Fast answer
The usual suspects, in order of savings: licenses assigned to departed employees, everyone on premium licenses when most staff need mid-tier, Azure VMs sized by guess and never right-sized, orphaned disks and IPs from deleted projects, resources running 24/7 that are used 45 hours a week, and paying month-to-month prices for workloads that qualify for 1-3 year reserved pricing.
Microsoft 365: the license audit that pays for itself
- Departed employees with active licenses - offboarding removes account access but often not the paid license. In a 40-person company, 5-8 zombie licenses is typical.
- One-size-fits-all premium licensing: frontline and light-usage staff on E3/Business Premium when F3 or Business Basic covers their actual usage at a fraction of the price.
- Duplicate tooling: paying for Zoom, Dropbox, and Slack alongside the Teams, OneDrive, and SharePoint you already own inside Microsoft 365.
- Add-on stacking: Power BI Pro, Visio, and Project licenses assigned broadly but used by three people.
The audit is mechanical: usage reports by user and app, mapped against license assignments. An honest review typically takes a week and recurs quarterly - it's built into our managed IT plans.
Azure: sized by guess, billed by the hour
Most SMB Azure VMs were sized during migration by someone being safe - and safe means oversized. A VM at 15% average CPU is paying for four times the compute it uses. Right-sizing is measurable and reversible: monitor for 30 days, resize, verify performance, repeat.
The other Azure classic: orphaned resources. Deleted a VM? Its disks, snapshots, and public IP addresses often live on, billing quietly forever. Every subscription we've ever audited had orphans - the record so far is a test environment from three years prior, still billing $400/month.
Pay for hours you use, commit to what's permanent
- Dev and test systems running 24/7 for a team that works 45 hours a week - auto-shutdown schedules cut those costs by ~65%.
- Steady production workloads on pay-as-you-go pricing when 1-3 year reservations or an Azure savings plan cut the same compute 30-60%.
- Old backups and file shares on premium storage tiers when archive tiers cost ~a large share less for data nobody touches.
The principle: match the pricing model to the workload's real shape. Commit to what's permanent, schedule what's intermittent, archive what's cold.
Governance: how it stays lean after the cleanup
A one-time cleanup decays back to sprawl within a year without three habits: budgets with alerts on every subscription (so surprises appear as notifications, not invoices), tagging so every resource has an owner and a purpose, and a quarterly review where the usage reports get read by someone accountable.
That accountability is the actual product when you hire cost governance - the checklist above is public, but somebody has to own it. It's part of how we run cloud services for clients: the bill gets reviewed like it's our own money.