Over the past few years, many businesses have done careful work on how employees sign in to Microsoft 365. Multi-factor authentication. Conditional access policies that allow only approved devices, or that block sign-ins from places your team never works. Monitoring that flags unusual logins. It’s a real front door with real locks.

Then AI arrived, and in many organizations it came in through a side entrance.

How AI Tools End Up Outside Your Controls

It usually happens innocently. An employee signs up for an AI tool with their work email and a password they chose. Another uses a personal account because it was faster. A department buys a team plan on a credit card. Each tool has its own login, its own MFA setting (or none), and its own list of who has access.

None of these accounts are governed by the conditional access policies you built. They don’t know whether the device is company-managed. They don’t check where the sign-in is coming from. When an employee leaves, their Microsoft 365 account is disabled on day one, but the AI account they created may stay active until someone remembers it exists, if anyone does.

That would be a minor issue if AI tools were minor applications. They aren’t. As we described in our last post, AI tools are often connected to email, files, and business systems, and they hold a history of what employees have shared with them. That makes them attractive targets, and the easiest way into an attractive target is the login nobody is watching.

Single Sign-On: One Identity, One Set of Rules

Single sign-on (SSO) means employees sign in to an AI tool using their company identity, typically Microsoft Entra ID for Microsoft 365 businesses, instead of a separate username and password. Most business-grade AI platforms support it, and Microsoft 365 Copilot uses your Microsoft 365 identity by design.

The change sounds administrative. In practice it moves your AI tools behind the front door you already built:

  • Your conditional access policies apply. If your policy requires a managed, compliant device for email, it can require the same for AI. If it blocks sign-ins from unexpected countries, that protection now covers AI too.
  • Your MFA applies. Users authenticate with the same strong methods they use for everything else, rather than a weaker setting chosen tool by tool.
  • Offboarding becomes one step. Disable a departing employee’s account and their AI access ends with it.
  • Sign-ins become visible. AI sign-ins appear in the same identity logs your monitoring already reviews, so unusual access to an AI tool is something you can see, not something you discover later.
  • Access follows roles. Group-based assignment means people get the AI tools appropriate to their job, and access can be reviewed alongside everything else.

You don’t build a second security program for AI. You extend the one you have.

A Sensible Order of Operations

If you’re starting from a mixed environment, here’s a practical sequence:

  1. Inventory what’s in use. Ask teams which AI tools they rely on. Sign-in and network data can fill in what people forget to mention.
  2. Standardize on business plans. Free and consumer tiers generally don’t offer SSO, admin controls, or the data protections a business should expect. Decide which tools the company supports and consolidate onto them.
  3. Connect them to your identity provider. Configure SSO and, where the platform allows it, require it so password-only sign-in is no longer an option.
  4. Extend conditional access. Apply the same device and location requirements you use for Microsoft 365. Consider stricter policies for AI tools connected to sensitive data.
  5. Review connections and permissions. Know which AI tools can reach email, files, or other systems, and confirm the underlying permissions are still right.
  6. Put it in policy. A short AI use policy that names approved tools and states that personal accounts aren’t for company data turns these controls into an expectation everyone understands.

Security That Makes AI Easier to Say Yes To

There’s a benefit here beyond risk reduction. When AI tools sit behind your company identity, leadership can approve new tools with confidence, employees get a familiar sign-in, and IT can grant or remove access in minutes. Good controls don’t slow AI adoption. They are what allow it to scale.

This is everyday work for Go West IT. We manage Microsoft Entra ID, conditional access, device management, and monitoring for businesses across financial services, legal, professional services, and more, and we help clients bring new applications, AI included, under the controls they already trust. If you’d like to know which of your AI tools are inside your front door and which are still using the side entrance, let’s take a look together.

Go Boldly. We’ll guide the way.

Most businesses we talk to are past the question of whether to use AI. Someone on the team is already drafting proposals with it, summarizing meetings, analyzing spreadsheets, or building a small tool that saves the operations team hours every week. The value is real, and it is arriving faster than most technology shifts we have seen.

The next question follows naturally: how do we secure it?

It’s a good question, and the market is eager to answer it with a new category of products. Some of those products will earn a place. But for most small and mid-sized businesses, the best place to start securing AI isn’t a new tool at all. It’s the operational security foundation you may already have, or may have been meaning to get to.

AI Raises the Value of Everything It Touches

To use AI well, you connect it to things. Your email. Your files. Your CRM, your calendar, your financial data. That’s where the productivity comes from: an assistant that can see your work can actually help with it.

It also changes the math for an attacker. A compromised AI account isn’t one more login. Depending on how it’s connected, it can be a single place to search a company’s email, read its documents, and review months of conversations where employees pasted in client details, contract terms, and internal plans. Instead of digging through a mailbox one message at a time, an attacker can simply ask.

Attackers are also using AI themselves, to write more convincing phishing, find unpatched systems faster, and work at a pace human defenders struggle to match. We covered that side of it in Detection at Machine Speed.

None of this is a reason to slow down on AI. It’s a reason to make sure the ground under it is solid.

The Foundation Is the Same. The Stakes Are Higher.

Here is the encouraging part. The controls that protect AI are largely the controls that protect everything else. Operational security is not a new discipline. It’s the set of fundamentals a well-run IT environment should already have in place. What has changed is how much rides on it.

  • Identity management. Every user has one company identity, protected with strong multi-factor authentication, with access that matches their role. When someone leaves, their access ends everywhere on the same day. This matters more now because AI tools inherit whatever access a user has. If permissions have drifted over the years, an AI assistant will find every file a user can technically open, including the ones they were never meant to see.
  • Device management. Every laptop and phone that touches company data is known, enrolled, and configured to a standard. AI increasingly runs on the device itself, through desktop apps, browser extensions, and agents that can take actions on a user’s behalf. A managed device is how you know those tools are running somewhere trustworthy.
  • Endpoint detection and response (EDR). Protection on every device that watches for malicious behavior and can isolate a machine when something goes wrong. If an attacker steals a signed-in session from a browser, this is often where it shows up first.
  • Patching. Operating systems, browsers, and applications updated on a disciplined schedule. AI is helping attackers find vulnerabilities faster than ever, which shortens the window between a fix being released and that weakness being used against you.
  • Monitoring. Sign-ins and account activity watched around the clock, by people and tools with the authority to act when behavior doesn’t look right.
  • Conditional access and single sign-on for AI apps. AI tools signed into through your company identity, rather than separate usernames and passwords, so the same rules that protect email protect AI: approved devices only, strong MFA, and access that disappears the moment an employee departs.

That last item deserves its own conversation, and it’s the subject of our next post.

If These Slipped to the Back Burner, Now Is the Time

For many businesses, these fundamentals have been “mostly done” for a while. MFA is on for most people. Most devices are patched most of the time. Monitoring exists, but nobody is quite sure who is watching after hours. That was a tolerable position when the stakes were lower.

AI changes that. When one account can reach this much, gaps that seemed minor become meaningful. The upside is that closing them is well-understood work. It doesn’t require guessing what threat comes next, and every step you take protects your AI tools and the rest of your business at the same time.

A practical way to begin:

  • Confirm every AI tool in use is a company account, not a personal one, and that you know who has access.
  • Review what your AI tools are connected to, and whether the underlying permissions still make sense.
  • Check that every device accessing company data is managed, patched, and running EDR.
  • Bring AI apps under your company sign-in so your existing conditional access policies apply.
  • Make sure someone is watching account activity outside business hours.

Opportunity on a Solid Footing

We’ve written before about whether your business is ready for AI and about the operational security practices that hold up in the age of AI attacks. This is where those two conversations meet. The businesses that get the most from AI will be the ones that can adopt it confidently, because they trust the foundation underneath it.

That foundation is what we build and maintain for our clients every day: identity, devices, patching, monitoring, and access controls, kept current so you can focus on what AI makes possible. If you’d like an honest look at where your environment stands before you expand your use of AI, we’re here to help.

Go Boldly. We’ll guide the way.

If you run a small or mid-sized business, the security market has a way of making you feel behind.

Every few months there’s a new category with a new acronym. Endpoint detection. Then extended detection. Then identity threat detection. Security posture management for the cloud, and a different one for SaaS. Each is legitimate. Each solves a real problem. And each arrives with the unspoken suggestion that if you don’t have it, you’re exposed.

For a company with forty employees and no security team, that’s an impossible position. You can’t evaluate a market you don’t have time to follow, and you can’t staff a function that enterprises fill with a dozen people. So the honest question isn’t “which of these do we need?” It’s a more practical one: who is responsible for knowing?

The Gap Isn’t Technology, It’s Adoption

Here’s something that gets lost in security coverage. The tools genuinely work. The detection capabilities available today are far better than what existed three years ago, and the good ones are no longer priced only for the Fortune 500.

The gap is that capability exists and doesn’t get deployed. Not because businesses are careless, but because adopting a new security platform is a project: evaluate the market, sit through demos, check whether the vendor is durable, negotiate, integrate it with what you already run, tune it so it doesn’t drown you in false alarms, decide who responds when it fires, and train someone to operate it. That’s real work, and it lands on people who already have full-time jobs.

Multiply that by the pace of the market, and the outcome is predictable. Sophisticated protection stays available and unused.

You Shouldn’t Have to Become a Security Company

This is the part worth saying plainly. A manufacturing company should be good at manufacturing. An advisory firm should be good at advising. Neither should have to develop an internal competency in evaluating identity threat detection platforms, any more than they would develop one in commercial insurance underwriting.

What they should have is someone whose job it is to know, whose incentives are aligned with theirs, and who brings the capability to them already working.

That’s the actual service. Not a product catalog, but a standing responsibility to watch the market, separate the genuinely useful from the well marketed, and put the useful things in place before a client has to ask for them.

What That Looks Like in Practice

Earlier this year we went looking at the identity threat detection space, because we could see the same pattern across our client base: compromises that started with a completely legitimate-looking Microsoft 365 login, where the attacker was already inside before anything flagged. The existing layers were doing their jobs and still missing it.

We evaluated the category, chose a platform purpose-built for Microsoft 365, confirmed it came with 24/7 SOC services behind it (a SOC being a Security Operations Center, staffed around the clock), and added it to Go Secured | Advanced Cloud and Go Managed | Comprehensive.

The part that matters for clients: from their side, that was not a project. No RFP. No evaluation cycle. No new line item to approve, no implementation to schedule, no platform to learn. The capability was simply there, running, monitored, and responded to, because that’s what they’re paying us to do.

That’s the model. Sophisticated on the back end. Uncomplicated for your business.

Questions Worth Asking Whoever Handles Your IT

This is useful whether or not you ever work with us. If you want to know where you actually stand, these four questions get you there quickly:

  • What has changed in our security stack in the last twelve months? If the answer is nothing, that’s worth a conversation. The threat landscape moved; a defense that didn’t is quietly falling behind.
  • Who is watching at 2 a.m., and what are they authorized to do? Monitoring that only produces a morning report isn’t response. Find out whether someone can actually act, and act without waiting for approval.
  • If our Microsoft 365 were compromised, how would we know, and how would we prove what was touched? The second half of that question matters as much as the first, particularly if you’re regulated or carry cyber insurance.
  • What are you watching in the market on our behalf? A provider who can answer this is doing the job. One who only responds to what you ask for is a vendor, not a partner.

If your current provider answers these well, that’s genuinely good news. If they can’t, you’ve learned something useful, and you know where to start.

Both Sides of the Coin

The security landscape is more complex than it has ever been, and it will keep getting more complex. That’s not a problem anyone is going to solve. And your experience of it does not have to be complicated. Those two things are entirely compatible, as long as someone is carrying the complexity for you.

The businesses that stay ahead aren’t the ones that chase every new category. They’re the ones who made sure somebody competent was accountable for knowing which ones mattered.

Where to Start

  • Curious where your Microsoft 365 environment stands? Our free Microsoft 365 Security Assessment gives you a prioritized, plain-language picture of how your environment is configured today. It also includes a look back across the last six months of tenant activity, so if something has already happened and is still sitting there unaddressed, you find out now rather than during an audit or an incident. No obligation, and useful even if you never call us again.
  • Wondering whether you’re getting this from your current provider? Let’s talk. We’ll tell you straight.

You don’t need to become an expert in every new category. You need a partner who already is. Go Boldly. We’ll guide the way.

Ask most business owners how they would find out about a security incident, and the answer comes back some version of: someone would notice. An alert would come in. A report would flag it. Somebody would look into it.

That answer made sense for a long time. It doesn’t anymore, and the reason isn’t that people got worse at their jobs. It’s that the clock changed.

We wrote recently about what happens after the login: how attackers now hold a valid Microsoft 365 session, move quietly inside your tenant, and never trip a traditional alert. The natural follow-up question is the one this post is about. Once that activity is detectable, how fast does the response have to be to actually matter?

Faster than a person can read.

The Math Stopped Working

Consider the sequence of a modern account takeover. An attacker gets a valid session. Within minutes, they create an inbox rule to hide their tracks. They register an application to keep access after any password reset. They add their own MFA method. They start reading. On a good day, the whole opening sequence takes less time than a coffee break.

Now consider the response side of that same window. An alert fires into a queue. It waits for someone to open it. That person checks whether it’s real, because most alerts aren’t. They escalate. Someone with authority decides to act. Then, finally, somebody ends the session.

Even run well, that’s a process measured in hours. The attack was measured in minutes. The gap between those two numbers is where data loss happens, and no amount of hiring closes it, because the problem isn’t effort. It’s arithmetic.

Attackers Adopted AI First

There’s an uncomfortable truth in security right now. The offensive side moved faster than the defensive side.

Attackers use AI to write phishing that reads like a colleague wrote it, in your industry’s vocabulary, referencing your actual vendors. They use it to sort through a stolen mailbox and find the payment thread in seconds instead of hours. They use it to run the same play against hundreds of businesses at once, which is why campaigns that used to target enterprises now land in small-business inboxes.

The practical consequence is that volume and speed both went up at the same time. More attempts, arriving faster, each one better tailored than the last. A defense built on human review was already strained. Against machine-speed offense, it is outmatched. Not because the humans are wrong, but because they are being asked to win a race they were never built to run.

Three in the Morning on a Holiday Weekend

There’s a second dimension to this, and it’s simpler. Attackers don’t keep your hours.

Compromises are deliberately timed for when nobody is watching. Friday evening. The middle of the night. The Wednesday before Thanksgiving. That timing isn’t incidental. It’s the whole strategy. An attacker who gets sixty hours of undisturbed access over a long weekend doesn’t need to be sophisticated. They just need the room.

This is why “24/7” in security has to mean something specific. Not a phone number that reaches someone eventually. Not a dashboard that gets reviewed Monday morning. Detection that is genuinely running at 3 a.m., paired with response that is authorized and able to act at 3 a.m., without waiting for anyone to wake up.

What Machine-Speed Detection Actually Looks Like

The good news, and there is real good news here, is that this problem has an answer, and it’s a well understood one.

  • Behavioral detection rather than rule-matching. Instead of checking a login against a list of known-bad indicators, the system studies how the session behaves. That’s what catches the compromise that passes every location check and every MFA prompt, because the sign-in looked perfectly legitimate. The behavior afterward didn’t.
  • Automated containment. When activity is clearly malicious, the response executes immediately: ending the attacker’s sessions, removing the devices and MFA methods they registered, deleting the inbox rules and revoking the applications they left behind. Machine speed on offense answered with machine speed on defense.
  • A forensic record built as it goes. The timeline of what was accessed, changed, and sent is assembled automatically, so the account of what happened rests on captured facts rather than on reconstruction after the fact.
  • Human judgment where judgment belongs. This is the part that gets lost in AI conversations. Automation handles the clock. A SOC, meaning a Security Operations Center staffed 24 hours a day, handles the questions that require a person: what this means for your business, what else to check, what to tell your clients, whether this was the whole of it. The machine buys the time. People decide what to do with it.

For Go West clients, this capability is delivered through Petra Security, an identity threat detection and response platform purpose-built for Microsoft 365, with 24/7 SOC services behind it. It is included in Go Secured | Advanced Cloud and in Go Managed | Comprehensive, not as a separate purchase or an add-on line.

Both Sides of the Coin

Here’s the honest framing. Attacks are faster, more numerous, and better written than they were two years ago, and that trend isn’t reversing. And the response to them is better than it has ever been: automated where speed matters, human where judgment matters, running while you sleep.

The businesses that come through these events well aren’t the ones that never got targeted. They’re the ones whose defenses were running at the same speed as the attack.

Where to Start

You can’t slow the attackers down. You can make sure nothing is waiting on someone to check a queue. Go Boldly. We’ll guide the way.

Most people still picture a breach as someone breaking in — cracking a password, slipping past the firewall, tripping an alarm. That picture is out of date. Today, the most damaging Microsoft 365 compromises don’t look like a break-in at all. They look like a normal login.

We wrote recently about how attackers now walk around multi-factor authentication — stealing session tokens, relaying real Microsoft sign-in pages, and abusing legitimate features to get a valid session without ever needing your password again. Once they hold that session, they’re inside your environment as a trusted user. Nothing looks wrong. And that raises two questions most businesses can’t answer:

If someone were inside your Microsoft 365 right now, would you see it? And if they had been, could you prove exactly what they touched?

The Breach That Doesn’t Look Like a Breach

When an attacker logs in with a valid session, the front-door defenses have already done their job and moved on. MFA verified the user. Email security scanned the inbound messages. Both are essential — and both are looking at the entrance, not at what happens once someone is inside.

So the attacker settles in quietly. They read email to learn your billing cycles and who approves payments. They create a hidden inbox rule that auto-files their own messages so you never see the replies. They register an OAuth app to keep access even after a password reset. They browse SharePoint and OneDrive for the documents that matter. None of this trips a traditional alert, because none of it is technically an “attack” — it’s a logged-in user doing ordinary things, just not the real user.

This is the gap. Verifying who logged in is not the same as watching what they do next.

What Modern Microsoft 365 Detection and Response Actually Does

Closing that gap takes a different kind of monitoring — one built specifically for identity-based threats in Microsoft 365. Here’s what that looks like in practice:

  • Behavioral login analysis. Instead of asking only “is this login from an expected location?”, it studies the behavior of the session itself. That’s how it catches the attacks that pass location checks — logins routed through residential proxies, sessions riding on a stolen token, credentials that already cleared MFA.
  • Watching activity after the login. It monitors what actually happens inside the tenant — new inbox rules, app registrations, MFA method changes, mailbox and file access, unusual sends and deletions, permission changes. The tell-tale signs of a takeover live in these actions, not in the sign-in itself.
  • Fast, automated containment. When something is clearly malicious, waiting hours for a human to notice is waiting too long. The response can move immediately — ending the attacker’s sessions, removing the devices and MFA methods they registered, and deleting the malicious inbox rules they left behind — to shut the door before real damage is done.
  • Forensic reconstruction. After an incident, it rebuilds the full timeline: how they got in, what they accessed, what they changed, what left the building. Not guesswork — a documented account of exactly what happened.

Behind Go West’s service, this capability is powered by Petra Security, an identity threat detection and response platform purpose-built for Microsoft 365, backed by our 24/7 Security Operations Center. The technology sees the threat; our SOC investigates and responds — around the clock, including the nights and weekends when these attacks are timed to land.

Why the Forensic Record Matters as Much as the Alert

For a regulated business — an RIA, a law firm, a family office — detecting the compromise is only half of it. The other half is being able to answer, with confidence, what happened.

After a Microsoft 365 incident, you may need to tell a regulator, a cyber-insurance carrier, or an affected client precisely what was and wasn’t accessed. “We think we contained it” is not an answer that holds up. A clear forensic timeline — the attacker’s infrastructure, the mailboxes and files they touched, the data that did or didn’t leave — turns a frightening, open-ended event into a defined, defensible one. It shortens the incident, supports the insurance claim, and lets you make disclosure decisions based on fact rather than fear.

That’s the difference between an event that controls you and one you control.

The Part That Has to Happen First

Here’s what surprises most businesses: a forensic timeline is only as good as the data behind it — and Microsoft 365 does not capture all of that data by default. The detailed logging that reveals which mailboxes were opened, which files were accessed, and what left your tenant has to be deliberately turned on, configured correctly, and retained long enough to matter — before an incident, not after. Plug even the best detection tool into a tenant that was never logging the right activity, and it has very little to work with. The evidence you’ll want on your worst day is created by choices you make on an ordinary one.

This is exactly why the smart move is to be proactive. Preventing, detecting, containing, and remediating a compromise aren’t four products you scramble to buy in a crisis — they’re one posture you put in place ahead of time. That means getting the foundation right first: logging and retention configured properly, detection tuned to the right activity, containment ready to act, and a response team already on call. It’s the work worth doing with a partner like Go West IT while everything is quiet — so that if the day ever comes, it’s all already in place.

Both Sides of the Coin

Here’s the confident version of all this: identity-based compromise is real and it’s rising — and it is now something you can genuinely see, stop, and account for. The businesses that come through these events well aren’t the ones that never get targeted. They’re the ones who detect fast, contain automatically, and can prove exactly what happened.

This protection is now included in our Go Secured | Advanced Cloud service and in Go Managed | Comprehensive — no separate purchase, no bolt-on. If Microsoft 365 is where your business runs, this is the layer that watches what happens after the login.

Where to Start

  • Worried something may already be going on — an odd login alert, an email that didn’t sit right, a payment that went to the wrong place? Our free, no-obligation Microsoft 365 Compromise Assessment looks for the specific signs of a takeover and tells you plainly what we find.
  • Just want to know where you stand? Our free Microsoft 365 Security Assessment gives you a clear, prioritized picture of your environment, in plain language.
  • Ready to have this running quietly in the background? Let’s talk about Go Secured Advanced Cloud.

You can’t stop every attacker from trying. You can decide whether they get to operate in the dark. Go Boldly — we’ll guide the way.

Multi-factor authentication was supposed to end account takeover. For years, the advice was simple: turn on MFA and attackers move on to easier targets. That advice is now out of date.

Over the past year, attackers have shifted to techniques that don’t break MFA — they walk right around it. And the uncomfortable truth is that even a business with excellent security can still lose money to these attacks, because the weak link may not be your network at all. It may be a supplier’s.

Here’s what’s actually happening, and what to do about it.

How Attackers Get Past MFA Today

Modern account compromise rarely involves guessing a password. Three tactics dominate right now:

Adversary-in-the-Middle (AiTM) phishing. The victim clicks a link and lands on what looks exactly like the Microsoft sign-in page. It is the real login — the attacker is simply relaying it through their own server in the middle. The user types their password, completes MFA on their real phone, and everything works. But in that moment, the attacker captures the session token Microsoft issues after login — the digital “pass” that proves you’re already authenticated. With that token, they don’t need your password or your MFA again. Microsoft observed a single AiTM campaign hit 35,000 users in three days, and one incident-response firm found MFA failed to stop the attack in 84% of the cases it investigated.

Device-code phishing. This one abuses a legitimate Microsoft feature. The attacker sends what looks like a routine prompt (“enter this code to finish signing in”). The victim enters the code on the genuine Microsoft page and approves it with their own MFA. Because every step happens on Microsoft’s real infrastructure, nothing looks wrong — but the approval hands the attacker a valid access token. In May 2026 the FBI issued a public warning about Kali365, a subscription “phishing-as-a-service” kit built specifically to run this attack at scale using AI-generated lures.

Token theft and quiet persistence. Both methods produce the same result: the attacker gets access and refresh tokens rather than a password. That means they can stay in the mailbox for as long as the token is valid — often without tripping a single alert — reading email, learning your billing cycles, and studying who pays whom.

The common thread: MFA verifies you once, at the front door. Once an attacker holds a valid session, the door stays open behind them.

What We Can Do About It

The good news: this is defensible. Go West IT works with clients on projects to harden Microsoft 365 against exactly these techniques — phishing-resistant sign-in methods, Conditional Access policies that restrict risky flows like device-code authentication, and token protection and session controls that limit what a stolen session can do. And for clients on our Go Secured Advanced Cloud service, we provide proactive 24/7 monitoring that watches for the tell-tale sign of a stolen session: a familiar account suddenly signing in from an unfamiliar place or device. Together, these measures dramatically reduce the odds that your accounts become the entry point.

But here’s the part that matters most, and the reason we’re writing this.

The Gap You Can’t Close With Technology Alone

You can do everything right and still lose money — because the compromised mailbox doesn’t have to be yours.

Picture a supplier you pay every month. Their mailbox gets taken over using one of the tactics above. The attacker doesn’t send you a clumsy fake. They sit quietly inside the supplier’s real inbox, read the real invoice thread, and wait. When the genuine invoice goes out, they reply from the supplier’s actual address, inside the real conversation, with one change: new bank details, payment due Friday.

Every check your team knows to make passes. Right sender. Right thread. Right invoice number. Right amount. Nothing in your security stack fires, because nothing in your environment was breached. The money goes out — and weeks later the real supplier calls asking why they haven’t been paid.

No firewall, no MFA, and no monitoring on your side can catch this, because the fraud rides in on a legitimate relationship you’ve trusted for years. This isn’t an IT gap. It’s an operations gap — and the fix is operational.

  • Verify every change of bank or wire details by phone, using a number from your own records — never a number or link from the email itself.
  • Make that callback a written, mandatory step in your payment process, so skipping it is a policy breach, not a judgment call. Apply it to every vendor, including the ones you trust most.
  • Ask your suppliers to do the same in reverse. Your mailbox is somebody else’s supplier thread.

Strong technology and strong procedure aren’t alternatives — you need both. We’ll help you lock down the technology. But the callback is the control that protects you when the breach happens somewhere you can’t see.

That’s the confident way to operate on today’s digital frontier: verify the change, then pay without hesitation.

Want to know where your Microsoft 365 environment — and your payment procedures — stand today? Our free Microsoft 365 Security Assessment gives you a clear, prioritized picture in plain language. Or if something already doesn’t sit right, let’s talk.

AI Adoption Is Change Management: Where Are We Going, and Why?

Most conversations about AI in business start with tools. Which model. Which license. Which app to try first. Those are fair questions, and they have real answers. But they are not the questions that determine whether AI actually takes hold in your organization — or quietly stalls after the initial excitement fades.

The harder, more important work is human. Adopting AI is one of the most significant change-management efforts a business will take on this decade. And change management, in the end, has never really been about technology. It has always been about people.

AI Changes the Work, Not Just the Toolset

When a new application shows up, people learn a few new buttons and move on. AI is different. It changes how work gets done — how a proposal gets drafted, how research happens, how a report comes together, how a decision gets informed. That kind of change reaches into how people see their own value and their own role.

Left unaddressed, that uncertainty shows up in predictable ways. Some people hesitate, waiting to be told it’s safe. Others quietly adopt tools on their own, outside any policy or oversight — the “shadow AI” problem that creates real exposure. Either way, the organization ends up with adoption that is uneven, unmanaged, and far below its potential.

None of that is a technology problem. It is a change-management problem. And it responds to the same things every successful change always has: intention, communication, and direction.

Lead With the Destination — and the Reason

The single most important thing a leader can do is tell people where the organization is going, and why.

The “where” gives people a destination to move toward instead of a tool to be handed. The “why” is what actually moves them. People adopt change when they understand the purpose behind it — when they can see how it serves clients better, removes friction from their day, or lets them spend their time on work that matters more. Without that reason, even the best tool becomes one more thing on the to-do list.

This is where a lot of AI rollouts go quiet. Leadership announces a license, sends a link, and expects momentum to follow. It rarely does. Direction and reason are what turn a tool into a shared goal.

What Intentional AI Adoption Looks Like

The businesses getting real value from AI tend to do a handful of things deliberately:

  • They set a clear direction. Not “everyone go experiment,” but a stated purpose: here is what we’re trying to improve, and here is why it matters to us and our clients.
  • They communicate early and often. People hear the reasoning directly from leadership, not through rumor. Questions and concerns are invited, not managed around.
  • They give people permission and a path. Clear guidance on what is encouraged, what is off-limits, and where to go for help removes the hesitation that stalls adoption.
  • They create champions. Early adopters who model good use and share what’s working carry the change further than any all-hands announcement.
  • They pair enablement with guardrails. A short, plain-language AI policy and basic training aren’t obstacles to adoption — they’re what let people move confidently, knowing the boundaries are clear.
  • They measure and celebrate progress. Naming real wins — hours saved, work improved — reinforces the direction and shows people the destination is worth moving toward.

Notice that most of this is communication, not configuration. The technology is the easy part. Bringing people along is the work.

The Guardrails Are Part of the Empowerment

It’s tempting to treat governance as the brakes on AI adoption — the thing that slows people down. In practice, it’s the opposite. When people know what data is safe to use, which tools are approved, and where the lines are, they stop hesitating and start using AI with confidence. Clear boundaries don’t restrict good work; they free people to do it.

This is why the people side and the security side of AI belong in the same conversation. Empowering your team and protecting your business aren’t competing priorities. Done well, they reinforce each other.

Moving Forward, Together

We’ve lived this ourselves. Go West IT builds its own tools, adopts AI across the firm, and has worked through the same questions we’re describing — how to point a team in a direction, how to explain the why, how to make adoption intentional rather than accidental. That experience is part of what we bring to the businesses we guide.

If your organization is standing at the edge of broader AI adoption, the most valuable thing you can do isn’t to pick the perfect tool. It’s to decide where you’re going, get clear on why, and bring your people with you.

That’s change management. It’s also leadership. And it’s the difference between AI that transforms how your business works and AI that never quite gets off the ground.

If you’d like a steady hand as you think through what intentional AI adoption looks like for your team, we’re here. Go Boldly — we’ll help guide the way.

The One Thing You Can Control: Operational Security in an Age of AI Uncertainty

The AI era has introduced a lot of unknowns. Threats evolve faster than they used to. Attackers now use the same AI capabilities that businesses do — to find weaknesses, craft convincing lures, and move quickly once they’re in. The landscape shifts month to month, and the honest truth is that no one can predict exactly what the next threat will look like.

For business leaders, that creates a genuine dilemma. How do you manage a risk you can’t fully see? How do you invest confidently when the ground keeps moving?

Here is the reassuring part. Amid all that uncertainty, there is one constant — and it’s entirely within your control.

Fundamentals Don’t Depend on Predicting the Future

Operational security is the set of foundational practices that reduce your exposure regardless of what specific threat comes next. It isn’t glamorous, and it doesn’t make headlines. But it is the difference between an organization that is genuinely hard to compromise and one that simply hasn’t been targeted yet.

The reason operational security matters so much right now is precisely because the future is uncertain. You don’t have to know which vulnerability an attacker will use if the vulnerability was already patched. You don’t have to anticipate a stolen password if that account required multi-factor authentication and was being monitored for anomalies. Strong fundamentals protect you against threats you can name today and threats that don’t exist yet.

That is a rare thing in security: work you can do now that pays off against a future you can’t predict.

The Fundamentals Worth Acting On

Operational security isn’t a single product. It’s a handful of disciplines that work together. None of them require you to guess what’s coming.

  • Device management. You can’t protect what you can’t see. Knowing every device connected to your business — and being able to configure, patch, and control it — is the foundation everything else rests on.
  • Vulnerability management. New weaknesses surface constantly. Routine scanning finds what’s exposed in your environment so you can fix it before someone else finds it first.
  • Identity protection and monitoring. Identity is the modern perimeter. Multi-factor authentication, conditional access, and active monitoring for unusual sign-ins stop the majority of account-based attacks — the most common way businesses get breached.
  • Permission controls. People should have access to what they need and nothing more. Least-privilege access limits how far any single compromised account or mistake can reach.
  • Credential management. A business password manager, enforced MFA, and the elimination of shared or generic accounts close one of the most reliable doors attackers rely on.
  • Gateway and network-layer protections. A well-configured firewall, network segmentation, and modern secure-access controls keep threats out at the edge and contain them if they get in.

Individually, each of these is straightforward. Together, they form a security posture that holds up whether the threat is a decade-old technique or something AI surfaced last week.

Certainty You Can Buy in an Uncertain Time

There’s a natural temptation, in a fast-moving threat landscape, to wait for clarity — to hold off until the picture settles or the next tool arrives. But the picture isn’t going to settle, and no single tool is the answer. The organizations that stay ahead aren’t the ones with a crystal ball. They’re the ones that did the foundational work early and kept it current.

That’s the opportunity hiding inside all this uncertainty. While much of the AI-era threat landscape is genuinely unpredictable, your operational security is not. It’s assessable, achievable, and durable. It’s the part of your risk you can actually take off the table.

A Steady Partner for the Work

Operational security is ongoing, not a one-time project — devices change, people come and go, new vulnerabilities appear, and configurations drift. Keeping all of it current, all the time, is exactly the kind of disciplined, unglamorous work that’s easy to let slide when you’re focused on running your business.

That’s the work we do. Go West IT manages device fleets, vulnerability scanning, identity protection, access controls, credential management, and network defenses across hundreds of businesses — so the fundamentals stay strong without becoming a distraction from the work that moves your organization forward.

You can’t control what the AI-era threat landscape does next. You can control whether your operational security is ready for it. If you’d like to know where your organization stands — and where the meaningful gaps are — we’re here to help you see it clearly and move forward with confidence.

Go Boldly. We’ll guide the way.

There’s an old saying that it’s better to be lucky than good. When it comes to protecting your business from cyber threats, that’s a gamble worth retiring.

We prefer a different version: the harder you work, the luckier you get. In cybersecurity, the organisations that look lucky when a major threat makes the news are almost always the ones that invested in operational security best practices long before the headline appeared; unglamorous, consistent work that quietly removes risk before any alert is published.

How AI Has Changed the Cybersecurity Threat Landscape

AI has fundamentally changed the economics of a cyberattack. For years, the raw materials for a breach – old leaked credentials, previously disclosed vulnerabilities, forgotten accounts, unpatched devices – have been sitting in the open. What’s changed is that attackers can now use AI to comb through years of those exposures and test them against live networks at a speed and scale that simply wasn’t practical before.

The FortiBleed activity is a recent example. It wasn’t a single, novel zero-day. It was a systematic recycling of vulnerabilities and credential leaks that were already known and already disclosed – assembled by AI into a working attack against internet-accessible firewalls and VPN gateways. The materials were old. The method was new.

That pattern isn’t unique to one vendor or one product. It’s the shape of the modern threat landscape: the cost of patience has dropped to near zero for attackers, which means the margin for “we’ll get to it eventually” has narrowed considerably for everyone else.

Lucky or Prepared? The Real Difference in Cybersecurity

Here’s the encouraging part. For organisations that had been practising sound operational security, an event like FortiBleed wasn’t an emergency; it was a normal course of business. Review the controls, confirm everything was in order, and move on.

That isn’t luck. It’s the compounding return on consistent work. The teams that hadn’t done the work spent that same week in a fire drill. Same threat, very different week.

The difference comes down to a handful of disciplines that, done routinely, quietly remove most of the risk before any alert is published.

Operational Security Best Practices: The Fundamentals That Work

Operational security isn’t a product you buy. It’s a set of habits you keep. The ones that matter most:

Routine firmware updates.

The devices that run your network – firewalls, switches, access points – need firmware maintained on a schedule, not whenever someone remembers. Outdated firmware is one of the most common ways old vulnerabilities stay exploitable long after a fix exists. If your team discovered FortiBleed through the news rather than through a patch notification, that’s a gap worth closing.

Managed software patching.

Patching should run on a defined cadence and be verified, not assumed. “The update was available” and “the update is installed and the system rebooted” are very different things, and only one of them protects you.

Strong credential management.

Most modern attacks don’t break in, they log in. A business-wide password manager, enforced multifactor authentication (MFA) on every account that matters, and the elimination of generic or default logins close off the easiest path an attacker has. Credentials are the new perimeter; treat them that way.

A minimised attack surface.

Management interfaces shouldn’t be reachable from the public internet. Unnecessary accounts and services should be turned off. Every door you don’t need is one you don’t have to defend.

Routine review of controls.

Security configurations drift over time as people change roles, projects launch, and systems are added. A regular review – confirming that the controls you put in place are still in place and still working – is what keeps a strong posture from quietly eroding.

None of these are exotic. That’s precisely the point. They reward consistency, not cleverness.

Why Operational Security Is Your Best Defence Against AI-Powered Attacks

In an environment where attackers move at machine speed, disciplined operational security may be your single most effective defence – paired with AI-assisted detection and response that can spot and contain trouble at the same pace threats are moving. One side reduces how much can go wrong; the other shortens how long it takes to catch what does.

But tools and tactics only go so far without one more ingredient: treating IT security as a core business competency, with genuine leadership support behind it. Threats like FortiBleed don’t reward the organisations that bought the most software. They reward the ones that operate well, every day – and that takes leadership deciding security is worth doing properly.

Strengthening Your Operational Security With Go West IT

Insisting that security be a core competency doesn’t mean building all of it in-house. The execution and the guidance can be entrusted to a partner who does this work every day – keeping firmware and software current, managing credentials and access, hardening your platforms, and reviewing your controls so nothing drifts.

That’s where Go West IT comes in: helping you put the right fundamentals in place, keep them running, and turn the next industry-wide scramble into a routine exercise.

The harder you work on the fundamentals, the luckier you’ll look when the next threat comes around. We’d be glad to help you do that work.

If you’d like to review where your organisation stands on operational security, we’re here.

Further reading: CISA, “CISA Urges Hardening Fortinet Devices After Reports of Credential Exposure” (June 2026)

Frequently Asked Questions About Operational Security

What is operational security in IT?

Operational security (OpSec) in an IT context refers to the ongoing discipline of maintaining controls that protect a business’s systems, data, and network. This includes routine practices such as firmware and software patching, credential management, MFA enforcement, minimising the attack surface, and regularly auditing that existing controls are still working. Unlike a one-time security audit, operational security is a set of continuous habits – the consistent work that reduces risk before a threat materialises.

What does the FortiBleed vulnerability mean for my business?

FortiBleed refers to a wave of attacks exploiting previously disclosed vulnerabilities in Fortinet devices – particularly firewalls and VPN gateways – combined with leaked credentials. Attackers used AI to systematically test known exposures against live networks. If your business uses Fortinet or similar edge devices and has not applied recent firmware updates or rotated credentials following prior disclosures, it is worth reviewing your posture. CISA published guidance in June 2026 specifically urging organisations to harden Fortinet devices and review credential exposure.

What is the single most important step I can take to improve my business’s cybersecurity right now?

If you’re starting from scratch, enforcing multifactor authentication (MFA) across all key accounts – email, remote access, admin portals – delivers the most immediate reduction in risk. Most successful breaches involve compromised credentials; MFA closes off that path even when a password is known. From there, a structured firmware and patching schedule, combined with a regular review of who has access to what, will address the majority of the remaining exposure.

Once a business starts using paid AI models like OpenAI or Anthropic’s Claude, something predictable happens. Someone – often not a developer by title – builds something. An app. An agent. A workflow that automates a task the team used to dread. And it works. It brings real, measurable value.

That’s the moment the ground shifts. Because in the age of capable AI, we’re all developers now – and AI app hosting is no longer a question just for technical teams. Deciding where your AI-built applications live, who controls them, and how they’re kept secure is a business decision. Getting the answer right early makes everything easier.

Why AI App Momentum Doesn’t Stop at One

Our experience at Go West IT is that once a business sees what’s possible, this doesn’t stay a one-time event. The first useful app inspires the second. A colleague sees it and builds their own. Within a few months, what started as one person’s experiment has become a small portfolio of tools the business is starting to rely on.

That’s really exciting, and it’s also where the important questions begin.

The Right Questions to Ask Before You Host AI Applications

As your team builds more, four questions quickly matter:

  • Where are these apps hosted and who is responsible for keeping them running?
  • Where does the code live, and who controls it if a key person leaves?
  • How do you manage access, is this internal only, or will clients interact with it?
  • How do you ship updates as the app evolves and improves?

These aren’t roadblocks. They’re the sign that something experimental is becoming something real. The businesses that answer them early turn a pile of promising experiments into durable, supportable assets.

How We Host AI-Built Apps at Go West IT: Our Azure Setup

We’re already well down this path internally, and we made a deliberate decision from the start: rather than scatter applications across personal accounts and consumer-grade tools, we’d build on a secure environment we already manage and trust.

In practice, that means:

Hosting in Azure. Our applications run in a managed cloud environment with redundancy, backups, and room to grow – not on someone’s laptop or a free-tier account that disappears when they change roles.

Identity and access through Microsoft Entra ID. Authentication and access management are handled consistently across our applications, so we always know who can reach what – whether the app is internal-only or touches client data.

Code managed in Azure DevOps. Our repositories live there, and we push code from DevOps straight to the application in Azure. That gives us version control, a documented change history, and a clean, repeatable way to ship updates.

The payoff is that as more applications surface – and they will – we’re not reinventing the approach each time. We have a repeatable, supportable process that protects three things at once: operational efficiency (we ship and maintain without chaos), intellectual property (the code your team creates is a business asset, governed like one), and operational security (access, hosting, and change management are controlled by design, not by accident).

How Much Does AI App Hosting Cost? Azure Pricing Explained

One of the most reassuring parts of this conversation is that the costs are knowable and modest relative to the value.

A basic application can run on an Azure Static Web App for approximately $9 per month. A more robust application – one that needs to run in a Linux container, for example – typically lands in the $100 to $200 per month range. Either way, you’re working with predictable monthly recurring costs you can plan around.

That predictability makes return on investment easy to evaluate. It also opens the door to a new metric worth tracking: Return on Tokens – how much did you spend, in AI usage and supporting infrastructure, to build and run that app, and is the value it returns worth it? That’s the difference between ‘we think this AI thing is helping’ and ‘we know exactly what this costs and what it returns.’

Building an AI App Hosting Strategy That Scales

If deciding whether your business is AI-ready was the first step, this is the natural next one: giving the things your team builds a secure, permanent, well-managed home – and a strategy that scales as the next app, and the one after that, arrives.

This is the path we’ve walked ourselves, which is exactly why we can guide you down it. Go West IT can help you stand up a secure hosting and development foundation – Azure for hosting, Entra for identity, Azure DevOps for code, so the value your team is creating with AI becomes something lasting, protected, and supportable.

You’re already building. Let’s make sure what you build has somewhere solid to stand.

If you’d like to talk through a hosting and development strategy for your AI-built applications, we’re here.

Frequently Asked Questions About AI App Hosting

What is the cheapest way to host an AI-built application?

A basic application can run on an Azure Static Web App for approximately $9 per month. This is suitable for lighter tools; internal dashboards, simple agents, or workflow automations that don’t require a dedicated server. More complex applications, such as those that need to run in a Linux container, typically cost $100 to $200 per month. Both options offer predictable monthly costs and enterprise-grade reliability.

What security considerations apply to AI app hosting?

Key areas to address include identity and access management (Microsoft Entra ID handles authentication and controls who can reach each application), code version control (Azure DevOps ensures a documented, repeatable deployment process), and data governance (knowing whether client data or internal data flows through each application and how it is stored and protected). Hosting on a managed cloud platform like Azure also provides redundancy and backup capabilities that consumer-grade tools do not.

What is ‘Return on Tokens’ in AI app development?

Return on Tokens is a metric for evaluating the value of an AI-built application against its actual running costs – including AI model usage fees and hosting infrastructure. It helps businesses move from “we think this is helping” to “we know exactly what this costs and what it returns,” making it easier to prioritize which applications are worth developing and maintaining.

The owner of this website has made a commitment to accessibility and inclusion, please report any problems that you encounter using the contact form on this website. This site uses the WP ADA Compliance Check plugin to enhance accessibility.