• Skip to primary navigation
  • Skip to main content
  • Skip to primary sidebar
  • Skip to footer
ControlUp Community

ControlUp Community

Connect, Learn, and Grow

  • Blog
  • Archives
  • Findings
  • Meetups
  • Videos
  • Events
  • Categories
    • ControlUp One Platform
    • ControlUp for Apps
    • ControlUp for Compliance
    • ControlUp Dashboards
    • ControlUp for Desktops
    • ControlUp for VDI
    • ControlUp Scripts & Triggers
    • ControlUp Synthetic Monitoring
    • ControlUp Workflows
  • Topics
    • Logos & Wallpaper
    • ControlUp.com
  • Join

Automating Windows 365 Cloud PC Rightsizing with AI Canvas: Cost and Performance Optimization

Posted on November 3, 2026

A new AI prompt has been developed for ControlUp‘s AI Canvas to automate the rightsizing evaluation of Windows 365 Cloud PCs within an organization. The prompt identifies every Windows 365 Cloud PC, determines its current plan based on CPU, RAM, and storage configuration, and analyzes recent performance metrics to assess whether the plan suits the workload. It estimates the potential savings or costs associated with resizing each Cloud PC and presents the results along with actionable recommendations.

The process begins by querying all Cloud PCs configured with the Microsoft Cloud PC provider. Devices not reporting back within the last 14 days are excluded from sizing recommendations but flagged separately since they may still incur licensing costs despite inactivity. The prompt calculates usage metrics over a recent 7-day window, focusing on CPU load (p95) and memory utilization (p95), plus storage consumption on the system disk. Oversizing is flagged when CPU and memory usage consistently remain low (under specified thresholds), prompting recommendations to downgrade one step in the plan ladder if the workload still fits comfortably within the smaller configuration. Conversely, oversizing occurs if resource usage exceeds defined upper limits, triggering an upgrade suggestion.

Storage sizing is handled separately, moving devices to plans with appropriately sized storage tiers based on usage, with thresholds designed to avoid undersized or oversized drives. Pricing uses Microsoft’s public Windows 365 Enterprise list prices as of November 2026, converted into cost differences reflecting the proposed resizing changes. The prompt outputs a comprehensive dashboard including a net monthly savings stat, a donut chart categorizing Cloud PCs by sizing verdict (right-sized, oversized, undersized, storage too small/large, not reporting, or unmatched), and a detailed table for all devices showing current and recommended plans, verdicts, pricing impact, VM size, region, and storage type. An AI-generated summary provides holistic findings, including potential savings, constraints on high-demand users, and caveats such as volume licensing discounts and workload peaks outside the measurement window.

Testing in lab environments indicates the prompt works reliably, producing clear and actionable results within acceptable execution times. Feedback suggested that the visual emphasis on total savings paired with detailed explanations improved usability. The solution currently covers Enterprise licenses only, with Business and Frontline licenses excluded from analysis, and the pricing reflects list prices that may differ from negotiated agreements. This prompt is intended to run as a single AI Canvas operation with efficient batch querying to stay within API limits, making it scalable for real-world tenant environments.

This AI prompt for Windows 365 rightsizing represents a powerful tool to optimize Cloud PC deployment costs and performance by providing IT teams with data-driven resizing recommendations and potential financial impact upfront. Further refinements may include extending support for additional license types and adjusting resource thresholds as usage patterns evolve.

For more information about Windows 365 Cloud PCs and rightsizing, see the official Microsoft documentation: https://docs.microsoft.com/en-us/windows-365/right-sizing-cloud-pc and ControlUp resources at https://docs.controlup.com.

Read the entire ‘Automating Windows 365 Cloud PC Rightsizing with AI Canvas: Cost and Performance Optimization’ thread below:

_Help us test a new prompt: Windows 365 rightsizing_

Before we add this to the Prompt Library, we’d like a few of you to run it on your own tenant. It finds every Windows 365 Cloud PC, works out which W365 plan it’s on, checks whether that plan fits the work being done on it, and puts a dollar figure on resizing.

_How to test_

1. Open AI Canvas and click _+ New canvas_
2. Paste the full prompt below into _Ask any question_ and send it
3. Reply in this thread with a your feedback if it’s working correctly or not
_What you should get_
• A stat with the net monthly saving in USD
• A donut of Cloud PCs by verdict: Right-sized, Oversized, Undersized, Storage too small, Storage too large, Not reporting, Not matched
• A table per Cloud PC: User, Device, Current plan, Recommended plan, Verdict, Reason, Monthly difference (list price), VM size, Region, Storage type
• A finding with the savings, the upsizing cost, the net, and what the numbers can’t see

_What we’d like to know_
• Did it find all your Cloud PCs, and is the _current plan_ right for each one?
• Do the recommendations make sense to you? Would you actually make these changes?
• Did it finish with all four pins? Did anything look off, or were any pins empty?
• Roughly how long did it take?

_Good to know_
• Prices are W365 _Enterprise_ US list prices (checked 3 Nov 2026). Your agreement price may be lower, and Business and Frontline licences aren’t covered.
• The thresholds are a first proposal: oversized below 40% p95 CPU _and_ 50% p95 memory, undersized above 80% CPU or 85% memory, and never downsized past 80% of the smaller plan.
• Cloud PCs that haven’t reported for 14 days get no verdict and are listed separately. Their licence may still be paying for nothing.

_The prompt_ (copy everything inside the box)

“`Rightsize our Windows 365 Cloud PCs and tell me what it is worth in dollars, in one pass of about 8 tool calls. *Find the Cloud PCs and their plan:* one query on `_devices` with `virt_desktop_provider` = 2 (Microsoft Cloud PC), returning name, console user, `last_communication`, `hw_cpu_vcore_count`, `hw_memory_total_bytes`, `os_disk_configuration` and `virt_properties`. Work out each Cloud PC’s plan from cores, RAM and the C: drive: RAM in bytes divided by 1024³ and rounded to the nearest plan size (34,307,309,568 bytes is 32 GB); storage is the `capacity` of the C: entry in `os_disk_configuration`, divided by 1024³ and matched to the nearest tier (549,162,299,392 bytes is 512 GB). Never use the `_all` disk fields: some older Cloud PCs also have a D: “Temporary Storage” drive that would inflate the total. Use `vmSize` from `virt_properties` only to confirm cores and RAM, and keep its `location` and `storageAccountType` for the table. Name every plan by its Microsoft name with “Windows 365” shortened to “W365” — “W365 Enterprise 8 vCPU, 32 GB, 512 GB” — and if cores, RAM and storage match no plan below, or the size is a GPU size, list the device as not matched rather than forcing it into a plan. *Leave out Cloud PCs that are not reporting:* any with `last_communication` older than 14 days gets no sizing verdict. List them as “not reporting” with their plan and price: their licence may still be assigned to a Cloud PC nobody uses. *Measure the work:* one aggregation on `device_status` over the last 7 days, filtered to the reporting Cloud PCs by device name, with the p95 of `cpuload` and the p5 of `memory_available_percentage` per device — p95 memory used is 100 minus that p5. Take C: drive fill from the same `os_disk_configuration` entry. If `device_health_ddc` returns a sizing verdict for these devices, mention it in Reason only where it disagrees with the measurements; if it returns nothing, decide from the measurements alone and say so in the finding. *Decide and recommend, one step at a time.* Compute oversized: p95 CPU under 40% and p95 memory used under 50% — move one step down the ladder (32/128, 16/64, 8/32, 4/16, 2/8, 2/4 in vCPU/GB), but only if the user’s p95 memory use would stay under 80% of the smaller plan’s RAM and p95 CPU under 80% of its cores; otherwise keep it and say why. Compute undersized: p95 CPU over 80% or p95 memory used over 85% — one step up. Storage: choose the smallest tier the recommended plan offers where the space used on C: is at most 75% of the tier, so a nearly full C: drive moves up a tier and a mostly empty one can move down. Give each Cloud PC exactly one verdict: “Oversized” or “Undersized” when compute changes (whatever storage does), “Storage too small” or “Storage too large” when only storage changes, and “Right-sized” only when nothing changes. *Price every plan with these Windows 365 Enterprise list prices*, USD per user per month, from Microsoft’s published price list (checked 3 Nov 2026): Windows 365 Enterprise 2 vCPU, 4 GB, 64 GB $28; 2 vCPU, 4 GB, 128 GB $31; 2 vCPU, 4 GB, 256 GB $40; 2 vCPU, 8 GB, 128 GB $41; 2 vCPU, 8 GB, 256 GB $50; 4 vCPU, 16 GB, 128 GB $66; 4 vCPU, 16 GB, 256 GB $75; 4 vCPU, 16 GB, 512 GB $101; 8 vCPU, 32 GB, 128 GB $123; 8 vCPU, 32 GB, 256 GB $132; 8 vCPU, 32 GB, 512 GB $158; 16 vCPU, 64 GB, 512 GB $277; 16 vCPU, 64 GB, 1 TB $315; 32 vCPU, 128 GB, 1 TB $665; 32 vCPU, 128 GB, 2 TB $765; GPU Select, 256GB $310; GPU Standard $537; GPU Super $1,029; GPU Max $1,914. GPU Select, 512GB has no published price: say so if it appears. Call them list prices and say the organisation’s agreement price may be lower. Report the downsizing saving, the upsizing cost and the net separately, per month and per year, and the not-reporting Cloud PCs as a separate possible saving. Pin in this order: (1) a `stat` with the net monthly saving in USD, with the number of Cloud PCs to downsize as its description; (2) a `donut_chart` of the Cloud PCs by verdict — Right-sized, Oversized, Undersized, Storage too small, Storage too large, Not reporting, Not matched; (3) a `data_table` of every Cloud PC, sorted by monthly difference with the biggest saving first, with exactly these columns in this order: User, Device, Current plan, Recommended plan, Verdict, Reason, Monthly difference (list price, USD), VM size, Region, Storage type — VM size, Region and Storage type are the `vmSize`, `location` and `storageAccountType` from `virt_properties` in separate columns. Put the evidence in Reason in one short sentence (“p95 CPU 2%, memory 36%; fits 4 vCPU / 16 GB at 71%” or “C: 96% full”), and leave prices, p95 values and drive fill out of their own columns. For a Not reporting Cloud PC, Recommended plan is “—”, the monthly difference is 0, and Reason names the last report date and its current price as a possible saving (“last reported 8 Sep; $158/month if the licence is no longer needed”), so the column still adds up to the stat; (4) an `ai_finding` with the monthly and yearly saving, the upsizing cost, the net, the not-reporting Cloud PCs as a separate possible saving, which users are most constrained, and what the figures cannot see — agreement discounts, Windows 365 Business or Frontline licences, and peaks outside the 7-day window. If no Cloud PCs are found, say so and stop; do not fall back to physical devices. Offer a resize `checklist`, an email to the affected users and a check of the not-reporting Cloud PCs’ licences as Next steps chips. Keep this to one pass so it finishes well inside the tool limit: batch each check into one fleet-wide query instead of repeating it per device, build every pin from what that pass finds, and offer anything deeper as Next steps chips.“`
Thanks! Every result helps, including “this didn’t work on our tenant”.


For my lab env it looks good…
The information I think is enough.
I do like the “fat number” with the savings and then the more or less detailed explanation in the left widget…

Personally I dont like to get my personal CloudPC downsized but thats a different topic


Thanks for testing it and sharing! Looks indeed good! Time to fill up your C drive so your right sized

Continue reading and comment on the thread ‘Automating Windows 365 Cloud PC Rightsizing with AI Canvas: Cost and Performance Optimization’.  Not a member? Join Here!


Categories: All Archives, ControlUp Dashboards
Topics: Cloud Computing, ControlUp Licensing, CPU, GPU, Microsoft, Microsoft Cloud PC, Microsoft Windows, Reporting, Scripts, Storage, Triggers, Unified Communications

Ask Us Anything, Connect, Learn, and Grow with the ControlUp Community!

Login to the ControlUp Community to ask us anything, stay up-to-date on what’s new and coming soon and meet other like-minded techies like you.

Not already a member? Join Today!

Primary Sidebar

ControlUp Academy

Enroll in ControlUp Academy for expert-led technical training, equipping you with skills to effectively deploy, manage, and grow your ControlUp investment.

Learn here >

Rotating Images

Hidden Gem from our Community on Slack!

ControlUp Betas - What's Coming Next?
NEW ControlUp Features - Stay Up-to-Date!
ControlUp Scripts - Scripting, Zero to Hero
Latest KB Articles - Be the First to Learn

Video Tutorials Library

Visit our technical how-to videos, offering step-by-step tutorials on advanced features, troubleshooting, and best practices.

Watch here >

ControlUp Blog

Check out the ControlUp blog for expert advice and in-depth analysis.

Read here >

ControlUp Script Library

Visit the ControlUp technical script library, which offers a multitude of pre-built scripts and custom actions for your monitoring and troubleshooting requirements.

See here >

ControlUp Support

Visit the ControlUp support home and to delve deeper into ControlUp DEX solutions.

Browse here >

Footer

      

ControlUp Community
Of Techie, By Techie, For Techie!

Terms of Use | Privacy Policy | Security
Dive Deeper, Learn more at ControlUp.com

  • facebook
  • twitter
  • youtube
  • linkedin

© 2023–2026 ControlUp Technologies LTD, All Rights Reserved.

We use cookies to ensure that we give you the best experience on our website. by continuing to use this site you agree to our Cookie policy..