How long does it take to upgrade a .NET application from .NET Framework?

It depends on size and on how much Web Forms or WCF code you have. Class libraries and MVC apps move fastest, while Web Forms pages need rebuilding. An audit with the .NET Upgrade Assistant gives a realistic plan, and incremental migration with YARP keeps the product live throughout.

Do we have to move to Azure to use Semantic Kernel?

No. Semantic Kernel runs anywhere modern .NET runs and supports several model providers, including OpenAI. Azure OpenAI and Azure AI Search are a natural fit for companies already on Microsoft cloud because of private networking, regional hosting and managed identity, but they are not required.

Will our data be used to train AI models?

With Azure OpenAI and other enterprise endpoints, your prompts and data are not used to train the provider models under their standard terms. We still mask personal data where possible, keep data in your chosen region, and enforce your existing access rules on every AI request.

Can you fix EF Core performance without changing the database?

Often, yes. Many slow pages come from how queries are written: loading too many columns, lazy loading in loops or missing AsNoTracking. We fix those first, then add indexes if needed. Schema changes are proposed only when they bring a clear improvement.

Upgrade .NET and Add AI With Semantic Kernel

Plenty of companies still run important products on ASP.NET Web Forms, ASP.NET MVC or WCF services built on .NET Framework. They work, but they are harder to host, slower to change and locked to Windows servers. If that sounds familiar, the goal is not a risky rewrite. The goal is to upgrade your .NET application in planned steps, fix what hurts today, and then add AI features that save your team time. This guide explains how Ecomsolver does that.

We follow the same three-part approach on every .NET engagement: audit and fix, fine-tune, then add AI and automation. Each part produces something you can see and measure, and each one makes the next safer.

Where Existing .NET Apps Usually Struggle

Framework age: .NET Framework apps, Web Forms pages with heavy ViewState, and WCF services that modern clients struggle to call.
Windows-only hosting: IIS servers that are patched by hand and difficult to scale.
Slow data access: Entity Framework queries that load whole tables, lazy loading that triggers hundreds of queries per page, and stored procedures nobody wants to touch.
Security gaps: outdated NuGet packages, old authentication code, secrets in web.config, and missing HTTPS or CSP headers.
Manual releases: deployments copied over remote desktop with no automated tests.
Integration failures: SOAP calls to ERP or payment providers that fail silently.

Step 1: Audit the Solution

We start with a structured audit of the solution, the database and the hosting environment.

Upgrade readiness: the .NET Upgrade Assistant and API compatibility analysis show which projects can move easily and which use APIs that no longer exist, such as System.Web, Web Forms or AppDomains.
Dependencies: every NuGet package checked for vulnerabilities, support status and a modern replacement.
Performance: profiling with dotnet-trace, dotnet-counters, Visual Studio diagnostics and SQL Server Query Store to find slow pages, queries and memory leaks.
Security: authentication flows, authorization rules, secrets handling and OWASP Top 10 risks.
Architecture and hosting: what runs where, how services talk, and what a move to Azure App Service, Azure Container Apps or AKS would involve.

You receive a ranked list of findings, an upgrade path for each project, and a written estimate.

Step 2: Quick Wins

Patch vulnerable packages and move secrets into Azure Key Vault or user secrets.
Fix the slowest EF queries with projections, eager loading where it helps, AsNoTracking for read-only queries and missing indexes.
Add Polly retry and circuit breaker policies to outbound HTTP calls.
Add response caching and Redis output caching to heavy read pages.
Move long-running tasks to Hangfire or a hosted background service.

Step 3: Migrate .NET Framework to Modern .NET

Modern .NET is cross-platform, faster, and runs well in Linux containers, which often lowers hosting cost and simplifies scaling. The migration path depends on what you have.

Class libraries and business logic

Shared libraries usually move first. We retarget them to .NET Standard or modern .NET so both old and new apps can use them during the transition.

ASP.NET MVC and Web API

These map fairly directly to ASP.NET Core MVC and minimal APIs. Controllers, routing, filters and dependency injection need changes, but the business logic usually survives.

Web Forms and WCF

Web Forms has no direct equivalent, so pages are rebuilt in Razor Pages, Blazor or a React or Angular front end. WCF services move to ASP.NET Core Web API, gRPC or CoreWCF when existing clients must keep working.

Incremental migration

For large apps we use the strangler fig pattern with YARP, a reverse proxy that sends some routes to the new ASP.NET Core app and the rest to the old one. Users see one product while pages move over one by one. Our ASP.NET development team has handled both incremental and full migrations.

Entity Framework to EF Core

EF Core is faster, but behaviors differ: lazy loading is opt-in, some LINQ queries translate differently, and client evaluation now throws instead of silently loading data. We cover these with integration tests before switching.

Step 4: Fine-Tune Performance and Delivery

EF Core and SQL

We use compiled queries for hot paths, split queries for large includes, bulk operations for imports, and Dapper where raw SQL is clearer. Query Store and Application Insights show which queries matter most.

Testing

xUnit or NUnit for business rules, WebApplicationFactory for API integration tests, and Testcontainers for SQL Server or PostgreSQL in CI. Playwright covers critical user journeys in the browser.

CI/CD and observability

GitHub Actions or Azure DevOps pipelines build, test, scan and deploy to Azure with slots or blue-green releases. OpenTelemetry and Application Insights give you traces, logs and metrics in one place, with alerts on error rates and slow requests. Health checks and structured logging with Serilog make problems easier to diagnose.

Step 5: Add AI With Semantic Kernel

Semantic Kernel is Microsoft’s open-source SDK for adding AI to .NET applications. It connects to Azure OpenAI, OpenAI and other providers, and lets you expose your own C# methods as plugins the model can call. It works with Microsoft.Extensions.AI abstractions, dependency injection, logging and configuration, so AI code looks like the rest of your ASP.NET Core app.

Core building blocks

Azure OpenAI: GPT and embedding models deployed in your Azure subscription and region, with enterprise data terms, private endpoints and managed identity.
Azure AI Search: keyword, vector and hybrid search with semantic ranking, plus security filters so users only see documents they are allowed to see.
Azure AI Document Intelligence: extraction of text, tables and fields from invoices, receipts, forms and scanned PDFs.
Plugins and function calling: C# methods marked as kernel functions so the model can look up orders, create tickets or check stock through your existing services.

Example AI features for .NET products

Internal knowledge assistant: a chat over policies, manuals and SharePoint content, using Azure AI Search for retrieval and citations in every answer. Our RAG development services cover the ingestion, chunking and evaluation side.
Document intake automation: uploaded invoices or claims parsed by Document Intelligence, checked by the model against business rules, and queued for approval.
Customer support copilot: suggested replies and case summaries inside your existing support screens, with agents approving every message.
Natural language reporting: users ask questions in plain English and the system maps them to safe, parameterized queries or predefined reports rather than free-form SQL.
Workflow automation: Azure Functions or background services that classify emails, route requests and update your CRM or ERP.

How a Semantic Kernel feature is built

Register the kernel, chat completion and embedding services in Program.cs, using managed identity to reach Azure OpenAI instead of API keys.
Build an ingestion job, often an Azure Function or hosted service, that extracts text, splits it into chunks, creates embeddings and writes them to an Azure AI Search index with metadata such as department, customer and permission groups.
Expose the business actions the assistant may take as plugins: small, well-named C# methods with typed parameters and their own authorization checks.
Write prompt templates that tell the model to answer only from retrieved content, cite sources and return a defined JSON shape.
Stream responses to the browser through SignalR or server-sent events so users see answers as they are generated.
Log every request with prompt version, model, tokens used and latency to Application Insights for cost tracking and debugging.

We keep AI logic in its own project with a clear interface, versioned prompt templates and one place for timeouts, retries and token tracking. ThinkEduAI, Ecomsolver’s own AI school-management SaaS, uses the same separation so AI features can change without destabilizing the core product. For broader AI work, see our AI integration services.

Risks to Plan For

Data privacy and compliance

Keep data in your Azure tenant and region, use private networking and managed identity, mask personal data in prompts, and apply the same authorization rules to retrieval that you apply to screens.

Cost control

Track tokens per feature, set quotas in Azure, cache common answers, and use smaller models for classification. Review usage every month alongside hosting costs.

Hallucinations and safety

Ground answers in search results, require citations, return structured JSON validated against C# types, and use Azure AI Content Safety filters. High-impact actions always need human approval.

Evaluation

Maintain a set of real test questions with expected answers, run it in the pipeline when prompts or models change, and sample production conversations regularly.

Checklist to Upgrade Your .NET Application

Upgrade readiness report for every project in the solution.
Vulnerable NuGet packages patched and secrets moved to Key Vault.
Top slow pages and queries profiled and fixed.
Migration path chosen: incremental with YARP or full rebuild.
Integration and browser tests around critical flows.
Automated pipeline with deployment slots and rollback.
Application Insights or OpenTelemetry dashboards and alerts.
AI use cases ranked, data rules agreed, budgets and evaluation in place.

Working With Ecomsolver

Ecomsolver brings more than seven years of software delivery from Jaipur, India, with a US office in Austin, Texas, and time-zone overlap for clients in the USA, UK, Europe and Australia. Every engagement includes a free consultation, a written fixed quote, an NDA, milestone payments, weekly demos and 100% code ownership. Prefer to grow your own team? Hire an ASP.NET developer who works inside your process.

Ready to upgrade your .NET app or add AI to it? Book a free audit and consultation and get a prioritized plan with a written quote.