Star Alliance Airline Soars with Alteryx Server
At a Glance
1000+
Workflows Governed
10
Departments Migrated
25
Alteryx Champions Trained
Overview
Service
Data Engineering & Infrastructure
Industry
Airlines
Stack

Machine Learning & Gen AI
Reconstructing Construction Progress with AI: A Claude Deployment for LCO
Author(s)
Technology Stack


The Challenge
LCO is a third-party construction consultancy delivering estimation, cost tracking, change order management, and progress monitoring for major industrial and infrastructure clients. As is typical in the industry, inputs arrive in inconsistent formats, contracts, invoices, Excel files, emails, handwritten notes, requiring arduous manual data entry and processing before project progress can be reconstructed into something ready for estimate review or cost analysis.
The Solution
Compass deployed Claude for Teams across LCO to support day-to-day productivity across the organization: drafting and summarizing emails, preparing for and following up on meetings, drafting documents and reports, and working through open questions on live projects, with MCP integrations connecting Claude directly to M365. For a construction consultancy, where the work is inseparable from document processing, that same productivity layer maps directly onto the core of the job. Claude is used to extract and normalize cost, quantity, and progress data directly from contracts, invoices, spreadsheets, and field notes, work that used to sit between the team and the estimate review or change order analysis they were actually there to do. Enablement went beyond an initial rollout: Compass advised on license sizing, established best practices, built out Skills and Projects tailored to LCO's workflows, and ran follow-up sessions that gave the team a forum to trade what was working and dig into specific use cases.
Impact
70%
reduction in manual time spent reconstructing project progress from source documents
100%
active adoption across the organization
2 months
from rollout to full-organization adoption
Stack

Our Client's Context
LCO operates across a portfolio of concurrent, deadline-driven engagements for major industrial and infrastructure clients, delivering the estimation, cost tracking, and progress reporting those projects run on. That input mix, contracts, invoices, Excel files, emails, handwritten notes, arrives with no shared format and no single source that reflects the current state of a project.
Before LCO can review an estimate, evaluate a change order, or judge whether a project is tracking to plan, someone first has to reconstruct that picture from whatever documents happen to exist. That reconstruction work is manual by nature, and across a busy portfolio, it consumes time better spent on the analysis clients are actually paying for.
Claude as the Extraction Layer
Across the organization, Claude sits inside the tools people already use for email, meetings, and drafting, handling the day-to-day writing and thinking that used to eat into billable time. That general layer is what made the document-heavy part of the job tractable: Claude ingests contracts, invoices, spreadsheets, and field notes, and extracts and normalizes the cost, quantity, and progress figures buried inside them, mapping whatever format a given project or engagement happens to use into LCO's own universal templates.
Because the same assistant is already part of how people work, that extraction step isn't a separate tool to learn, it's an extension of what they're already doing. MCP integrations extend that reach further, connecting Claude directly to M365, SharePoint, and LCO's internal construction management platform, so extraction and analysis happen where the team's data already lives rather than requiring anyone to move it in and out by hand.
Data Trapped in Every Format But One
No Structured Source of Truth
Every tracking input, a contract, an invoice, an Excel file, a written note, holds a piece of project progress, but none of them are structured the same way. That inconsistency is far more pronounced across projects than within any single one, so reconstructing a current, accurate view meant manually reading and reconciling formats that shifted from engagement to engagement.
Manual Work Ahead of the Real Work
Estimate review, cost tracking, and change order analysis all depend on that reconstructed picture being right. Meaningful time went into extracting and organizing data before anyone could start the judgment-based work at the center of the job.
Concurrent Engagements, Compounding Time Loss
With a portfolio of active projects running on tight deadlines, the manual document handling on any one engagement was small. Across all of them at once, it added up to a real drag on capacity.
Interested in learning more?
Fill out the form below and we'll get back to you as soon as possible.
Thank you for contacting us. We'll be in touch soon.


