Skip to content
Devsoft

Article

Zero Trust for Carolinas businesses on the Microsoft stack: what AI threats are changing

AI is making phishing, credential theft, and lateral movement faster and harder to detect. Here is what Zero Trust looks like in practice on the Microsoft stack for North and South Carolina businesses in 2026.

By Devsoft Solutions

Zero Trust started as a network security concept: stop assuming that anything inside your perimeter is safe. A decade ago, implementing it meant expensive third-party tooling and a multi-year architecture project most mid-market businesses could not justify. In 2026, the calculus has shifted on both sides of that equation. The threats are worse, and the tools to address them are already inside most Microsoft 365 subscriptions.

We work with businesses across North and South Carolina on Microsoft 365 and Azure security. The Zero Trust conversation is no longer optional for most of them. Here is what is driving that shift, and what Zero Trust actually looks like on the Microsoft stack for a 50 to 500 person Carolina business.

Why AI has made the old perimeter model untenable

The perimeter model assumed attackers would have to do hard things: find a vulnerability, get past the firewall, brute-force a password. AI has made several of those hard things cheap and scalable.

Phishing at scale and at quality. Phishing emails used to be identifiable by tone, grammar, and generic greetings. AI-generated phishing can impersonate the CFO’s writing style, reference last week’s board meeting (scraped from LinkedIn), and target a specific individual with specific context. The volume of high-quality targeted attacks has increased sharply in the last eighteen months. Charlotte financial services firms, Research Triangle biotech companies, and Greenville manufacturers are all seeing this in their security logs.

Credential stuffing, faster. AI accelerates the speed at which stolen credential lists from unrelated breaches get tested against Microsoft 365 and Azure portals. A list of 100,000 stolen username and password combinations from a retail breach gets tested against your tenant in minutes. If your employees reuse passwords, this works.

Social engineering that is harder to detect. AI-generated voice and video can simulate a known contact requesting an urgent funds transfer or credential reset. The executive impersonation attacks that used to require a skilled human actor now run with less skill and at higher volume.

The result is that the old model of “trust everyone inside the network, block everything outside” has collapsed. Remote work finished the network perimeter as a meaningful security boundary. AI finished the belief that a strong password and perimeter firewall constitute adequate security.

Zero Trust replaces the perimeter assumption with a different one: trust nothing by default, verify every access request explicitly, and limit the damage when something does get through.

What Zero Trust means on the Microsoft stack

Zero Trust is a framework, not a product. On the Microsoft stack, it maps to a set of specific controls across identity, devices, and data. Most of these controls are available in Microsoft 365 Business Premium, E3, and E5. The question is whether they are configured.

Identity: Entra ID and Conditional Access

The identity plane is where most Zero Trust implementations start and where the return is highest per hour of work.

Entra ID (formerly Azure Active Directory) is the identity provider for the Microsoft stack. Conditional Access is the policy engine that sits in front of it. Together they enforce the Zero Trust principle of “verify explicitly” at every sign-in.

The baseline configuration:

  • Multifactor authentication for every user on every sign-in, with no exceptions for location or device. Password plus second factor is the floor. Phishing-resistant authentication (FIDO2 keys, certificate-based authentication) is the target for privileged accounts and high-risk roles.
  • Block legacy authentication protocols (IMAP, POP, SMTP AUTH, older Outlook clients) that cannot satisfy MFA. These protocols are the entry point for the majority of credential-based attacks against Microsoft 365 tenants.
  • Entra ID Identity Protection (available in P2 licenses, included in E5 and Business Premium) to score every sign-in for risk. Impossible travel, anonymous IP, atypical location, leaked credentials. High-risk sign-ins force a password change and MFA before the session continues.
  • Privileged Identity Management to put admin roles in time-limited elevation rather than permanent assignment. An administrator should not be a Global Admin at 2 AM on a weekend by default.

For Carolinas businesses in regulated industries, Charlotte banking and insurance especially, the Conditional Access audit log becomes a material artifact for regulatory examinations. Every policy decision is logged. Auditors can see what was enforced, when, and for whom.

Devices: Intune and Microsoft Defender for Endpoint

Zero Trust requires knowing the state of the device that is making the access request. An Entra ID sign-in from a corporate-managed, fully patched device is a different risk profile than the same sign-in from an unmanaged personal laptop.

Intune is the device management plane. The Zero Trust configuration:

  • Enroll all corporate devices (Windows, macOS, iOS, Android) in Intune and define a compliance baseline. What constitutes a compliant device: minimum OS version, disk encryption enabled, antivirus up to date, screen lock configured.
  • Use Conditional Access policies that require device compliance for access to sensitive applications. SharePoint, Teams, Exchange, and any line-of-business applications that hold regulated data should require a compliant device.
  • For BYOD scenarios where employees use personal phones for email, use Intune App Protection Policies (MAM without MDM). These enforce encryption and cut-and-paste restrictions on the application level without taking full management of the personal device. This is the right balance for most Carolinas mid-market businesses.

Microsoft Defender for Endpoint (included in Business Premium, E5, and available as an add-on) extends visibility to what is happening on the device itself, not just whether it passed a compliance check at sign-in. It detects malware behavior, lateral movement, and attacker tooling.

For Upstate South Carolina manufacturing firms dealing with OT/IT convergence, the device management boundary is more complex. Production equipment cannot always run Intune. The Zero Trust approach there is network segmentation and tighter controls on the office-side devices that do have access to cloud resources, with monitoring on the network boundary between OT and IT environments.

Data: Microsoft Purview and Information Protection

Verifying identity and device state reduces the risk of unauthorized access. Information protection limits what attackers can do if they get through anyway.

Microsoft Purview (included in E3 and E5, available as an add-on) provides sensitivity labels that travel with documents. A document labeled Confidential is encrypted by the label itself. Even if an attacker exfiltrates the file, they cannot open it without a valid Entra ID credential in the tenant.

For Carolinas businesses in healthcare, the sensitivity label approach maps directly to HIPAA obligations around PHI. For defense contractors in the region working toward CMMC Level 2, the same label infrastructure provides part of the documentation trail for controlled unclassified information handling.

The practical starting point is not a comprehensive labeling taxonomy. It is identifying the two or three document types that would cause the most damage if exfiltrated: client contracts, employee PII, financial records, intellectual property. Label those first, enforce the policy for those document types, and expand from there.

The AI-specific additions in 2026

Beyond the foundational Zero Trust controls above, two AI-specific configurations are worth implementing for Carolinas businesses that are deploying Microsoft 365 Copilot.

Oversharing risk from Copilot. Copilot surfaces content from across the Microsoft 365 tenant based on the signed-in user’s permissions. If SharePoint permissions are overly permissive, Copilot will surface content the user technically has access to but was never intended to see. This is not a Copilot bug. It is a permissions problem that Copilot makes visible. Before deploying Copilot broadly, run a SharePoint permissions audit. The Microsoft 365 admin center and Purview data map can identify sites and libraries with overly broad sharing.

Copilot interaction data governance. Microsoft 365 Copilot interaction logs are stored in Exchange Online and surfaced through Purview eDiscovery and Audit. For regulated industries, these logs are subject to the same retention and litigation hold requirements as email. Establish a retention policy for Copilot interaction data before deployment, not after a regulatory inquiry.

A Zero Trust maturity sequence for Carolina businesses

The full Zero Trust architecture is not a single project. It is a sequence of increasing maturity. A practical sequence for a 100-person Carolinas business:

Phase one (weeks one through four): Identity baseline. MFA for all users. Block legacy auth. Break-glass accounts. Conditional Access report-only mode, then enforce. This stops the majority of credential-based attacks.

Phase two (months two and three): Device management. Enroll corporate devices in Intune. Define compliance policies. Require compliant device for high-risk application access. Defender for Endpoint on managed devices.

Phase three (months three through six): Data governance. Sensitivity labels for the highest-risk document categories. SharePoint permissions audit. Retention policies aligned to compliance requirements. Purview audit log review cadence established.

Phase four (ongoing): Threat monitoring. Microsoft Sentinel or Defender XDR for correlation across logs. Incident response runbooks. Quarterly review of Conditional Access policy effectiveness.

Each phase has measurable security outcomes. Each phase is also a legitimate stopping point if budget or internal capacity requires it. Phase one alone, done well, eliminates the most common attack vectors we see compromising Carolinas Microsoft 365 tenants.

What Zero Trust does not solve

A few things worth naming:

Zero Trust does not replace security awareness training. A well-configured Conditional Access policy still allows a user who has authenticated to click a malicious link or approve a fraudulent payment. Human judgment remains in the chain.

Zero Trust does not fix application vulnerabilities. If a line-of-business web application has SQL injection vulnerabilities, Zero Trust controls at the identity and device layer do not protect the application’s data.

Zero Trust does not eliminate the need for backup and recovery. Ransomware that runs under a legitimately authenticated user session is not stopped by identity controls. Immutable backup, tested recovery, and incident response planning are separate requirements.

What Zero Trust does: it dramatically reduces the attack surface available to an adversary who has stolen a credential or compromised a device. In the AI-accelerated threat environment of 2026, reducing that attack surface is not a best practice. It is the baseline.


Devsoft Solutions works with businesses across North and South Carolina on Microsoft 365 security, Entra ID configuration, and Zero Trust implementation. If you are evaluating your current security posture or building a Zero Trust roadmap, get in touch.