The IT AI Stack
Build a governed AI operating layer for your IT team in 30 days.
Organization Edition: one legal entity, internal sharing, no seat cap: $349
“Battle-tested in production environments.”
The situation
Every vendor deck says your IT team should be "agentifying everything". Nobody hands you the other list: which of your activities should never run unattended, which need a named human gate, and which are fine as a script. So AI arrives as scattered experiments, ungoverned chats, and one enthusiast’s agent nobody else can maintain.
And when security asks what the agents can touch, or a manager asks what all this actually returned, the honest answer is usually a shrug. That shrug is what kills IT AI programmes, not the technology.
What this is
The IT AI Stack is a governed operating layer, not a prompt pack: 100 IT activities classified into five operating modes, ten agents built and tested on a real Microsoft 365 tenant, the eleven workflows that wrap them in human gates, and the 30-day sprint that rolls it out with a gate at the end of every week.
The testing ships with it. Every agent comes with its paste-ready instruction file, a smoke fixture, and a pre-calculated answer key, so "it works on my tenant" is something you prove in an afternoon and re-prove after every platform change. 67 counted test runs; the four defects found on the way are documented in the pack, not hidden.
What is inside
- The 100-row IT AI Blueprint: every major IT activity classified (40 agent-led with review, 29 AI-assisted, 16 automate, 13 never automate, 2 human only), with impact and effort per row
- Ten tenant-tested agents: spec sheets, paste-ready instruction fences with character counts, and the build checklist that flips the permissive platform defaults
- The fixtures folder: one smoke-test input and one answer key per agent, plus the two-turn pressure test the refusal gates survived
- Eleven gated workflows, fifty engineered prompts, the AI Implementation Decision Matrix, and a one-page governance pack with mechanisms instead of posters
- A 21-page Playbook PDF and a 10-tab workbook: the map with your own posture and owner columns, a pre-filled estate register, the full 67-run test ledger, and the value scorecard your IT manager reads at day 30
- The 30-day sprint: week 0 paper, weekly gates, and a close-out where cutting an agent that earned no use counts as governance working
Who it is for
- IT managers and team leads who want AI working inside real guardrails, with something to show an auditor
- M365 admins holding an "do something with Copilot" mandate and no method
- The person who will be asked, in three months, what the AI programme returned
Who it is not for
- Teams that want autonomous access decisions or autonomous production changes. This stack refuses both by design; the never-automate list is part of the product.
- Consultants delivering it to clients as paid training: that needs the Practitioner licence (email mathieu@kesslernity.com), not either edition here.
Not sure yet? Start free.
First 10 IT AI Opportunities is free and covers the same ground at a smaller scale. If it is useful, this is the full version.
Download First 10 IT AI OpportunitiesQuestions people actually ask
What is the difference between the two editions?
The files are identical across the two editions; you are choosing usage rights, not features. Individual ($99) licenses one named purchaser for their own work. Organization ($349) licenses one legal entity: share it internally, deploy across the IT function, no per-seat cap. If a colleague wants their own copy, that is either a second Individual purchase or one Organization upgrade that covers everyone.
What does "tenant-tested" actually mean?
Each agent was built in Microsoft 365 Copilot Agent Builder on a real tenant and run against fixtures with pre-calculated answer keys: 67 counted runs, all passing in final state, with every initial failure documented in the pack. It does not mean independently reproduced or validated by Microsoft, and the pack’s TESTING-SCOPE page says exactly what the testing establishes and what it does not.
What happens when Microsoft changes the platform under it?
That is why the validation pack exists. After any platform change notice, you re-run the smoke fixtures within the week, and a failed smoke freezes the affected workflow until a re-run passes. Nothing is future-proof; the stack ships the instrument that tells you when something broke.
Why do only 16 of 100 activities get automated?
Because unattended and irreversible are different variables. Detection, disabling and drafting can run unattended, because a mistake is a ticket. Deletion, destruction and declarations cannot, because a mistake is an incident. 69 of the 100 rows keep a person in the loop, and if a vendor tells you those numbers should be higher, ask who profits from that answer.
How is this different from the Deployment Kit?
The Deployment Kit is the rollout playbook for Copilot itself. The IT AI Stack builds what comes next: the agent layer on top, tested, gated and measured. They address different layers of the same programme.
Buying this with a company card?
Most buyers expense it. Your receipt email carries a Generate link that opens an invoice generator: add your billing address, put your legal company name and any reference your finance team needs in the Additional notes box, and download the PDF. If you have an EU VAT number, enter it at checkout and no VAT is charged. Here is the approval email, so you do not have to write one.
Subject: Approval request: The IT AI Stack ($99, one-time) Hi [Manager], I would like to expense a one-time $99 purchase: The IT AI Stack by Kesslernity. What it is: Build a governed AI operating layer for your IT team in 30 days. What we get: - The 100-row IT AI Blueprint: every major IT activity classified (40 agent-led with review, 29 AI-assisted, 16 automate, 13 never automate, 2 human only), with impact and effort per row - Ten tenant-tested agents: spec sheets, paste-ready instruction fences with character counts, and the build checklist that flips the permissive platform defaults - The fixtures folder: one smoke-test input and one answer key per agent, plus the two-turn pressure test the refusal gates survived - Eleven gated workflows, fifty engineered prompts, the AI Implementation Decision Matrix, and a one-page governance pack with mechanisms instead of posters Why it is worth it: it is a one-time purchase with future updates included, and it costs less than an hour of any consultant we would otherwise ask. The author publishes the open-source Copilot prompt library that Microsoft's CMO for AI at Work cited as "battle-tested in production environments". A downloadable invoice is available from the purchase receipt for expenses. [Your name]
The IT AI Stack
Build a governed AI operating layer for your IT team in 30 days.
Get it for $99Licensed, not sold. The full License & Terms apply. Kesslernity is an independent publisher: this is independent analysis, not affiliated with, sponsored by, or endorsed by Microsoft. It is practitioner guidance, not professional, legal, or financial advice.
Often bought alongside
The IT AI Quickstart Pack
A working slice of the IT AI Stack methodology, not a prompt dump.
$29→M365 Copilot Deployment Kit
The full rollout playbook plus paste-ready artifacts. Roll out without the rework, governed from the start.
$97→The Kill This Process Kit
Stop automating work that should not exist.
$249→