Seller motivation. Motivation is tracked per tract, so the team can see who may be ready to sell and where each conversation stands.
- Personal dashboard. Each person sees their follow-ups, active deals, an uncontacted-owner queue they can claim and assign, and the pipeline by stage.
During the build, Landon Schott, PRP's Data Manager, said:
"This is already exactly what we're asking for, if not more."
Claude does the import, and the team reviews it
PRP's title lead sheets often packed several facts into a single cell. LandHub accepts one structured format, with one fact per column.
A custom Claude skill, run in Claude Cowork by PRP's own team, handles PRP's title lead sheet imports in production. Claude breaks apart packed cells, converts each sheet into the LandHub format, and imports it through a LandHub MCP tool. LandHub validates every row and reports any errors. Claude shows the team member running the import every suspected duplicate owner, and that person confirms or rejects each one before LandHub saves anything. Re-importing the same sheet updates changed values instead of creating copies.
Claude Cowork works the CRM directly
The MCP server gives Claude Cowork a secure connector to LandHub, so Claude can read and update LandHub records. It runs in production alongside the application, with tools covering owners, contacts, tracts, holdings, deals, notes, activities, citations, and imports.
A land team member can ask Claude to bring them up to speed on an owner before a call, then ask it to log the call notes and set the follow-up for next Tuesday. Both land in the shared CRM.
Claude works with the same permissions as the person using it in Claude Cowork. That person signs in through PRP's Microsoft Entra ID tenant, and every tool call goes through the same business rules as the LandHub application. Before Claude deletes a record or changes a tract's recorded owner, it shows the user what will change, and nothing happens until that user confirms. LandHub's communication tools only log activity. They cannot send email, send texts, or place calls.
That shared access is what makes PRP's land team AI-native. Claude reaches the same records the team works in, with the same permissions each person already has.
Results: before and after
These are Perpetual Resource Partners' own figures.
- New title lead sheet cleanup: 2 to 4 hours per sheet before. 15 to 30 minutes now, including review.
- "What's our position in this section?" or "Have we contacted this owner?": 15 to 30 minutes of digging through files or asking around before. Under a minute now, by filtering to the section or opening the owner card.
- Duplicate owners across title runs: tracked by hand in research notes before. Flagged at import now, and confirmed by a person.
Landon described those name mismatches across title runs as "a real double-contact risk that LandHub's duplicate matching addresses."
"The level of collaboration and responsiveness throughout the project exceeded our expectations."
Landon Schott, Data Manager, Perpetual Resource Partners
A pattern for other records-heavy teams
Land and title work is one example of a common problem, where records are spread across spreadsheets and the relationships between them live in hand-written notes. At PRP, the approach was to put the records in one system the team owns, then connect that system to Claude, which works with the same permissions as the person using it.
If your team's work lives in spreadsheets that only make sense side by side, we can walk you through how we deployed LandHub for PRP and what it would take to build and connect a system like it that your team owns.
