CIAOPS Image comparison page

Every time a new image model from Microsoft becomes available I use the same test prompt to see what the result is and compare that to previous attempts. The last one of these was here:

https://blog.ciaops.com/2026/08/15/new-microsoft-image-model-2/

With new models coming out all the time, the latest is MAI-Image-2.6-Flash available today, I figured a better option going forward is to put them all on a single web page for comparison which I have now done here:

Screenshot 2026-09-05 085040

https://directorcia.github.io/Office365/image-comparison.html

You can view all the model results side by side as well compare the selection you want using the selector in the top left.

I’ll keep adding new results using my standard image test prompt and posting the results here to have a reference going forward.

Let me know if you have any other image models I should test and post.

The Inbox Tells the Truth

image

I have seen this happen more than once. A business owner turns up every day, answers the same messages, attends the same meetings, and wonders why the energy has disappeared. Nothing is technically broken. Clients remain. Invoices still go out. But the work has become something to survive rather than something to shape.

That is usually when people look for a new tool, market, or strategy. Sometimes that is needed. But I reckon there is a smaller test that tells the truth faster.

Look at the last thing you sent.

Not the polished proposal. I mean the ordinary email you sent to a client, supplier, or your own team. If it landed in your inbox from someone else, would you read it properly? Would it make the next step clearer?

If the honest answer is no, that is not just a writing problem. It is a leadership signal.

Your inbox reflects your business

Email is easy to dismiss because it feels routine. That is why it matters. The routine is where culture leaks out.

A vague email usually points to vague thinking. A rushed reply often means no one has stopped to decide what actually matters. A long message that says little normally means the sender is hoping volume will do the work that clarity should have done.

I am not judging that from a distance. I have written plenty of those messages myself. The kind where the subject line is weak, the point arrives too late, and the recipient has to work out what I want.

When that becomes normal, I give away control.

Copilot should challenge the thinking

This is where Microsoft 365 Copilot becomes interesting, but not for the reason many people expect.

The lazy version is to ask Copilot in Outlook to write the email and then send whatever appears. That may save time, but it does not fix the real problem. It can make it worse, because unclear thinking is now wrapped in cleaner sentences.

The better use is to make Copilot test your intent.

Before sending an important reply, ask: what is the decision in this message? What action am I asking for? What could the reader misunderstand? Can this be shorter without becoming blunt?

That is a different relationship with the tool. Copilot is not the ghost-writer. It is the mirror. It helps me see where I have avoided making the point.

The same applies in Teams. If a channel conversation needs ten messages to agree on one next step, the issue is probably not Teams. It is the absence of a clear owner, decision, or sentence.

Small standards rebuild confidence

When a business feels flat, the temptation is to chase a big reset.

Momentum often returns through smaller acts of discipline.

Write the subject line like it matters. Put the decision near the top. Tell people what happens next. Use Copilot to tighten your thinking, not decorate your uncertainty. Move the discussion into Teams when the email thread has become a swamp, then summarise the decision so the next person does not have to excavate it.

That may sound basic. It is. That is the point.

Businesses rarely lose their edge all at once. They drift. Standards slip in small places first. The inbox gets dull. Meetings become foggy. Documents become harder to read. Everyone keeps moving, but fewer people know why.

The way back starts in one ordinary place.

Open your sent items. Read your own work as if you were the client. If it would not earn your attention, fix the standard there.

That is where authority starts to return: with one clearer message, one cleaner decision, and one less piece of work that asks the reader to do your thinking for you.

The Algorithm Is Choosing for You

image

A few weeks ago, I caught myself doing something I suspect many people do every day. I opened LinkedIn looking for one thing and, twenty minutes later, realised I’d spent my time reading something completely different.

It wasn’t an accident.

The platform had quietly decided what it thought I should see. The posts, the comments, the videos, the suggested connections. All selected and arranged by an algorithm making thousands of decisions on my behalf.

And it’s not just social media anymore.

The same thing is now happening with AI.

The New Digital Gatekeepers

For years, we worried about search engines deciding which websites appeared on the first page of results. Today, that influence runs much deeper.

Your social media feed determines which opinions receive your attention. News aggregators decide which stories deserve prominence. Video platforms choose what appears next. Every click, pause and interaction becomes data that fine-tunes the content delivered back to you.

Most people believe they’re making independent choices about what they consume.

In reality, they’re often choosing from a menu that someone else prepared.

The concerning part isn’t that algorithms exist. We need them. The volume of information available today makes manual filtering impossible.

The concern is that most people rarely stop to consider who is tuning the filter.

AI Is Taking This Further

The shift becomes even more interesting when we look at AI chatbots.

When you ask a question in ChatGPT, Copilot, Gemini or another AI tool, you’re not receiving the entire internet. You’re receiving a curated response generated from a combination of training data, system instructions, ranking mechanisms and safety controls.

Again, none of this is inherently bad. In fact, it’s often incredibly useful.

But it does mean that the answer you see is the result of countless decisions made before you ever typed your prompt.

I’ve noticed this in my own work with Microsoft 365 Copilot.

Ask Copilot to summarise a long Teams meeting and it determines what information is important enough to include. Ask it to analyse an email thread in Outlook and it decides which points deserve emphasis. Ask it to draft a document in Word and it makes choices about structure, tone and content.

The productivity benefits are enormous.

However, it also means we’re increasingly consuming information that has already been filtered, prioritised and interpreted on our behalf.

The Real Skill Is Learning to Challenge the Output

I don’t think the answer is to reject algorithms or AI.

That ship has sailed.

The answer is developing the habit of questioning what you’re shown.

When I read something that strongly reinforces my existing view, I ask myself whether I’m being shown an opposing perspective. When Copilot gives me an answer, I often ask a follow-up question from a different angle. When an AI summary highlights a particular issue, I sometimes go back to the original source material.

The organisations that get the most value from AI will be the ones that understand this distinction.

Using AI well isn’t simply about writing better prompts.

It’s about recognising that every response is a starting point, not a final truth.

Don’t Outsource Your Thinking

One of the biggest risks I see isn’t that AI becomes smarter.

It’s that people become less curious.

Algorithms are incredibly good at feeding us information that feels familiar. AI is incredibly good at giving us answers that sound complete. Together, they can create the illusion that we’re seeing the whole picture when we’re really seeing a carefully filtered version of it.

The irony is that tools like Microsoft 365 Copilot can actually help us become better thinkers when used correctly. Ask it to challenge your assumptions. Ask it to find alternative viewpoints in a SharePoint document library. Ask it to identify gaps in a proposal you’re drafting in Word.

Use the technology to expand your perspective rather than narrow it.

Because whether it’s your social media feed or your favourite AI assistant, someone or something is helping determine what you see.

The important question is whether you’re still deciding what to think.

We have a new LLM in Copilot winner

image

A while back I compared all the different LLM’s (i.e. GPT, Claude) in various Copilot services (i.e. Chat, Cowork) here:

https://blog.ciaops.com/2026/08/17/comparing-llms-in-copilot-services-round-7-final/

The ultimate winner was:

Opus 5 (Cowork)

which cost US$22 to complete the work.

Microsoft has just announced that Fable 5.1 is available in Cowork so I took it for a spin through the same prompt I used in the recent testing. The results are here (as judged by Grok):

https://github.com/directorcia/general/blob/master/Copilot/Comparisons/20260902-Cowork-Fable51-Grok-eval.md

and the new winner is now:

Fable 5.1 (Cowork)

you can see the result for yourself here:

https://github.com/directorcia/general/blob/master/Copilot/Comparisons/20260816/20260902-Cowork-Fable-51.docx

along with all the others outputs:

https://github.com/directorcia/general/tree/master/Copilot/Comparisons/20260816

Here is an update on the costs in Cowork for each model:

Fable 5.1 = 3,453 credits

Fable 5 (Preview) = 3,387 credits [Model no longer shown]
Fable 5 (Copilot)(Preview) = 3,373.8 credits [Model no longer shown]

GPT 5.6 Sol = 2,800.50 credits
Sonnet 5 = 2,509.3 credits
Opus 4.8 = 2,487 credits [Model no longer shown]
Opus 5 = 2,240 credits

GPT 5.5 = 1,200 credits
GPT 5.6 Terra = 260 credits

In summary then, the new best model, Fable 5.1, produces the best results from my standard prompt and is also the most expensive to use using that prompt.

The comparison report for all document paramteres has been updated here:

https://github.com/directorcia/general/blob/master/Copilot/Comparisons/20260816-File-paramters.md

The next updated model that we expect to see will probably be GPT 6.0 which is expected in the October timeframe if not sooner. As soon as it becomes available I’ll test it and update this information.

The Game You Are Actually Playing

image

I see a lot of MSP owners get into trouble at the point where the business starts to feel slow.

The pipeline is not moving the way it used to. The team meeting feels a bit flat. The same service issues keep coming around again. Nothing is broken enough to panic, but nothing is improving enough to create energy either.

That is a dangerous place to sit, because a quiet business rarely stays quiet for long.

Borrowing from the wrong playbook

When progress gets hard, the temptation is to look sideways. You see another MSP posting about a new offer, a new pricing model, a new automation stack, or a new AI service, and the instinct is to copy it.

I understand why. Doing something feels better than sitting with the problem. A fresh tactic gives you the sense that movement has returned.

The issue is that another provider may not be playing the same game as you.

A fifty-person MSP with dedicated account managers, mature project delivery, and a bench of engineers can make moves that will punish a ten-person business still trying to standardise its support desk. A one-person consultant can make promises that a larger team cannot deliver consistently. A cybersecurity-focused provider can price and position differently from a generalist MSP still carrying legacy client expectations.

The tactic may be sound. It just may not belong at your stage.

Progress needs context

This is where I think Microsoft 365 Copilot becomes useful, not as some magic answer, but as a way to get clearer about the current state of the business.

Before chasing the next idea, I would rather see an MSP owner ask Copilot in Teams to summarise recurring themes from recent leadership meetings, or use Copilot in Outlook to pull together the common patterns in customer escalations. If the team keeps discussing noisy clients, delayed projects, margin pressure, or weak documentation, that is a signal.

Then take that signal into Excel. Look at ticket trends, project slippage, licence margin, and client profitability. The goal is not to produce a beautiful report. The goal is to work out what level of the game you are really in.

Are you trying to win new business when delivery is already stretched? Are you selling AI readiness while your own SharePoint permissions are messy? Are you chasing strategic advisory revenue while your QBR process is still a slide deck built the night before?

That is not failure. That is feedback.

The next move should fit the level

Good strategy is not about copying the smartest-looking move in the market. It is about choosing the move your organisation can actually absorb.

For one MSP, that might mean tightening onboarding before launching a new managed Copilot service. For another, it might mean cleaning up agreement profitability before hiring another technical resource. For another, it might mean using Planner and Teams to create a simple weekly operating rhythm, so the leadership group can see what is stuck before it becomes expensive.

None of that sounds glamorous. That is partly the point.

A business usually regains energy when it can see progress again. Not noise. Not theatre. Real progress. A decision made. A blocker removed. A client segment clarified. A service standard improved.

The MSPs I see do well are not always the ones with the cleverest tactic. They are the ones with enough discipline to ask, “What game are we actually playing right now?” and enough honesty to answer it.

That is where momentum comes back.

Not from copying someone else’s move, but from making the next right move for your own business.

Standards Are Not Gatekeeping. They Are Protection.

image

I have sat in plenty of rooms where someone suggests lowering the bar to get more people involved. Make the rules softer. Reduce the expectations. Stop being so particular about process, documentation, security, handover, client communication, whatever the current friction point happens to be.

It always sounds reasonable at first.

But there is a pattern I have seen too many times to ignore: when a room has no clear standards, it does not attract the best people. It attracts the people who prefer ambiguity because ambiguity lets them avoid accountability.

That is a hard lesson for many MSPs to learn.

The bar tells people what matters

Standards are not about making life difficult. They are about making expectations visible.

If your team has a standard for documenting client environments in SharePoint, then everyone knows where the source of truth lives. If meeting notes and actions are captured in Teams or Loop, people know what was agreed and who owns it. If Copilot is being used to summarise client discussions, draft follow-up emails in Outlook, or prepare service review notes in Word, there should be a clear expectation about what information is appropriate to include and what must stay out.

Without that, everyone invents their own version of good enough.

That is when things drift.

One person writes detailed notes. Another leaves three vague lines in a ticket. One engineer checks conditional access changes against a peer review process. Another makes the change directly because “the client needed it quickly.” One account manager records agreed actions. Another relies on memory and an inbox search three weeks later.

The issue is not that people are bad. The issue is that the organisation has allowed different definitions of acceptable work to exist at the same time.

That never scales.

Lowering standards rarely solves the real problem

When standards feel hard to meet, the answer is not always to lower them. Sometimes the answer is to make them easier to follow.

That is where Microsoft 365 and Copilot can help, but only if the business already knows what good looks like.

For example, if your MSP expects every client meeting to end with clear actions, use Copilot in Teams to help summarise the discussion, then review and clean it up before sending it from Outlook. If your team struggles to keep project notes consistent, create a simple Word or Loop template that defines the structure. If client documentation is scattered, fix the SharePoint libraries and permissions before telling people to “just ask Copilot.”

Copilot does not remove the need for standards. It exposes where they are missing.

If your process is messy, Copilot will help you move faster through the mess. If your permissions are sloppy, Copilot will show people more of what they already had access to. If your team has never agreed what a good client handover looks like, Copilot will not magically create that discipline for you.

AI makes good standards more valuable, not less.

A serious business needs serious expectations

I am not talking about perfection. No MSP runs perfectly. Every team has busy days, awkward clients, rushed escalations, incomplete tickets, and moments where someone has to make a judgement call.

But there is a difference between flexibility and having no standard at all.

A good standard says, “This is how we do things here.” It gives strong people confidence because they know the work will not be undermined by shortcuts. It gives newer people a path to improve. It gives leaders something fair to manage against.

Most importantly, it protects the business from becoming a place where the lowest acceptable behaviour becomes normal.

If you want a better team, better clients, and better outcomes from tools like Copilot, start by being clear about the room you are building.

The people worth having will not be scared away by standards.

They will be attracted by them.

Stop Asking People To Buy Before They Care

image

A lot of businesses get marketing backwards.

They start with the offer. The demo. The booking link. The licence. The proposal. The “contact us” button sitting proudly at the bottom of a page that almost nobody has any reason to trust yet.

I see this in the MSP world all the time. An MSP wants more clients, so the instinct is to push harder. More ads. More campaigns. More landing pages.

The problem is not always visibility. Sometimes the market has not been given enough reasons to remember you.

Familiarity Comes Before The Sale

People rarely buy serious services from strangers. They buy from names they have heard before, ideas they have seen before, and people who have already helped them think more clearly.

That does not mean giving everything away. It means giving enough useful thinking away that the right people start associating your name with practical answers.

For an MSP, that might be a short post explaining why unmanaged devices are still a risk. It might be a simple onboarding checklist. It might be a plain-English explanation of Microsoft 365 backup, conditional access, or why Copilot needs clean information architecture before it becomes genuinely useful.

The point is not to sound clever. The point is to be useful before anyone is ready to buy.

If a business owner has seen your guidance in LinkedIn posts, Teams community discussions, newsletters, and client-facing articles, you are no longer just another supplier. You become part of how they understand the problem.

That matters.

Your Best Ideas Should Not Be Locked Away

Some MSPs worry that sharing too much will remove the need for the client to engage them.

I think the opposite is usually true.

A client who reads your explanation of a security issue and realises there are five more things they had not considered is not less likely to need help. They are more likely to respect the work. Good content does not replace expertise. It reveals the depth of it.

This is where Microsoft 365 gives MSPs a practical advantage. You can build a simple content rhythm inside the tools you already use. Capture repeated client questions in a Teams channel. Use Copilot in Word to turn rough notes from a technical conversation into a first draft. Store approved client education pieces in SharePoint so the team can reuse and improve them. Ask Copilot in Outlook to help shape a reply when a prospect asks, “Why do we need this?”

None of that replaces your judgement. It just reduces the friction between knowing something useful and getting it in front of the people who need to hear it.

Trust Is Built In Advance

The best time to earn attention is before the sales conversation starts.

By the time someone asks for a proposal, they have often already made a judgement. They have looked at what you say. They have noticed whether your advice is consistent. They have probably decided whether you understand their world.

That is why content is not decoration. It is part of the buying journey.

For MSPs, this matters because trust is the product before the product. Clients are handing over access to systems, identity, data, devices, security controls, and business continuity. They are not just buying labour. They are buying confidence.

You do not create that confidence by shouting louder at the last minute.

You create it by showing up with useful ideas before the prospect is ready to sign anything.

So my view is simple. If you want people to buy from you later, help them think better now.

Be remembered for useful work before you ask for commercial action.

That is not soft marketing. That is how trust compounds.

Copilot Has to Earn Its Seat

image

The more I speak with SMBs about Copilot, the more I think the price objection is often a cover story.

Yes, cost matters. It always does in a small business. A few extra licences can feel very real when margins are tight and every subscription renewal is already under scrutiny. But when I push beyond the first objection, the real issue is usually not, “This costs too much.”

It is more like, “I cannot yet see how this changes the way we work.”

That is a very different conversation.

A tool is easy to ignore. A better way of working is not.

If Copilot is positioned as another clever AI thing, most business owners will nod politely and move on. They have already seen plenty of clever AI things. Some staff are probably already using personal AI tools. Someone in the office has used it to write a birthday message, summarise a web page, or produce a passable first draft of something.

Interesting? Yes.

Worth paying for across the business? Not automatically.

The problem is that many Copilot conversations start with the product rather than the pain. They start with “Copilot can draft in Word” or “Copilot can summarise email in Outlook” or “Copilot can recap meetings in Teams”. Those are useful examples, but they are not business cases by themselves.

A business does not buy a licence because Word has a new button. It buys when the proposal that used to take half a day now has a solid first draft before lunch. It buys when the owner stops spending Sunday night catching up on email threads. It buys when Teams meeting notes no longer depend on the most organised person in the room remembering to write them up.

That is where the value starts to become visible.

Start with the job, not the SKU

The strongest Copilot deployments I see do not begin with “how many users should we licence?” They begin with one annoying, expensive, repeated business problem.

A sales team that loses time assembling proposals.

An operations manager who spends Friday afternoon turning messy updates into a weekly report.

A service desk where documentation is always the thing everyone agrees matters, but nobody has time to do properly.

A business owner who is constantly dragged into the same internal questions because the answer exists somewhere in SharePoint, but nobody can find it quickly.

Those are the entry points.

Once the business can see Copilot working inside that flow — Outlook for client follow-up, Teams for meeting recaps, Word for first-draft proposals, SharePoint as the source of truth — the conversation changes. It stops being about whether an AI licence is expensive. It becomes about whether the current way of working is more expensive.

And often it is.

Not always in a line item on the P&L. More often in delay, rework, missed follow-up, inconsistent communication, and staff doing low-value admin instead of useful work.

Capacity is the better conversation

For SMBs, I think the better framing is not productivity theatre. It is capacity.

Can the same team produce more proposals without hiring?

Can the owner get out of the inbox and back into decisions?

Can new staff get productive faster because Copilot can help them understand previous work, meeting context and internal documents?

Can service desk notes become consistent enough that the next technician does not have to rediscover the same answer?

That is the business value. Not magic. Not transformation by press release. Just less drag in the work that already exists.

This is why I do not like broad licence rollouts as the starting point. They create cost before confidence. A smaller, focused pilot around one measurable business problem is far more useful. Pick the problem. Pick the people closest to it. Use Copilot against real work. Measure what moved.

If the outcome is obvious, expansion becomes easy.

If it is not obvious, buying more licences will not fix it.

Copilot has to earn its seat in the business. When it does, the renewal discussion becomes much less about cost and much more about what the business can now do with the same people, the same hours, and far less friction.