Article
Microsoft Graph is the unified data layer behind every Microsoft 365 service. Carolinas manufacturers, healthcare systems, and professional services firms are building custom apps on top of it to automate specific workflows, surface AI-driven insights, and transform how their operations run.
The AI transformation conversation in the Carolinas has centered on Microsoft 365 Copilot and Azure-based machine learning. Both are real and worth pursuing. But there is a quieter transformation underway that tends to generate more durable returns: businesses building custom line-of-business applications on Microsoft Graph that automate specific workflows, surface data that was previously trapped in separate systems, and increasingly feed into AI-powered decision layers.
We build these applications for manufacturers, healthcare organizations, and professional services firms across North and South Carolina. What follows is the practical picture: what Microsoft Graph is, why it matters for AI specifically, where Carolinas businesses are getting real value from it, and what it takes to build on it well.
Microsoft Graph is the unified API layer that sits across all of Microsoft 365: Exchange Online, Teams, SharePoint, OneDrive, Planner, Calendar, Azure Active Directory, and more. When you make a call to Microsoft Graph, you can read a user’s email, list the files in a SharePoint site, see who attended a Teams meeting, query the org chart, or pull task assignments from Planner. One authentication model. One endpoint. One consistent set of permissions.
For most businesses, the significance of this is not obvious until you try to build something. The moment you realize that the identity, the org structure, the documents, the communications, and the calendar are all accessible through a single API that your existing Microsoft 365 licenses already cover, the cost calculus on custom applications changes considerably.
Before Graph, building a custom approval workflow might have required a database to store user identities, a file storage system, email integration, and calendar logic. All of that infrastructure already exists in your Microsoft 365 tenant. Graph exposes it.
Microsoft 365 Copilot is built on Microsoft Graph. When Copilot summarizes a meeting, drafts a document, or answers a question about a project, it is using Graph to access the underlying content: the meeting recording in Teams, the documents in SharePoint, the emails in Outlook, the tasks in Planner.
When you build custom applications on Graph, you are building on the same data layer that Copilot uses. That alignment has two practical implications:
Graph apps become AI-ready without extra work. Data that your custom app writes into SharePoint, Planner, or Exchange is immediately accessible to Copilot. A work order status that your field technicians update via a custom mobile app appears in the same data layer that Copilot reasons over when a project manager asks about job status.
Graph connectors extend Copilot to external data. If your business runs systems outside Microsoft, such as an ERP, a custom CRM, or a job management platform, Graph connectors let you index that content into the Microsoft 365 semantic index. Copilot can then search and reason over it alongside your native Microsoft content. The custom development work to build a Graph connector is substantially less than building a full integration, and the return is Copilot awareness of your business-critical external data.
Greenville and the Upstate South Carolina manufacturing corridor run on approvals. Production sign-offs, quality hold releases, supplier deviation requests, engineering change notices. These workflows exist in every plant. In most of them, they still run on email chains, shared drives, and institutional memory about who needs to approve what.
Custom Graph applications replace these with structured workflows: a request form that creates a SharePoint item, routes it through an approval chain pulled from Active Directory, sends Teams notifications at each step, records decisions with timestamps and rationale, and feeds a Power BI dashboard that shows the operations manager which approvals are overdue and by how long.
The AI layer arrives when you connect the historical approval data to Azure OpenAI. Patterns that took an experienced quality manager to spot, such as which supplier deviations correlate with rework events, become queries anyone can run in plain language.
North Carolina’s healthcare landscape sits between large academic medical centers in the Triangle and Raleigh-Durham, a growing network of regional health systems, and a significant rural coverage gap in the eastern part of the state. Referral coordination is a persistent operational problem. Patients fall through between systems. Specialists do not receive the right records. Primary care providers do not hear back.
Custom Graph applications for healthcare referral management work within the Microsoft 365 environment that most health systems already have. A structured referral created in SharePoint, with patient context attached and a Teams-based communication thread with the receiving provider, beats an email with a PDF attachment that may or may not reach the right person.
This category requires careful attention to HIPAA and the Microsoft compliance framework. The short version: Microsoft 365 and the Graph API can be configured for HIPAA-eligible workloads, and doing so correctly is a prerequisite before any patient data moves through custom Graph apps. We have written about that configuration separately.
Charlotte’s financial services and professional services firms run a significant volume of billable knowledge work on tight timelines. The data about that work, who is working on what, which deadlines are at risk, which clients are behind on document delivery, is distributed across email, Teams, SharePoint, and individual calendars.
Custom Graph applications for project and client visibility pull from that existing data rather than requiring staff to duplicate entries into a separate project management tool. A client portal that shows the status of a matter or engagement, populated from SharePoint and Planner without manual status updates, eliminates the Friday afternoon email round-robin that consumes project staff time at most professional services firms.
The AI layer on top of this data, surfacing which matters are running over estimated hours, which clients have not responded to document requests, or which team members are at capacity, gives leadership the operational picture they are currently assembling by hand.
When a Carolina business comes to us with a workflow to automate, the first question is not what to build but how to build it. The options are meaningfully different.
Power Apps with Dataverse or SharePoint is the right choice for forms-based workflows that do not require complex business logic, where non-technical staff will need to maintain or extend the app, and where the user base is internal. Power Apps has its own article, but the short version: it is genuinely useful for a large category of internal tools and significantly underutilized in most Microsoft 365 organizations.
Custom .NET with the Microsoft Graph SDK is the right choice when the workflow is complex, the integration surface extends outside Microsoft 365, the performance requirements exceed what Power Apps delivers, or the user experience needs to match a quality bar that a low-code tool cannot reach. Custom .NET apps on Graph are more expensive to build and require ongoing development support. They are also more capable, more maintainable at scale, and easier to extend as requirements change.
Graph connectors for Copilot are the right choice when the goal is not to build a user-facing application at all, but to make external system data searchable and available to Copilot and Microsoft 365 Search. If your business runs Salesforce, ServiceNow, SAP, or a custom ERP, and you want Copilot to answer questions that require data from those systems, a Graph connector is the path.
Most mid-market Carolinas businesses end up with a combination: Power Apps for the simple internal workflows, custom .NET for the mission-critical ones, and Graph connectors to bring in the external data that the organization depends on.
The three phases that determine whether a Graph application delivers its promised return:
Discovery: mapping the workflow before touching code. The most common failure mode in custom app development is building the right technology for the wrong process. Before any development begins, the workflow needs to be documented: who initiates it, what information is captured, what happens at each decision point, who approves, where the exceptions live. At this stage, a surprising percentage of workflows turn out to be simpler than assumed, or to already have a Power Apps solution that does not require custom development.
Permissions architecture: getting this right is non-negotiable. Microsoft Graph operates on a delegated and application permission model. Getting it wrong means the app either cannot access the data it needs or can access data it should not touch. For organizations in regulated industries, this step requires legal and compliance review before development begins, not after.
Integration surface mapping. If the app connects to systems outside Microsoft 365, each of those connections needs to be designed explicitly: authentication, data mapping, error handling, retry logic. In manufacturing environments, this often means integrations to ERP systems that were not designed with modern APIs in mind. The integration work is frequently the longest part of the project.
Development itself, for a well-scoped Graph application, typically runs four to twelve weeks depending on complexity. The range is wide because the scope varies significantly between a straightforward approval workflow and a multi-system operational dashboard with AI components.
Every Graph application generates data. Decisions, timestamps, volumes, patterns over time. For most businesses, that data sits in SharePoint lists and is reviewed occasionally by someone who exports it to Excel.
The AI layer becomes viable when you have enough data to look for patterns, the patterns carry business value if surfaced early, and the people who would act on those patterns are currently finding them too slowly or not at all.
For a manufacturing quality workflow, that might mean flagging supplier deviations that match the profile of past failures before a hold decision is made. For a professional services firm, it might mean identifying which matters are tracking to exceed budget based on early activity patterns. For a healthcare referral system, it might mean surfacing referrals that are approaching the point where patients typically disengage from the process.
The common thread is: the AI is not replacing a human decision. It is surfacing a pattern fast enough that the human can make the decision with more context and less lag. That is a realistic and valuable use of AI in the Microsoft stack.
The businesses that get the most from Graph application development are not the ones that start with the biggest vision. They are the ones that identify a specific workflow that is currently running on email and shared drives, scope a clean solution to that workflow, build it well, and measure what changes.
The second project is usually faster and more ambitious than the first, because the organization now has a working pattern for how to scope, build, and maintain a Graph application. The third project often incorporates an AI layer, because by then there is enough data in the system to make pattern detection worthwhile.
For most Carolinas businesses, the workflow candidates worth starting with are the ones that combine high volume, structured inputs, and a documented approval chain. Quality sign-offs. Budget approvals. Client onboarding. Referral routing. These are the workflows where the improvement is measurable, the business case is clear, and the scope is contained enough to deliver something in a quarter rather than a year.
Devsoft Solutions builds custom Microsoft Graph applications for manufacturers, healthcare organizations, and professional services firms across North and South Carolina. If you have a workflow that is running on email and shared drives and you want to understand what a Graph application would look like for your business, get in touch.
— Continue reading
AI has raised the ceiling for Power Apps and lowered the cost of custom .NET development. For Carolinas businesses deciding where to build internal tools, the old rules of thumb no longer apply. Here is an updated framework.
Legacy .NET Framework line-of-business apps are quietly blocking AI adoption, cloud migration, and developer hiring for businesses across North and South Carolina. Here is what the modernization path actually looks like in 2026, including how AI tooling is compressing the timeline.
AI adoption is accelerating across the Carolinas, and it is creating data privacy exposures that most businesses have not yet mapped. Here is what North Carolina's current legal landscape requires, what is changing at the federal level, and how to keep your AI rollout from becoming a compliance liability.
— Related services
Microsoft Copilot rollout, custom copilots with Copilot Studio, and Azure OpenAI integrations for the parts of your business where AI actually pays back.
Tailored Microsoft 365 plans, deployment, and ongoing support — designed for businesses that need reliable productivity infrastructure without the in-house overhead.
Azure infrastructure, Microsoft 365 environments, and the operational glue between them. Right-sized, secured, and supported.
