Techwerks 27

bw-car-vehicle

CIAOPS Techwerks face to face returns to Melbourne CBD on Thursday the 22nd of May 2025. The venue for the event will be:

Melbourne City College
Level 9, 120 Spencer St
MELBOURNE VIC 3000

The course is limited to 20 people and you can sign up and reserve your place now! You reserve a place by completing this form:

http://bit.ly/ciaopsroi

or by sending me an email (director@ciaops.com) expressing your interest.

The content of these all day face to face workshops is driven by the attendees. That means we cover exactly what people want to see and focus on doing hands on, real world scenarios. Attendees can vote on topics they’d like to see covered prior to the day and we continue to target exactly what the small group of attendees wants to see. Thus, this is an excellent way to get really deep into the technology and have all the questions you’ve been dying to know answered. Typically, the event produces a number of best practice take aways for each attendee. This event will largely focus on AI including Copilot, Agents and Agent creation, etc with a special focus on what is relevant for small business.

Recent testimonial – “I just wanted to say a big thank you to Robert for the Brisbane Techworks day. It is such a good format with each attendee asking what matters them and the whole interactive nature of the day. So much better than death by PowerPoint.” – Mike H.

The cost to attend is:

Gold Enterprise Patron = $50 ex GST

Gold Patron = $90 ex GST

Silver Patron = $180 ex GST

Bronze Patron = $360 ex GST

Non Patron = $720 ex GST

I hope to see you there.

SharePoint Agents PAYG costs

image

To get a better idea of the costs of using SharePoint Agents, I’d suggest you have a look at:

https://techcommunity.microsoft.com/blog/spblog/consumption-based-pricing-for-sharepoint-agents/4389591

with the highlight being:

Under the PAYGO model, customers are billed $0.01 per message. Each interaction with a SharePoint agent uses thirty-two (32) messages, so customers are billed at $0.32 per interaction with SharePoint agents. The PAYGO meter uses your Azure subscription as the payment instrument, ensuring seamless integration with existing billing processes. This meter is available worldwide.

and

There are no in-product feature differences between the PAYGO meter, and the SharePoint agent included in the Microsoft 365 Copilot license. Users have the same capabilities and benefits, regardless of the billing model they choose.

Thus, with each interaction being $0.32, let say that typically a user will interact with SharePoint agents three times during any inquiry. That makes it about $1 per enquiry. If we now say that an average user will make 20 inquiries per day, that is $20 per user per day. Multiply that across all the users in an organisation and you can see how it could get very expensive very quickly.

Clearly then, pay as you go SharePoint agents is for very low volume of usage across the organisation, typically one enquiry per day. Otherwise, it make more sense to buy a full license of Microsoft 365 Copilot for the user in question because they effectively get unlimited SharePoint agent enquiries as well as a personal AI assistant plus more.

If you combine any other pay as go usage of Copilot, such as with Copilot Studio as I have outlined before, then it make far more sense to get a full Microsoft 365 Copilot for those who need to use any AI tools. However, pay as you go billing does provide you the flexibility to mix and match with full Microsoft 365 Copilot licenses. If you have a business with 5 major users and 20 casual users then teh starting point is for those 5 users to have full Microsoft 365 Copilot license, while the rest simply use an Azure subscription to cover any incidental costs until the point when another person in the business needs a full license.

To keep control of any SharePoint or Copilot pay as you go, you shoudl always set up a budget in Azure as I have outlined before with Security Copilot

Pay as you SharePoint agents do provide a degree of flexibility of quickly and easily enabling AI across your SharePoint information for your whole organisation but if usage of AI starts to grow then so too will the costs, and potentially quite dramatically if appropriate limits are not configured. The best option with pay as you go SharePoint agents then is its use in combination with full Microsoft 365 Copilot licenses for users who need to use AI extensively in their jobs, while casual users can remain on the pay as you go option. The good news is that you do have the flexibility to mix and match with the two types of licenses as needed and Azure does give you the added benefit of being able to turn off immediately where Microsoft 365 Copilot licenses are typically an annual commitment.

Copilot Studio PAYG costs

Now that I have set up pay as you go (PAYG) Copilot Studio via an Azure subscription, the next big question is what are the costs likely to be? These are somewhat hard to quantify exactly because it ‘depends’ on a lot of factors.

Start with:

Copilot Studio licensing here:

https://learn.microsoft.com/en-us/microsoft-copilot-studio/billing-licensing

which says:

  • Pay-as-you-go: $0.01 per message

but then it isn’t a simple ratio of 1 question = 1 message, oh no. You need to look at this:

Message scenarios

which gives you this table:

Screenshot 2025-03-13 140428

The example Microsoft provides is:

Diagram illustrating various Copilot Studio events and their corresponding billing events.

Each interaction with an agent might utilize multiple message types simultaneously. For example, an agent grounded in a tenant Microsoft Graph could use 32 messages (30 messages for the Microsoft Graph grounding, and two for generative answers) to respond to a single complex prompt from a user.

Agent costs depend on an agent’s complexity and its usage.

Inside the Power Platform admin center, under licensing and Copilot Studio I see this:

Screenshot 2025-03-13 141042

if I drill into this a little more I find:

Screenshot 2025-03-13 141024

Ok, so 2,040 messages is the usage.

I then waited and checked my Azure billing for the period and it reports:

Screenshot 2025-03-13 134801

which is AU$20.30 for Copilot Studio usage across those 2,040 messages I suggest. If you divide the cost by the messages you come out to around that suggested $0.01 per message as expected.

How does that relate to usage? Again, hard to exactly quantify as I was the only user and I was building and testing an autonomous agent with Copilot Studio for around 8 hours roughly. Thus, that means an average of 255 AI message per hour or 4.25 messages per minute.

Based on that, the best estimate (rule of thumb) I could give you would be, based on ‘average use’ across a typical day (8 hours), for a single user using Copilot regularly throughout the day the cost is going you around AU$20 per user for that 8 hours of sustained usage.

I fully appreciate this is nowhere near exact but, so far it is the best average I can come up with for sustained daily usage.

If we assume that a ‘normal’ user is not going to using AI in the sustained manner across the whole day we could then apply say a 50% usage discount and settle on around AU$10 per user per day for an ‘average’ user using Copilot resources in an ‘average’ way per day. More intensive usage would be considered around AU$20 per user per day I suggest.

In summary then, via my imperfect observations and calculations I would suggest to you that if you do indeed implement Copilot service via Pay As You Go (PAYG) then the ‘typical’ costs you can expect would be around AU$10 per user per day up to AU$20 per user per day. If this was sustained across a full month then you would be looking at $300 per average user per month which is way above the cost of a full license of Microsoft 365 Copilot whih which would be a flat fee of around AU$45 per user per month.

This is the best estimate I can give you and your costs and usage will vary but I think $10 per user per day for average Copilot use on a PAYG plan is as good as any place to start.

Clearly then, if your users are planning on sustained Microsoft 365 Copilot usage a paid license of Microsoft 365 Copilot is a much more effective investment from what I can determine.

Copilot for M365 Administration

image

A while back, I spoke about the fact that Microsoft was bringing Copilot to many of the administration portals in Microsoft 365. As you can see above I now have Copilot in my Microsoft 365 admin center.

image

You can see that it provides a pretty comprehensive answer when I ask it a common administration task such as resetting a users password.

All of this is available to all administrators provided there is just ONE full license of Microsoft 365 Copilot in a tenant! You can read more at :

Copilot in Microsoft 365 admin centers

I would expect to see Copilot for M365 administrator start to appear in more and more places in the future. How much easier will email troubleshooting appear when it arrives there?

Need to Know podcast–Episode 339

In this episode I try and explain Microsoft 365 Copilot Chat and where it fits into the Microsoft 365 environment, which is challenging without visual but hopefully it gives you a better idea. In essence Microsoft 365 Copilot chat is free to all Microsoft 365 users and provides the same AI across just the web but with commercial data protection. You can extend those capabilities by adding metered billing if desired before stepping up to the full paid version of Microsoft 365 Copilot.

I also have Microsoft Cloud news and updates for you as well as part of this episode so tune in.

Brought to you by www.ciaopspatron.com

You can listen directly to this episode at:

https://ciaops.podbean.com/e/episode-339-copilot-chat/

Subscribe via iTunes at:

https://itunes.apple.com/au/podcast/ciaops-need-to-know-podcasts/id406891445?mt=2

or Spotify:

https://open.spotify.com/show/7ejj00cOuw8977GnnE2lPb

Don’t forget to give the show a rating as well as send me any feedback or suggestions you may have for the show.

Resources

@directorcia

Join my shared channel

CIAOPS merch store

Become a CIAOPS Patron

CIAOPS Blog

CIAOPS Brief

CIAOPSLabs

Support CIAOPS

Access Copilot anywhere on the canvas in Word for the web

Microsoft Lists forms: What’s New

Diagnose Safe/Blocked Senders Issues in Microsoft 365

Do more with Copilot and agents

Microsoft Copilot help & learning

Securing Identities: 10 recommendations for building a stronger identity security posture

Accelerating the Anomalous Sign-Ins detection with Microsoft Entra ID and Security Copilot

Get visibility into your DeepSeek use with Defender for Cloud Apps

Microsoft Loop, a year in review: greater integration and richer collaboration

Introducing new Surface Copilot+ PCs for Business

Copilot for all: Introducing Microsoft 365 Copilot Chat

The impact of AI on the MSP business model

image

Today, I liken the impact of AI for MSPs to the scene inside the garbage crusher in the original Star Wars movie. On one side is the impact AI is having on the existing MSP model around configurations and maintenance and on the other is the challenge of how to provide new AI services to customers.

The starting point is to examine the current MSP business model which is largely based on a reactive approach, that is, get paid for fixing issues after they occur. The managed service that most MSP’s sell is a kind of insurance policy. This means the client agrees to pay a regularly fee per month and for that the MSP will ensure they are available to resolve issues that occur during that time period at no additional cost to the customer. The incentive then for the MSP is to thus reduce the chances of problems occurring by configuring systems to be as reliable as possible. However, no matter how much this is done things will still go wrong and require a reactive approach to resolution.

Enter AI. It is clear that AI is become more and more integrated into services sold to the customer. One simple example is Microsoft Security Copilot. This service can look across all the security signals in an environment and assist with investigations and incident response. We are beginning to see Microsoft Security Copilot being extended from a stand alone chat style interface to direct integration with Intune. This means that it can report and troubleshoot on policies used throughout Intune. At the moment this integration is largely just reporting but that ability to actually make changes and configure Intune directly cannot be far away. AI will soon be able to do the job of the MSP with Intune I would suggest.

Likewise, Security Copilot is also available in Defender XDR and Entra ID. I’m sure it won’t be long before it appears in places like Exchange Online and SharePoint Online as well as across the whole Microsoft 365 administration portals. In fact, that capability is already in preview (Copilot in Microsoft 365 admin centers). It won’t be long before it is available for every tenant. The Microsoft 365 administration portals used to be the sole purview of the MSP. No longer, AI will take up a lot of the load and probably allow customers to do most of the administration tasks that an MSP does today such as resetting accounts, creating users, assigning licenses, etc.

AI is really good a evaluating data and them coming to a decision about what option is best in a given circumstance. It is in fact probably going to be able to better evaluate the security of a customers environment and determine what settings should be enabled or disabled to provide this. What it will soon be able to do is actually take those recommended actions. In a world where AI is automatically handling the administration of a Microsoft 365 environment, what now is the role of an MSP, given the AI is largely doing what they used to do for a fee?

How will AI mitigate those challenging errors that also occur for users that you can’t plan for you may ask? Take a look at this example from the keynote at the last Microsoft Build conference :

https://youtu.be/8OviTSFqucI?si=j0oI1kbmbRgrvaSe&t=1260

(at time stamp 21:00 if needed)

It shows Copilot playing Minecraft by reacting to what is on the screen directly. Now extend that concept to desktop support where the AI is constantly watching and can interact directly with a user if an error appears. It may also get to the stage where the AI takes care of the error immediately without an interaction from the customer or MSP. AI today has the capability to see and talk based on its environment. As this matures I surely see it challenging the traditional help desk concept, especially for MSPs. Still not convinced? Take a look at this video of ChatGPT 4o interacting with data on the screen:

https://www.youtube.com/live/DQacCB9tDaw?si=j-KvPcNJwypvk1U9&t=1105

( at time stamp 18:34 if needed)

and remember that was back in May 2024! The capabilities have only increased since then (hello Sora)!

Can AI do every maintenance role that an MSP can do for their customers today? Not yet, but I very confident that it will do more and more over time (aka the walls are definitely coming in thanks to AI).

If AI is reducing the maintenance side of the MSP managed service model, where is the opportunity selling AI services to customers? When a customer wants Microsoft 365 Copilot, they simply buy a license and assign it to a user. That’s it! Microsoft 365 Copilot will automatically appear for the user as an icon that will open Copilot Bizchat so they now have an AI agent they can generate answers from. Microsoft 365 Copilot will automatically appear in all their desktop apps as well such as Word, Excel, Teams, and so on without the need for further configuration. Microsoft 365 Copilot will also automatically appear in SharePoint and Onedrive. The list goes on, Forms, Loop, etc, etc all without ANY further configuration.

As for AI training tools, they are already available such as Prompt Coach and are free. There is also the Copilot Prompt Gallery that the user has access to, again for free. There are also services like the Microsoft Copilot Academy available for free and integrated into the Microsoft 365 Copilot subscription. The list of free embedded training material is extensive. This is going to challenge an MSP to provide provide something that is better than what is already available and how will an MSP be able to charge a fee for that when quality embedded training is already available for free?

Once Microsoft 365 Coplot is in place I can’t see how it will need any maintenance. It doesn’t need password resets, it doesn’t need delivery troubleshooting, it doesn’t need to restored, it will just work. It won’t break or required support as other services MSPs supported did. In a world where services don’t require a managed maintenance service, how does the tradition MSP revenue model apply?

It is important to remember that Microsoft 365 Copilot doesn’t have any settings, such as for security. It relies on existing services like Entra ID, SharePoint permissions, DLP policies, Data Labelling and so on. These security settings really should already be in place prior to Microsoft 365 Copilot being enabled and once configured they largely won’t require any form on ongoing maintenance. As I have also suggested previously, I think the AI itself will play a bigger and bigger role in evaluating and acting to ensure Microsoft 365 environments remain constantly secure. Once again, the need to ongoing maintenance is reduced or eliminated which means another hit to the MSP business model.

The direction that most vendors like Microsoft are encouraging MSPs to move to is around building ‘apps’ or ‘agents’ for their customers to solve business challenges. The challenge there for MSPs, as I have called out before, is that fact that the majority are not skilled or experienced in the ‘creator’ model we see today. That model means taking tools such a Copilot Studio to create these ‘agents’. The big change to the MSP business model is thus a shift from reactive to proactive. Unfortunately, I just do not see any evidence of MSPs in general understanding or embracing this as part of their business. Most continue to place their faith in the old reactive business model, which introduces huge risks for their business. The biggest of these is that it allows an ‘AI aware’ provider to solve customer challenges with agents and then potentially scoop up the best of the business from the customer.

These are the reasons why I see most traditional (infrastructure focused) MSPs being stuck in the proverbial Star Wars compactor. Moving to an AI business model is a huge change in approach and it can certainly be done but I am not seeing it being embraced. To me, it harkens back to the early days of the cloud but I feel the AI transformation will have a far greater impact on MSPs of today than the cloud ever did. It is not too late to include a true focus on delivering AI effectively to customers while also using AI to minimise tenant maintenance costs but any effective strategy cannot be grounded in the status quo. You can’t expect to continue to apply the same old MSP business model and expect to be successful. The AI model is different. The AI model is proactive. The AI model is about code.

The walls are closing in from both sides on the traditional MSP business model from what I see and there is precious little time to escape. Much like in the Star Wars movie, the saviours to the compactor conundrum will be the bots (R2-D2 and C3PO in the movie), but not unless you invoke them.