Hayete Gallot, Chief Customer Experience Officer at Google Cloud, decrypts for generative AI market in recent months and details the new AI capacities deployed by Google.
JDN. In 2022, Google sounded an internal alarm, evoking a “red code” in the face of Google’s risk of downgrading if the teams did not redouble efforts on generative artificial intelligence. Three years later, Google is at the forefront of innovations in the sector. How did the company manage to reinvent itself?
Hayete Gallot. I think we have always been leaders on AI. When you look at our search engine, it’s machine learning. Transformers are we who created them. Our investment in research, especially on Deepmind, is enormous. On the Tensor Processing Unit (TPU), we created our own equipment because we could not find hardware that could deliver what we needed. I think we have always been ahead, but we were not very good at telling our story.
At the start, we focused on the general public with our search engine and our automatic summaries, and we were confident on these products. But we realized that the business market represented enormous potential for AI. We have therefore restructured our teams and announced reorganizations, especially in terms of engineering. It is now much better organized. The efficiency you observe today is the result of our new approach. We better define our priorities, we organize our engineering teams differently, and we optimize collaboration between Deepmind and our other divisions to fully exploit their innovations. This is what allows us to launch products like Imagen, Veo or Lyria.
Project Mariner made a lot of talk about him during Google I/O. Could you specify the key uses of this agent and how it is distinguished from other market solutions?
Project Mariner is really innovative because it aims to demystify AI agents for the general public. The objective is to build confidence by showing concretely what agent can do. Its principle is simple: on a browser, you enter a request, and the agent begins to work for you. Its particularity is to document each step in its process, providing references and a final result. It is a great opportunity to make it clear what an AI agent really is.
In 2023, the AI was still at the experimental stage in most companies. In 2024, we attended its first deployments in production. In 2025, you observe in your customers first signs of maturity in the adoption of AI?
I would say that we are in a real transition from the POC to the scale. Today, business decision makers have integrated into the decision -making process. Companies have worked a lot on their processes and how to move from an idea to a large -scale deployment. When they decide to move forward, they meticulously examine several aspects: security, principles, scalability, costs, and above all how to anticipate and control expenses. What is particularly interesting is the diversity of the sectors involved: from Renault to LVMH, via construction and Galeries Lafayette. All sectors are now in competition, with an approach focused on business processes. The objective is to promote wide adoption, by forming users.
What are the products of the Google Cloud that your customers adopt the most?
Our approach with Gemini in Workspace was to democratize AI. We have set up an incentive tariff model that has largely facilitated its adoption. The result is impressive. All users use it. On the other hand, Notebooklm is also an excellent example, with 100,000 user companies in a few months. We also note a marked adoption in the AI sector for customer service. Companies want to invest in innovation and seek to optimize their efficiency.
Finally, there is the whole developer ecosystem. Companies cannot recruit enough developers. All code development and developer assistance tools therefore become crucial, because existing developers must produce more code. Cybersecurity follows the same logic. Faced with the lack of analysts, security agents coupled with Threat Intelligence now allow more people to contribute to the ecosystem.
If we now look at the market, in particular by comparing North America and Europe, do you still see today differences in the adoption of AI technologies?
At first, Europe actually accused a certain delay. Today, AI adoption levels are quite similar between North America and Europe. The real difference is now around the challenges of technological sovereignty. All companies, whether global or local, share a common concern: innovate without compromising their competitiveness, while meeting regulatory requirements.
Act (General artificial intelligence) arouses many expectations. Does Google work on this subject, and do you anticipate new concrete use cases?
The evolution is fascinating. We have gone from the assistants to the agents, and now some evoke Act. But today, we are resolutely in the era of agents. Customers are pragmatic. Their concern is not theoretical but operational. They wonder how to build an agent, how to orchestrate them, how to create confidence with users, how to help them concretely? This approach is reassuring because it keeps the course on the real added value.
