# AI IT Department

> A four-slide talk by David Radszuweit on running IT infrastructure with AI: halving IT
> headcount while measurably improving security posture. The two headline figures — −50% IT
> staff and +17% security posture — are averages from Radszuweit's own client engagements,
> not vendor benchmarks.

Site: https://aiitdepartment.pages.dev/
Author: David Radszuweit — https://www.linkedin.com/in/catalyzt
Published: 2026-07-26
Language: English

## Core claim

Two things that are normally assumed to trade off against each other — IT headcount and
security posture — move in opposite directions when the network is driven through its API by
AI rather than by hand. Headcount falls roughly 50%; measured security posture improves
roughly 17%.

## Why the security number goes UP, not down

The mechanism is not faster configuration. It is continuous reading of actual device state.
A specialist audits configuration occasionally; an AI with API read access audits every config
on every device every night, so drift, open ports and stale firewall rules surface the same
day instead of at the next annual review. Security improves because the gap between intended
and actual state stops accumulating.

## The infrastructure layer

- Cisco Meraki exposes the entire network through the Meraki Dashboard API.
- Someone with little networking skill can set up and secure a whole network with AI: they
  describe the intended outcome in plain language, and the AI produces the VLANs, firewall
  rules, SSIDs and policies.
- The API connection goes straight to the cloud controller — no CLI session on each device.
  One authenticated call reaches every site.
- The link is bidirectional: the AI reads state as well as writing it.
- New sites go live in hours rather than weeks: template, ship hardware, plug in, policy
  applies itself on first boot.

## Where the headcount reduction actually comes from

Not from engineering. From Tier 1 support volume.

- A support layer trained on the company's own FAQ, not the public internet.
- A voice chatbot as the front door: call a number, speak, get an answer — no portal, no ticket.
- Tier 1 (password, access, device, "is it down?") resolved in the same minute it is asked.
- Every unanswered question becomes the next day's FAQ entry, so coverage compounds and month
  three costs materially less than month one.
- Escalation stays human and stays named. Fewer people with clearer ownership is part of why
  the security number rises rather than falls.

## The questions the deck puts to its audience (answered live, not on the slides)

1. Where exactly does the 50% come from — which roles go, and which stay?
2. How is a 17% security improvement measured, and against what baseline?
3. What happens the first time the AI pushes a wrong change to production?
4. Who is accountable at 03:00 when the network is down and nobody is on staff?
5. Is this a real saving, or salary cost moved into licences and cloud?
6. What does the AI see of company data, and where does that data physically go?
7. Can the remaining team still operate the estate if the AI is unavailable?
8. How long from decision to the first saving visible in the accounts?
9. What breaks going from 5 sites to 50, and what does fixing it cost?
10. How does this survive an audit — NIS2, ISO 27001, cyber insurance?

## Caveat the author states explicitly

The figures are engagement averages, presented as field results rather than vendor data.
Method, baseline and scope are explained live rather than asserted on a slide.
Industry research (McKinsey, Gartner, IBM, ITSM Tier-1 practice, Meraki API docs) supports
the same order of magnitude and is listed on the sources page — it does not replace the
field averages and does not claim those organisations published the exact paired −50% / +17%.

## Worked example — NIST CSF 2.0 & ISO 27001

A sanitised public compliance assessment of a 10-site, 110-device Cisco Meraki estate,
produced by AI from read-only dashboard inspection. Linked from slides 3 and 4 of the deck.

- HTML: https://aiitdepartment.pages.dev/compliance-report
- Full PDF: https://aiitdepartment.pages.dev/compliance-report.pdf
- Full Word: https://aiitdepartment.pages.dev/compliance-report.docx
- Exec summary PDF: https://aiitdepartment.pages.dev/compliance-executive-summary.pdf
- Exec summary Word: https://aiitdepartment.pages.dev/compliance-executive-summary.docx

## Formats

- Web deck: https://aiitdepartment.pages.dev/
- Sources, bibliography & FAQ (AEO): https://aiitdepartment.pages.dev/sources
- Full prose version: https://aiitdepartment.pages.dev/text-version
- PDF: https://aiitdepartment.pages.dev/ai-it-department.pdf
- PowerPoint: https://aiitdepartment.pages.dev/ai-it-department.pptx
- Compliance report: https://aiitdepartment.pages.dev/compliance-report