Do we need Microsoft 365 Copilot, or is ChatGPT enough?

It depends on where your work already lives. If your email, files and meetings are in Microsoft 365, Copilot is usually the lower-risk choice because it works inside tools you already control, with permissions you already manage. A standalone tool can be excellent for drafting and long documents, but it is another place your data lives, so it needs its own rules. We will tell you which fits rather than which carries the better margin.

Is Copilot safe with our client information?

It is as safe as the permissions underneath it. Copilot only returns what the signed-in person already has access to, so over-permissioned folders and stale sharing links become visible the moment it goes live. That is why we fix SharePoint structure, access and sensitivity labels first, and why we test it with real accounts before a rollout.

Will AI replace any of our staff?

In the work we do, it removes the drafting, filing and searching that nobody enjoys, and it does not decide anything on its own. The realistic change is that the same team handles more volume, with a person reviewing the output. Where a client wants AI in a decision about people, we insist on a human in the loop and a policy that says so.

What does an AI engagement cost?

It depends on how much groundwork is needed. A policy and readiness assessment is a fixed, small piece of work. A Copilot roll-out with permissions, labels, agents and training is quoted per environment after the discovery session, and automation work is scoped per process. You get a number before anything starts, and we will tell you if the honest answer is that you are not ready.

Is this legal advice?

No. We are an IT provider, not lawyers. We help you work out what the technology does, what data it touches, and what practical controls to put around it, and we point at the obligations that apply in Canada. Your own counsel should confirm the policy, and we will happily work through it with them.

Will you use our data to train anything?

No. We do not train models on client data. Where a vendor's plan uses your content to improve its service, we make that clear before a licence is bought and help you choose a configuration that does not. Vendor data terms are reviewed at the start and revisited when they change.

What happens when a new AI tool turns up?

You call us and we assess it against your policy: what data it touches, where that data goes, what the vendor's terms say, and whether it fits your existing stack. You get a written view rather than a trial that quietly becomes production. That is part of ongoing support, not a separate project.

AI consulting & automation · Toronto & the GTA

Your team already uses AI. Nobody has written the rules.

We turn scattered, unapproved AI use into something you can defend: a plain use policy, files and permissions that are actually ready, Microsoft 365 Copilot or the automations worth having, and training on the work your people really do.

Policy and permissions first, licences second. Then the same engineers who built it keep it working.

Draft your AI policy 416-479-0505

Plain AI use policy, written with you
Files and permissions tidied before Copilot
Training by role, on your own work
Supported by the engineers who set it up

The order that works

01
Rules
Which tools, which data, who signs off.
02
Permissions
Who can see what, before an assistant looks.
03
Rollout
One real job at a time, measured.
04
Training
By role, on their own work.

The licence is the last decision, not the first.

Free starting point

Draft a plain AI use policy in a minute.

Choose the tools you will allow and the data you hold. You get plain rules you can put in front of staff today — then we can help you finish it and back it with actual settings.

Which AI tools will you allow?

What sensitive data do you hold?

Could AI help decide anything about people?

Your policy, first draft

6 rules

1.Only these AI tools may be used for work, signed in with a work account: Microsoft 365 Copilot.
2.Personal AI accounts and free consumer tools are not used for anything to do with the business.
3.Customer names, contact details and account numbers are only used in approved tools whose data terms confirm the information is not used to train the model.
4.AI may suggest, but a person reviews and owns every decision that affects a customer, employee or supplier.
5.A person reads and owns everything AI drafts before it leaves the business.
6.Anything odd an AI tool does, or anything sent by mistake, is reported to IT immediately.

Finish it with us →

A starting point, not legal advice. Before an assistant goes live, check who can see what in SharePoint and OneDrive — it shows people what they already have access to.

What has to change

AI that saves time without leaking data.

Buying a licence takes ten minutes. These four things decide whether it earns its place or becomes a subscription nobody opens.

Everyone knows the rules

One plain policy covering which tools are approved, what may go into them, and who signs off on the output.

Approved tools
Data rules
Human sign-off

Copilot sees the right files

SharePoint structure, permissions and sensitivity labels are sorted before anyone is given a licence.

SharePoint and OneDrive
Sensitivity labels
Least access

Training on their own work

Each team learns on the tasks they actually do — quoting, filing, reporting — not a generic product demo.

By role
Coaching
Team champions

Someone to call

When a tool changes, an agent misbehaves or staff have a new question, you call the same engineers who built it.

Support
Updates
Advice

Where it pays off today

The jobs AI is genuinely good at.

We would rather do five things well than promise a transformation. These are the tasks where we consistently see hours come back, with a person still approving the output.

Document review

Pull key terms and dates from long contracts and policies. A person confirms the result before it is relied on.

Meeting notes

Summaries and action items from Teams meetings, filed where the team can find them instead of in someone's notebook.

First drafts

Proposals, quotes and follow-ups drafted from your own past examples, so your team edits rather than starts from nothing.

Invoices and forms

Data extracted and posted into your accounting or ERP system, with a review step before anything is posted.

Staff questions

A single-purpose agent that answers IT and HR questions and hands off to a person when it is unsure.

Why AI projects stall

Six ways this quietly fails.

None of these are model problems. They are preparation problems, and each one is avoidable before a licence is bought.

Licences before groundwork

Copilot bought before the files, permissions and labels were ready, so it answers with the wrong things and trust is gone.

No policy at all

Staff guess what is allowed. Half stop using AI, the other half use their personal account and paste in client data.

Generic training

A demo that never touches anyone's real job. People leave impressed and change nothing about how they work.

Permissions too broad

Copilot surfaces files people technically had access to and should never have seen. The rollout becomes an audit.

No baseline

Nobody timed the task before, so nobody can prove the gain and the subscription looks like another cost line.

IT-only rollout

The teams who would actually use it were never in the room, so the tools solve the wrong problems.

The one nobody admits

Some of your team is already using personal AI accounts on work devices, pasting in client details to get a quote written faster. That is data leaving the building with no record of where it went. The fix is not a ban — it is an approved tool and a clear rule, which is exactly what the policy above is for.

Platform choice

Copilot, ChatGPT, Claude or Gemini?

The tool matters far less than the data and the rules around it. Here is how we usually think about the choice — including the reason to say no to one.

ToolUsually fitsWatch for

Microsoft 365 CopilotBusinesses already living in Outlook, Teams and OfficeOnly as good as your SharePoint permissions

ChatGPT businessA flexible standalone tool across several teamsAnother place your data lives, outside Microsoft 365

Claude businessLong documents, policy work and careful writingAnother place your data lives, outside Microsoft 365

Gemini in Google WorkspaceBusinesses already on Google WorkspaceRarely the right pick for a Microsoft-based office

Our view, not a vendor claim. Check each provider's current data terms for the plan you buy.

The engagement

What we actually do, step by step.

Most engagements start with discovery and a readiness check. Not every one goes all the way to agents and automation — some businesses only need the first three steps done properly.

Discovery

Which AI tools are already in use, which jobs are worth targeting, and how long those jobs take today so there is a baseline.

Set the rules

Acceptable use policy, data rules and a decision on where AI may not be used at all.

Sort the foundations

SharePoint structure, permissions, sensitivity labels and data retention, so an assistant only finds what the person should see.

Roll out, one job at a time

Copilot in Word, Excel, Outlook and Teams for the first use cases, with settings documented as we go.

Agents and automation

Single-purpose agents for IT, HR or customer questions, plus Power Automate for invoices, forms and document routing.

Train by role

Hands-on sessions on real work, champions in each team, and follow-up once the novelty wears off.

Measure it

Compare the baseline with what actually changed, and drop whatever did not earn its licence.

Support and advise

Ongoing support, new tool evaluations and a written position on each, so you are not deciding from a vendor pitch.

The Canadian picture

What the rules say about your client data.

You do not need a legal team to run AI sensibly, but you do need to know where your obligations sit. These are the questions we ask in a discovery session.

We are an IT provider, not lawyers. We explain what the technology does and what controls to put around it, and your counsel confirms the policy.

PIPEDA and your client data

PIPEDA applies to personal information you collect in the course of business. Putting a customer's details into an AI tool is a disclosure, so the vendor's data terms, retention and location matter — and a free consumer account is rarely an appropriate place for it.

Automated decisions

Quebec's privacy law requires disclosure when a decision about a person is based exclusively on automated processing. If AI ever shortlists, scores or prices, the policy and the vendor settings have to reflect that. Elsewhere in Canada, disclosure obligations are tightening.

Records and accountability

You need to be able to say which AI tools are in use, what they were used for, and who approved the output. That is an audit trail question, and it is much cheaper to design than to reconstruct.

Clients

What people say when we pick up.

Long-standing relationships across manufacturing, professional services, transport and healthcare — many measured in years, not months.

“Sun IT Solutions has given us predictable monthly IT budget costs, eliminating surprises and downtime. They provided us with the right IT solution and a comprehensive business plan to meet all future needs.”

Kevin Stanley

President, Skyreach Window Cleaning Inc.

“They are quick to respond to any issues or questions we have and Lloyd always offers solutions so that the same problems don't arise again. Thank you to Lloyd and his team — we can focus on our business while they take care of our IT needs.”

Mike Mintsopoulos

President, Truck Equipment & Service Co. Ltd

“Following our serious server crash the service they provided our business was the absolute best. A technician was onsite almost immediately and worked non-stop to get us back up and running, with almost no downtime.”

Amelia Folco

President, Jupiter Industries Inc.

“Sun IT Solutions convinced us to switch to a VoIP telephone system. It has been virtually trouble free and our telephone costs dropped by more than 80%. We have no hesitation whatsoever in recommending them.”

Richard Waite

President, Salix Systems Limited

AI consulting FAQ

The questions we get before a laptop is opened.

Ask an engineer directly on 416-479-0505 — including the awkward questions.

With a policy and a permissions check, not a licence. Agree which tools are approved, where client and employee data may not go, and who signs off on AI-drafted output. Then check who can already see what in SharePoint and OneDrive, because any assistant shows people exactly the files they already have access to. Only after that does a roll-out make sense.

AI sits on top of your systems. Get those right too

Talk to an engineer

CybersecurityAI raises the stakes on phishing, identity and data access.Cloud servicesMicrosoft 365 and SharePoint foundations done properly.Managed IT servicesPatching, monitoring and support around the new tools.Why Sun ITHow we work, and what we will not pretend to know.

Start with a readiness check

Find out what Copilot would actually see.

We will review who can currently see what in SharePoint and OneDrive, where client data sits, and which two or three jobs are worth starting with. You get a written summary and a policy draft — whether or not you buy a licence from anyone.

Book an AI readiness check 416-479-0505

Rules first, licences second. If your files and permissions are not ready, we will say so before anyone spends money.

Offices in downtown Toronto and Markham · remote consulting across Canada