When
you started working with GeneXus, you built something that goes far beyond a software system. Transaction by transaction, rule by rule, you accumulated years of knowledge about how your business works. That is your Knowledge Base: the data model, the business logic, the relationships between entities, the rules that define how the system operates. All of it is contained in one place, in a structured and coherent way.

Today, with the explosion of generative artificial intelligence, tools are emerging that promise to connect that KB with LLMs like ChatGPT, Claude, and many more. The promise is tempting: giving AI access to all the context of your system so it can help you develop faster and at lower costs.

But before connecting your KB with any AI tool, there is a question worth asking carefully: do you know exactly where your data goes, and who has access to it once it leaves your infrastructure?

The mechanism few explain

Some KB-AI integration tools work in a way that is important to understand in detail. The process is, in essence, as follows:

• Your Knowledge Base is exported to flat files: structured text that represents the content of the KB.
• Those files are sent to an external LLM so the model can read them and reason about them.
• The LLM generates responses or code based on that information.

However, the problem is not in the idea of connecting the KB with AI. The problem is in the middle step: your KB leaves your infrastructure, in plain text, toward external servers (which you do not control), with the security controls that third-party provider decides – or does not decide – to implement.

This is a concrete operational consideration that must be taken into account by the GeneXus Community, especially by teams developing Mission-Critical Systems.

Data you should know

When
a company connects its KB with an AI tool that exports data to external infrastructure, it is creating a link with a third party.

According to the IBM Cost of a Data Breach 2025 study, 97% of organizations that reported AI-related security incidents did not have adequate access controls over their AI tools. In 2025, the absence of governance left a global average cost of USD 4.44 million, including legal costs, forensic investigation, notifications, customer loss, and operational recovery time.

Exposing how your business works through a tool that exports its contents to an external LLM – without auditable logging, without per-operation access controls, and without formal governance – is falling into the shadow AI trap (use of AI tools without corporate supervision or control), applied to your most critical asset.

What makes the manufacturer different

When
working with AI agents and a GeneXus Knowledge Base, the technical process is similar in any tool: GeneXus objects are converted into a textual representation of the KB objects. That textual representation can be manipulated with different source management IDEs, such as Visual Studio Code, or any agentic tool that assists development tasks.

Luis Murillo, GeneXus Technical Support Manager and expert in Legacy System Modernization, explains it this way: “With GeneXus for Agents, the recommended path is to use CODA CLI, the Globant tool that integrates naturally with Glob.AI OS, and this is where our main distinction lies, since Globant has signed confidentiality contracts with LLM providers (such as Anthropic, OpenAI, etc.). That means the client’s intellectual property is protected and cannot be used by LLM providers for training: your code in text cannot be used to train models or shared with third parties. It is not something you have to configure; it is already on.”

In this way, the chain of custody is as follows:

Client – GeneXus + Globant – LLM provider (confidentiality guaranteed by default)

In contrast, with third-party integration tools, one more player is added in the middle: the provider company itself, with which you will need to sign equivalent confidentiality agreements.

Client – external provider – LLM provider (confidentiality to be agreed separately)

And this is not a minor detail. That is the difference between your business model traveling with or without a confidentiality contract.

There is also a second important difference. GeneXus is the provider of the symbolic AI that generates application code.

This distinction matters for several additional reasons:

GeneXus built the schema for modeling reality in a knowledge base – it did this more than 30 years ago

No one knows its structure more deeply than the manufacturer. That translates into more precise, safer operations with less risk of incorrect interpretations.

Every Operation Is Logged

The governance model includes per-operation logging, role-based access control (RBAC), and on-premises support. This allows you to audit who did what and when.

No Additional Cost

GeneXus for Agents is included for all users who are up to date with maintenance and use GeneXus Next or GeneXus 18. It is not a separate service with its own monthly invoice.

Questions you should ask before deciding

If you are evaluating how to connect your KB with Generative AI, or if someone on your team is already doing so, there are some concrete questions your team should be able to answer before moving forward:

• Does business data leave your infrastructure in plain text? If so, is there a confidentiality contract with the LLM provider?
• Does the tool add an intermediary between your business and the LLM? What guarantees does that intermediary offer regarding the handling of your data?
• Is confidentiality activated by default, or do you have to configure it?
• Are there auditable access controls over what information the tool shares with the LLM?
• Do you have logging of operations performed on your KB through that integration?
• What does the provider’s service agreement cover in the event of a breach involving your KB data?

If any of these questions does not have a clear answer, it is worth pausing before activating the integration. The speed that AI brings is worthless if the price is exposing the most sensitive asset of your system.

Your Business Knowledge Is Your Competitive Advantage

The GeneXus KB is the structured representation of years of learning about how your business works: what rules apply, how entities relate, what processes are critical, what exceptions exist. It is the kind of knowledge that takes years to build and that, if it falls into the wrong hands, can be reproduced or used by third parties.

When
you connect that KB with AI using GeneXus for Agents with CODA CLI, that knowledge travels with confidentiality guarantees active by default, through a chain that the manufacturer itself controls, using living technology that will continue to evolve. What changes is that AI agents can now operate on that knowledge in a structured, precise, and GeneXus engine-validated way.

AI does not have to mean giving up control. GeneXus for Agents is available now, ready to connect your favorite agents with your KB natively, securely, and at no additional cost.

For more information, we invite you to visit the GeneXus For Agents page and the technical documentation available on GeneXus Wiki.

You may also be interested in reading:

GeneXus for Agents: Development with GenAI without losing control

The Problem with Prompt-Based Development

GeneXus in the Era of Agentic Development