IT Support Manager (AI) — Technical Assessment

Time limit: ~60 minutes.

---

Context

You've just taken over IT at Twin Hearth Studios, a 40-person indie game studio. The SaaS portfolio is a mess: it lives across spreadsheets and emails, nobody knows who pays for what, there are unused seats, duplicate tools, and people who left months ago still have access.

Your first job is to turn that chaos into something clean and decision-ready.

You are given the studio's data as it actually exists, in /data/:

• saas-portfolio.csv — every SaaS subscription: app, owner, seats (total/used), monthly cost, renewal date, SSO status. It's dirty: duplicates, blanks, inconsistent names.
• employees.csv — the people: who's active and who has departed (with departure dates).
• app-access.csv — who currently has access to which app.

---

The task

Produce one decision-ready document (Markdown or PDF, your call) that delivers:

1. A clean source of truth for the SaaS portfolio — deduped, normalized, with total monthly/annual spend.
2. What to cut or renew — unused seats, departed employees still holding seats, duplicate tools, over-provisioned licenses — with the estimated savings.
3. Access & security risks — apps without SSO, and accounts belonging to people who already left.
4. A reusable onboarding/offboarding checklist — per system, so the next hire or departure is handled the same way every time.

Use AI however you like to get there — to parse the mess, cross-reference the data, find the savings, draft the checklist. We care about the result and your judgment, not that you typed it by hand.

In 2–3 lines, tell us which AI tool(s) you used and how.

---

What we're looking for

Area	Weight
Accuracy — did you find the real savings, orphaned access, and risks in the data	35%
Judgment — what you'd cut/keep and why; sane prioritization	25%
The onboarding/offboarding checklist (clear, complete, reusable)	20%
Effective use of AI to do the work	10%
Clarity of the final document	10%

We are not grading code or polish. A clean spreadsheet + a tight memo beats a fancy app with wrong numbers.

---

How to submit

1. Fork this repo.
2. Create branch submission/<your-name>.
3. Commit your final document (and any script/workflow you used) to /output/.
4. Open a Pull Request to main.
5. In the PR include: time spent and how you used AI.