Skip to content
Close
SEARCH
Careers
Contact
Careers
Contact
MS-Template_cover01-1-3.png

 

Client story · IT services

Secure enterprise AI deployment: How CBTS governs Claude for 2,300 employees

Before asking a single client to trust AI with real work, CBTS ran the deployment across our own business. We’ve turned our experience into a blueprint for mid-market organizations to compete at scale in an AI-driven world.

image (9)

Industry

IT services

Scale

2,300 employees

Deployed

Anthropic Claude for Enterprise, under one governed control plane

Timeline

First 250 users in March 2026, firmwide within four months

Key outcomes

< 3 months

Payback on the full AI program investment

CFO-Verified

~$2M

Closed-won revenue

Salesforce-Anchored

0

Major security incidents attributable to AI

Security-Validated

2x faster

Incident containment with agentic SOC

Security-Validated

The CFO’s test

“Most AI ROI numbers don’t hold up once you ask how they were calculated. Ours do, because I don’t sign off on anything until I’ve checked it myself. Every ROI calculation ties back to actual results I can see in the P&L. This discipline ensures our AI spend is tied to actual returns.”

Shannon Mullen

Chief Financial Officer, CBTS

Right (12)

The situation

Fast is easy. Governed is hard. Companies need both.

Mid-market organizations face particular risk from AI pilots that scale activity without organizational readiness. That gap closes when governance comes before rollout.

With a cloud-native environment that requires complete protection of its infrastructure and data, CBTS took a rigorous approach to securing and governing AI at scale before rolling out Anthropic Claude for Enterprise to all 2,300 employees.

The approach

Governance first, then scale

We became our own first client. Before we deployed anything, we worked through four questions:

  • Data exposure. How do you connect AI to real systems without exposing sensitive data?
  • Proof of value. How do you prove ROI when most initiatives cannot quantify value for the CFO?
  • Fragmented readiness. How do you move fast when every team is figuring out AI on their own?
  • Workforce trust. How do you get a workforce to trust and adopt AI at scale?

Each answer had to hold up in production, not on paper. Our CEO-sponsored, company-wide Anthropic Claude for Enterprise rollout started with 250 users in March 2026 and within four months scaled firmwide, relying on the same discipline we sell to clients: resilience, incident response, and disaster recovery automation.

Over the course of those four months, six workstreams carried the rollout, each covering a different focus and each shipping with its own controls rather than inheriting them later:

We didn’t get everything right the first time. Early red-teaming caught access permissions far broader than they should have been, and the audit trail caught an agent pulling from a repository it was told to avoid. We fixed both before scaling.

The CISO’s view

“Governance was not the brake on this program. It was the reason we could hit the accelerator at all. Every agent runs inside controls I can see, and nothing reaches production without clearing review.”

Chris DeBrunner

Chief Information Security Officer, CBTS

The outcome

This is how the mid-market wins with AI

We used our own deployment to prove that AI could scale without compromising security, governance, or accountability. Three numbers carry that proof: what the tool chain stopped, how much of the estate the control plane covers, and what share of agents cleared review before going live.

Now we have a proven framework for giving clients an AI control plane that identifies, authenticates, and authorizes every agentic interaction in real time.

You don’t have to be a technology company to achieve similar results. In fact, our earliest and largest wins came from finance reporting, proposal response, and employee support. These are workflows that every mid-market organization runs. What’s more, it’s not just technologists contributing to Skill Forge. Our 1,300+ active builders also include sellers, finance analysts, and operations staff.

Put the CBTS blueprint to work

We have packaged our own experience into CBTS Forge AI: advisory services to identify the work worth doing, a governed control plane that shows you who is using what and at what cost, reusable skills and agents that run real workflows, and the team that operates them in production after go-live. Start with one piece or all four. Anthropic underpins our internal work, but Forge AI is model-agnostic by design. It deploys on whatever model and environment your requirements call for.

GettyImages-2133181950 2

Figures reflect CBTS internal systems of record as of the reporting period. Results described are specific to the CBTS environment and are not a guarantee of outcomes in any other organization.