AI in existing software

Add AI to the system your team already uses

We add AI features to the application your team already uses: extracting data from documents, organising requests and drafting replies. Employees check the result in a familiar program, instead of reading documents and filling in forms manually. We specialise in .NET applications with Microsoft SQL Server databases.

Let’s talk about your business

30–45 minutes online. Get in touch and we’ll arrange a time.

What AI features can you add to your application?

We extend applications that can be developed further: your own business systems or software with available source code and the rights to modify it. We specialise in .NET and MSSQL. If you use a packaged ERP, we first check its add-ons and integration options.

We can add features that support reading and interpreting content, for example:

  • Filling in forms from documents. Data from a message, PDF or scan goes into a draft form for an employee to check.
  • Organising requests. The system proposes a category and routes the case to the appropriate queue according to agreed rules.
  • Draft replies. The case history and selected data are used to prepare a proposed message for editing and approval.
  • Correspondence summaries. An employee receives a summary of a long history, with access to the sources.
  • Document search. A user asks a question and receives an answer based on material available to them.

We choose one feature to start with and test it using real examples. AI should reduce the total workload, including checking and corrections.

When is it worth adding AI to an existing system?

This is a good option when the application supports the process, holds the required data and is where your team works every day. An employee can run the new feature in the current form and check its result there.

First, we check whether the task needs AI. If data has a fixed format or a decision can be described by a simple rule, an ordinary software feature may be easier to maintain.

If your team mainly needs to ask questions about data from several systems, see an AI assistant on company data. When the current system does not support the core process, we first need to assess extending it.

How do we implement an AI feature in an application?

  1. We understand the task and system. We establish what takes time, which data is available and how the application can be extended. We check the code or interfaces and the deployment process.
  2. We define the output and how to check it. We specify what AI should prepare, what an employee approves and which mistakes matter. We also collect difficult and incomplete examples.
  3. We test the model and cost. We compare quality, response time and correction effort. We agree where data is processed.
  4. We add the feature to the system. We introduce format checks, permission and cost controls, and error handling.
  5. We launch it in stages. First for selected users or case types. We also assess results on examples not used when preparing the solution.

If the scope is clear after our conversation, we prepare a proposal. If the application or feasibility needs examining first, we suggest a separately priced review or test.

Example scenario: service requests

A company receives email service requests in its own application. An employee reads the message, identifies the customer and equipment, and selects a case category.

After AI is added, they see proposed field values and a draft reply in the same form. They check and approve them. Incomplete or ambiguous requests continue to be handled manually.

AI integration with .NET and Microsoft SQL Server

We specialise in .NET and C# applications with Microsoft SQL Server (MSSQL) databases. We have over 20 years of experience designing, programming and implementing business systems. Read more about DO IT GLOBAL’s experience.

We can add an AI feature as an application module or a separate service that exchanges data with it. The application continues to handle permissions, rules and data storage; the model prepares the output for a defined task.

When choosing the technology, we allow for a future change of model or provider. Each such change requires quality to be checked again on your company’s tasks.

AI in the cloud or on a company server?

The feature can use an AI cloud service or a model running locally. The choice depends on the task, data requirements, answer quality and cost of use.

For a cloud service, we check its conditions: data use for training, storage location and duration, and access. If information needs to stay at your company, we assess a local or offline option. We select hardware based on task tests and the expected workload.

How do we control results and keep work running?

We tailor controls to the task and the consequences of a possible mistake:

  • Checking the result. The system verifies required fields and rules, while an employee approves the output where their judgement is needed.
  • Working without AI. We prepare and test a way to continue work when the model or connection is unavailable.
  • Cost control. Limits on data size and call volume help keep spending within the agreed budget.
  • Investigating errors. We log the necessary events, with defined access and retention periods.

Quality is assessed through test results and data checks. A model’s declaration that it is confident in an answer is not enough.

How much does adding AI to an application cost?

The price depends on the feature, system condition, available interfaces and the required output checks. Before work starts, we agree the scope, price and acceptance criteria.

We first review a system built by another supplier to assess the possibility of changes. Read more about taking over application development and how we work.

After launch, you can ask us to provide maintenance and development. If you are an IT company developing an application for a client, see our partnership options.

Frequently asked questions

What does integrating AI with an existing application involve?

Integrating AI with an existing application lets you use AI within that application. For example, the model prepares data extracted from a document or proposes a request category. The application checks the result against its rules and shows it to an employee.

Can you add ChatGPT to a business application?

This is a common way of describing a feature that uses an AI model. In the option described here, the application connects to a chosen service through an API or to a local model. It does not use an employee's personal chat account. We agree the provider, costs and data processing conditions separately from the integration itself.

Does the whole system need to be rewritten to add AI?

A separate module or service connected to the application is often enough. After reviewing the code and available interfaces, we identify the required changes and price the implementation.

Does the system have to be written in .NET?

We specialise in .NET, C# and Microsoft SQL Server. We assess connections to applications in other technologies individually, for example through a separate API service. This requires available interfaces or cooperation with the person developing the system.

Can you add AI to a system built by another supplier?

We can assess this after reviewing the code, documentation and how the application runs. We also check rights to make changes and access to environments. If part of the system needs work first, we explain this before pricing the implementation.

Will AI have access to the whole database?

In the solution we design, the model receives the data needed for a specific task. The application or a dedicated service handles retrieval and writing, with access controls. We define approval rules so that the model cannot independently bypass the system's rules.

How much does adding AI and then using it cost?

We price implementation after defining the feature, application condition and required connections. We estimate model, infrastructure and support costs separately. For a cloud service, call volume and data size matter; for a local model, hardware, workload and maintenance matter. Before implementation, we check the cost using representative examples.

Can AI work without sending data to the cloud?

Yes, if the entire required flow — model, document reading, search and storage — runs locally. The choice depends on the task and quality on your company data. Hosting the application on a company server does not by itself mean an external AI service will not receive data.

Which activity in your application would you like to improve?

Describe what your team currently does manually and which system handles that work. We will get in touch to arrange a free conversation.

Let’s talk about your business