Despite all the praise generative AI receives, there are still few use cases where it genuinely saves companies time or money. Most businesses remain in the exploratory and testing phase, aiming to stand out from competitors. To fully harness the potential of this new technology, three new concepts must be integrated into how we think about and adopt generative AI. Here’s what they are:
1. Moving from “Dialogue” to “Button”
At MTL Connecte this past October, Philippe Harel, Director of the AI Practice at Umanis, introduced a key idea during the conference titled “Generative AI: Reinventing Tools and Reassessing Strategies.”
“If you need to engage in a dialogue with a generative AI agent, it generally means the tool is poorly designed. Instead of typing out a question for 20 seconds, the interface should already anticipate our needs and suggest something. Ideally, all we’d need to do is press a button, rather than type, to trigger a process. The AI should already be aware of the user’s requirements.”
Harel notes a shift in approach to developing user experiences:
“Generative AI is no longer just human-centric but tool-centric, focused on the product or interface, adapting it to be more flexible based on what we know about user needs.”
2. Moving from Chatbots to Creating Digital Workers (with Agentic AI)
Having conversations with AI is great… but enabling it to carry out tasks is even better. That’s precisely the promise of “agentic AI” (from “agency,” meaning the capacity to act autonomously).
In an interview with Les Affaires, Vasi Philomin, VP of Generative AI at Amazon Web Services (AWS), described the next generation of virtual agents as “digital workers.”
According to Philomin, these “chatbots” will handle returns, inventory changes, or refund requests by directly accessing the necessary APIs, without requiring any human supervision.
3. Creating Custom Virtual Agents
One major challenge for generative AI remains its high energy consumption, driven by the powerful models behind tools like ChatGPT, Gemini, Copilot, and others. However, at MTL Connecte, during the talk “Catalyzing the Enterprise with AI: A Practical Guide for Successful Integration,” Marc Boyer, Director of Cloud Services for Google in Eastern Canada, highlighted a growing trend toward developing “local” AIs:
“In B2C, we’re seeing a federation of models where AI capabilities are accessible only through major data centers and cloud providers like Amazon, Microsoft, or Google. But we’re starting to federate models that enable local processing power. We’re already seeing this with new Pixel phones or iPhones equipped with neural network chips, allowing certain tasks to be performed without querying the cloud.”
Boyer cited two examples: photo editing and character additions—tasks that could be accomplished solely with a smartphone’s processing power.
This is another avenue to explore for unlocking the full potential of generative AI in the future!
training.isarta.com 
