---
id: "lib-2026-10-01-where-ai-agents-are-showing-real-it-savings"
type: "library"
title: "Where AI agents are showing real IT savings"
lang: "EN"
org: "CIO.com (G. Gross)"
source_type: "Practitioner post"
date: "2026-10-01"
week: "2026-W41"
url: "https://nosov.ai/library/lib-2026-10-01-where-ai-agents-are-showing-real-it-savings/"
original_url: "https://www.cio.com/article/4228524/where-ai-agents-are-showing-real-it-savings.html"
tier: "B"
tier_name: "Useful, limited"
tier_reason: "CIO.com feature by Grant Gross collecting named CIOs' and vendors' self-reported savings from IT agents; useful practice and a worked hidden-cost example, but no method behind any figure."
summary_basis: "full-text"
original_minutes: "6"
tags: ["business transformation", "ai technology", "ai agents", "roi", "hr service delivery"]
---

# Where AI agents are showing real IT savings

Tier B (Useful, limited): CIO.com feature by Grant Gross collecting named CIOs' and vendors' self-reported savings from IT agents; useful practice and a worked hidden-cost example, but no method behind any figure.

Summary based on the full text.

## In short

A CIO.com feature (1 October 2026) on where AI agents cut IT costs. Three areas recur: tier 1 and 2 IT support, coding agents that let firms build tools instead of buying them, and cloud cost control. Named CIOs at LaunchDarkly and West Monroe and the CTO of KamiwazaAI report savings. A sceptical CTO from Glokal AI argues that savings are often inflated because pilots run on easy tickets and the cost of checking agent work rarely enters the business case.

## Key numbers

- 40%: reduction in yearly managed service provider costs reported by West Monroe after an IT and HR support agent
- 2,700 hours: estimated yearly operational time saved by the West Monroe support agent
- about $1 million: spending LaunchDarkly says it avoided on two optimisation projects by building tools with coding agents
- €60,000 to €36,000: worked example of paper savings shrinking once reopened tickets and agent checking are counted, about 60% of the pilot promise

## What it reports

- LaunchDarkly (CIO Rhonda Baldwin): about $1 million avoided on two optimisation projects, $120,000 avoided by building its own asset-management tool with coding agents, about $50,000 a year saved on tier 1 support.
- West Monroe (CIO Kevin Rooney): an IT and HR support agent in Teams or Slack handles password resets, access requests and tickets; 40% lower yearly managed service provider costs and an estimated 2,700 hours a year saved.
- KamiwazaAI (CTO Matthew Wallace): cloud-monitoring agents that track new systems and budget approvals cut cloud spend by 70% immediately.
- Glokal AI (CTO Jeet Pattanaik) worked example: 10,000 tier 1 tickets a month at about €15 each; the agent closes 40%, €60,000 a month on paper. If 15% reopen (€9,000) and checking costs about €15,000, the real reduction is about €36,000.
- His rule: agents fit work that is high volume, well documented and cheap to undo; anything that changes permissions or touches production is not ready.
- He warns that automating tier 1 removes the place where juniors learn the job.

## How they got it

Interviews with named practitioners and vendors. Figures are as reported by the interviewees; none is independently checked, and one cited study is a vendor-Forrester ROI study.

## Limits

- All savings are self-reported, with no baseline or method.
- Several sources sell AI agent products.
- The hidden-cost example is illustrative, not measured.

## Why it matters for companies

The worked example is the part to reuse: before approving a support-agent business case, add reopen rates and the human time spent checking agent actions, and expect real savings near 60% of the pilot figure. The selection rule (high volume, documented, cheap to undo) is a practical filter for HR service desks as much as IT. The junior-pipeline point deserves a plan: if tier 1 work disappears, decide where new staff will learn.

Original: https://www.cio.com/article/4228524/where-ai-agents-are-showing-real-it-savings.html