AI-ready IT starts with the boring stuff: a Milwaukee MSP checklist
0
0
Right now every business owner is getting the same pitch. Add AI. Automate the busywork. Finally get useful answers out of your own data.
Some of it is real. A lot of it is hype. And the part nobody in the pitch deck wants to talk about is the IT foundation it all sits on.
Here is the uncomfortable version. If a former employee still has a live Microsoft 365 login, if half your files live in people’s personal OneDrive folders, if nobody has tested a backup in two years, or if the Wi-Fi falls over every time the conference room fills up, AI will not fix any of that. It will just run faster over the same mess.
We work with businesses around Milwaukee that want IT to stay out of the way: fewer outages, real security, support that actually picks up. Lately those same clients are also asking about AI tools and automation. Our answer rarely changes. Before you trust a model or a dashboard, make sure you trust the systems feeding it.
So here is the checklist we would run before anyone spends real money on AI.
AI does not clean up messy IT
An AI tool is only as good as the data and permissions sitting behind it.
If your sales numbers live in one system, project notes live in another, and the version everyone actually trusts is a spreadsheet on someone’s desktop, that is not really a data problem. It is an operations problem wearing a data costume. Shared passwords, stale accounts, aging laptops, personal phones nobody manages, cloud folders with no clear owner: same story.
Basic, sure. It is also exactly where most AI projects quietly die. The model was never the hard part. The hard part is knowing which data is current, who is allowed to see it, what breaks when something breaks, and how fast you can undo a mistake.
Good managed IT does not turn you into a data science shop. It makes the groundwork boring and reliable, which is the only condition under which the fancy stuff works at all.
Start with identity and access
One question first: who can get into what?
For most businesses the honest answer is “we’re pretty sure we know.” That stops being good enough the moment an AI tool can search across your files, read mailboxes, reach into cloud apps, or trigger actions on its own. Loose permissions used to be a quiet risk. Point a capable tool at them and the risk gets loud.
Where to start:
– Turn on multi-factor authentication for email, cloud apps, VPN, and admin accounts. All of them.
– Kill logins for former employees and vendors who moved on.
– Get rid of shared logins wherever you can.
– Audit your Microsoft 365 admin roles, mailbox delegation, and external sharing.
– Use groups and roles, not one-off permissions nobody remembers granting.
– Write down who owns each major system.
None of this is interesting. It is the difference between a helpful AI tool and a fast new way to leak data.
Know where your data actually lives
Most small businesses are sitting on more data stores than they think.
Files in SharePoint, Teams, OneDrive, Dropbox, an old server share, the accounting software, CRM exports, email attachments, the phone system, the scanner folder, and a few spreadsheets that have survived three office moves and two owners. Nobody planned it this way. It accumulated.
Before you wire anything up to a dashboard or an AI tool, map the sources that matter. You do not need an architecture diagram. A plain list does the job:
– Where does customer information live?
– Where do the financials live?
– Which system is the real source of truth for contacts, projects, tickets, inventory, jobs?
– Who is allowed to change those records?
– How do old files get archived or deleted, if ever?
– What should never be pasted into a public AI tool?
This usually surfaces a couple of habits people would rather not admit to. Good. Far better to find them now than during an incident.
Patch the boring holes before they get expensive
Patching, endpoint protection, firewall rules, locked-down Wi-Fi, a real device inventory, monitoring that someone actually watches. None of it is exciting. All of it decides whether anything you build on top holds up.
You cannot run analytics on flaky laptops and an unstable network, or off a server that has been “fine for now” since 2019. Your analyst feels those problems as delays. You feel them as lost hours and support bills.
A support process worth paying for can answer:
– Which machines are encrypted and protected, and which ones slipped through?
– Are the operating systems and the common apps actually patched?
– Is the team on managed company devices or their own?
– Is guest Wi-Fi walled off from your business systems?
– Are the firewalls and access points documented anywhere but one guy’s memory?
– Do alerts land somewhere a human reads them?
This is also where being local still earns its keep. Remote support is great right up until the problem is a dead access point, a firewall swap, a cabling fault, or a workstation that will not boot twenty minutes before payroll closes. For Milwaukee-area shops with a warehouse, a clinic, or a production floor, someone has to physically show up. That someone should not be a four-hour drive away.
Backups matter more once automation gets involved
Backups are easy to ignore until the one day they are the only thing that matters.
Automation raises the stakes because it touches a lot of records fast. Someone bulk-edits the wrong field. A sync pushes corrupted data. A workflow gets pointed at a system before anyone understands the blast radius. Forget ransomware for a second. Ordinary human mistakes do plenty of damage on their own, and now they do it quickly.
“We have backups” is not a backup plan. Better questions:
– What is actually backed up, and what quietly is not?
– How often do they run?
– How far back can you restore?
– When did anyone last test a restore? An untested backup is a hope, not a plan.
– If your main IT contact is on a plane, who else can recover the data?
– Are cloud files, email, and SaaS apps covered, or just the servers?
Give your data people a way to get unstuck
Data work dies by a thousand small IT cuts.
A user cannot reach a folder. A Power BI refresh breaks because someone rotated a credential. A laptop is out of memory. The one API key the team needs is buried in somebody’s inbox. A firewall is silently blocking the tool. And nobody can tell whether the issue belongs to Microsoft, the vendor, the ISP, or internal IT.
That is the MSP’s lane even when it is nowhere near the actual model. The job is to keep the environment stable, documented, and supportable so the business can use its tools without a daily fire drill. At minimum, your team should have:
– One clear place to ask for access or report a problem.
– Documented systems, vendors, licenses, and renewal dates.
– A real onboarding and offboarding process, not a tribal-knowledge ritual.
– Standard device builds for the people handling sensitive data.
– Someone who owns coordinating with outside vendors.
– Logging and alerting on the systems that matter.
That structure buys back time. It also stops a project from quietly depending on one person remembering how everything is wired.
Why the local angle still matters in Milwaukee
A lot of IT genuinely does not care where you are. Some of it cares a great deal.
Milwaukee-area businesses tend to run a mix of cloud tools and very physical hardware: conference room gear, printers, barcode scanners, shop-floor PCs, network closets, firewalls, access points, cameras, phones, and that one ancient line-of-business app the whole company secretly depends on. When the physical stuff fails, the fix means someone tracing a cable, swapping hardware, or arguing with an internet provider on your behalf.
Local also means context. A law office, a machine shop, a nonprofit, and a medical clinic do not use technology the same way, and they should not get the same cookie-cutter plan. For anyone rolling out AI or analytics, that difference is often what separates a clean launch from months of avoidable tickets.
The checklist, before your next AI or analytics project
Run this before you approve a new AI tool, reporting platform, or automation workflow:
– Review active users, admin accounts, and vendor access.
– Require MFA on every core system.
– Inventory company devices and confirm they are patched, encrypted, and protected.
– Identify the data sources the project will actually touch.
– Decide the source of truth for each type of data.
– Review cloud folder permissions and external sharing links.
– Define what data is allowed into AI tools and what is strictly off limits.
– Test a restore on anything you cannot afford to lose.
– Document vendors, support contacts, license owners, and renewal dates.
– Write a one-page incident plan for account compromise, data loss, and ransomware.
– Give employees one obvious place to ask for help.
This will not make you “AI-ready” by Friday. It will show you the weak spots that would have wrecked the timeline later, while they are still cheap to fix.
How Powerful IT Systems helps
We handle managed IT, support, cybersecurity, backups, cloud, and network projects for businesses around Milwaukee.
We are not here to drown you in jargon or sell a shiny tool before the basics work. The goal is an environment that is easier to secure, easier to support, and easier to build on. In practice that looks like cleaning up Microsoft 365, tightening endpoint protection, documenting the network, testing backups that have never been tested, fixing onboarding, and helping you decide which risks actually deserve attention first.
If AI, analytics, or automation is on your roadmap, start at the bottom of the stack. Get the accounts, devices, data, backups, and support process ready for the extra weight. The interesting work goes a lot better from there.
About Powerful IT Systems
Powerful IT Systems is a Milwaukee-area managed IT services provider working with small and midsized businesses on IT support, cybersecurity, Microsoft 365, backups and disaster recovery, cloud management, and network projects. More at https://powerful-it.com.
Artificial Intelligence – The Data Scientist