Our R&D Team at the International AISA’25 Conference on Practical AI Applications

Artificial intelligence is advancing rapidly. But for companies, the real question remains the same: how do you turn these advances into reliable, useful solutions that business teams can actually use?

That is exactly the focus of the work our R&D team presented at the 3rd edition of the international AISA’25 conference (Artificial Intelligence and Smart Applications), a scientific event dedicated to AI applied to real-world challenges.

From research to operational use

The contributions presented at AISA’25 follow a clear objective: narrowing the gap between AI’s technical capabilities and the concrete needs of organizations.

Our research focuses on three key challenges encountered in the field:

• Understanding business language, even when it is imprecise or non-technical
• Converting an intention expressed in natural language into automated actions
• Interacting reliably with complex digital interfaces

These challenges are central to process automation and process management initiatives.

What these advances change in practice

1. Better understanding of business requests

Research on language model reasoning makes it possible to better interpret complex, incomplete, or ambiguous texts.
For organizations, this means fewer misinterpretations and automations that better reflect operational reality.

2. Workflows generated from natural language

With approaches such as Text2Workflow, a process can be described in a few sentences and automatically transformed into an executable workflow.
The direct benefit: greater autonomy for business teams and significantly shorter implementation times.

3. More reliable interactions with existing applications

Research on user interface understanding using multimodal models (text + image) improves the ability of AI systems to operate across diverse application environments.
The result: more stable automations, even when interfaces change.

AI designed to last

Presenting this work in an international scientific setting is not a theoretical exercise.
It is a key step to test, compare, and validate approaches before integrating them into solutions used on a daily basis.

This approach ensures:

• more robust technologies,
• better explainability of AI-driven decisions,
• continuous improvement built on solid foundations.

Looking ahead: more accessible and better-controlled AI

This research directly feeds the evolution of our process automation and process management solutions.
Our goal is straightforward: put AI at the service of business teams, without unnecessary complexity, and with real user control.

In the short to mid term, this translates into:

• tools that are easier to configure,
• more reliable automated processes,
• faster adoption by operational teams.

Applied research is not an end in itself.
It is a practical lever to build AI solutions that are useful, controlled, and aligned with today’s and tomorrow’s business challenges.