Back office ops · Production

McCarthy Holdings transforms dispersed construction knowledge into AI-powered advantage with Glean

The problem

With 100 million files spanning decades, critical information — policies, procedures, project records, material specifications, historical data — was dispersed across SharePoint, Microsoft Teams, Procore, and other systems. Employees lost time searching for answers that existed somewhere in the organization but weren't readily reachable.

First attempt

McCarthy evaluated Microsoft Copilot as part of a broader push toward AI-powered knowledge access, but found its search and retrieval experience opaque and lacking the LLM flexibility and user experience the team envisioned.

Workflow diagram · grounded in source
1
Employee information need
trigger
“McCarthy employees identified a growing challenge accessing information across many distributed tools”
2
Knowledge graph retrieval
ai_action
“Knowledge that once required digging across multiple systems now surfaces in seconds. In one case, a user located a file containing a 14-year-old 3D building model almost instantly — a search that previously demanded significant time fro…”
3
Purpose-built agent execution
ai_action
“With Glean's agent capabilities, McCarthy teams are now building and deploying purpose-built agents that automate complex, data-intensive workflows”
4
Human review of agent outputs
human_review
“has already helped its human counterparts review more than $2.7 million in change orders”
5
Governed agent publishing
validation
“a smaller, vetted group holds publishing rights for company-wide deployment — structured enough to be responsible, open enough to drive real innovation across the organization”
Reported outcome

McCarthy estimates a conservative two hours saved per employee per week company-wide, with corporate adoption reaching 90% and field and operations teams nearly 60%.
Their most-used agent logged more than 40,000 runs in its first two months, and an agent has helped human counterparts review more than $2.7 million in change orders.

Reported metrics
Time saved per employee per weektwo hours saved per employee per week
Corporate team adoption90%
Field and operations team adoptionalmost 60%
Most-used agent runs in first two monthsmore than 40,000
Show all 7 reported metrics
time saved per employee per weektwo hours saved per employee per week
corporate team adoption90%
field and operations team adoptionalmost 60%
most-used agent runs in first two monthsmore than 40,000
change orders reviewed by agentmore than $2.7 million
reliance on HR supportcutting reliance on HR support
onboarding timereducing onboarding time
Reported stack
GleanMicrosoft CopilotSharePointMicrosoft TeamsProcore
Source
https://www.glean.com/resources/customer-stories/mccarthy-holdings-inc
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

McCarthy estimates a conservative two hours saved per employee per week company-wide, with corporate adoption reaching 90% and field and operations teams nearly 60%.

What tools did this team use?

Glean, Microsoft Copilot, SharePoint, Microsoft Teams, Procore.

What results were reported?

Time saved per employee per week: two hours saved per employee per week; Corporate team adoption: 90%; Field and operations team adoption: almost 60%; Most-used agent runs in first two months: more than 40,000 (source-reported, not independently verified).

What failed first in this deployment?

McCarthy evaluated Microsoft Copilot as part of a broader push toward AI-powered knowledge access, but found its search and retrieval experience opaque and lacking the LLM flexibility and user experience the team envi…

How is this back office ops AI workflow structured?

Employee information need → Knowledge graph retrieval → Purpose-built agent execution → Human review of agent outputs → Governed agent publishing.

Related back office ops cases