2024 was a pivotal year for artificial intelligence (AI). Improved performance, new services, etc.Generative AI have made a series of announcements. They validated the durability of this technological revolution, despite the obstacles to its adoption, particularly in France. From now on, we'll have to make do with AI! So what are the prospects for 2025? An update from Hervé Druez, a specialist in AI and the integration of marketing into corporate marketing and communications strategies.
According to a Statista study published in January 2025, almost half of companies are not currently planning to integrate AI into their operations.
What does 2025 look like? Will AI continue to progress and become a mainstream part of business?
Training
Artificial intelligence in the enterprise: the state of the art
Discover how AI can optimise your processes, improve your personal productivity and support the growth of your business, through tools, concrete use cases and successful examples. The ethical, security and strategic issues associated with its adoption are also addressed.
Ever more powerful tools, but...
Since the beginning of 2023, generative AI publishers have been making a steady stream of announcements. In particular, OpenAI has continued to lead the way with its ChatGPT4 version and 01 preview, and new services as varied as SearchGPT, advanced voice, projects and the release of Sora as a video generator.
Training
Generative artificial intelligence, from ChatGPT to MultiGPT
What are the new business models for generative artificial intelligence (GAI) within companies? With an overview of the various existing solutions (ChatGPT, Gemini, Midjouney, etc.), find out how GAI meets the specific needs of certain business areas (health, employment, etc.).
Participants speak out:
"A very good approach to discovering this vast field of generative AI. Excellent exchanges with the participants, who were naturally motivated by the quality of the trainer's performance. Good balance between practical cases, theory and discussions. Especially in 2 days."
"Fun training easily accessible to all levels. Practical exercises with real impact. I was very sceptical about AI when I came and I came away from the course with the knowledge to better understand, use and criticise the tool."
Increased international competition
Competition is set to intensify in 2025. Financial resources will make all the difference in the battle between the technology players, and some surprises are in store.
For example, Grok, which has been heavily backed by Elon Musk, could spring a surprise at a time when OpenAI is under financial pressure. Could it be dropped by Microsoft? Some observers think so.
In short, there could yet be a redistribution of players. And that's without taking into account the entry of Chinese players who are exacerbating the competition, such as the recent release of DeepSeek 3.0.
What are the current limits?
Despite performance gains, the main market-leading LLMSs (ChatGPT, Gemini, Claude, Mistral, etc.) are playing in a tight pack. They seem to be facing a kind of glass ceiling with error rate which are still very high on the various benchmarks.
For example: these error rates are higher than 20% on the MMLU-Pro test covering 12,000 questions and 14 skill areas (biology, business, chemistry, law, mathematics, etc.).
Will we have to wait for new algorithms to see significant progress in reducing the error rate? This is one of the challenges for 2025.
Diversity of specialist tools
A number of specialist AI applications made their mark in 2024: Heygen, Midjourney, Gamma App...The applications of the Horowitz ranking of the world's top 100 most-frequented AIs are becoming increasingly well-known.
But beyond the general applications, for each functionality, for each sector, "best of breed" solutions are beginning to emerge, with undeniable added value. As shown by the ranking of corporate AIs recently published by 574 investthe SNCF's investment fund. For example, the legal world now has recognised players such as Jus Mundi and Doctrine.
The challenge for a company is to identify the specific applications it needs to equip its teams with.
The latest trend is the development of applications with new functions. Personalisation, memory, improved archiving, the possibility of creating intelligent assistants to help users maximise their use...
Examples:
- GPTs or ChatGPT Task: a simple way of setting up the delivery of an information watch to your inbox every morning.
- The project logic present on Claude, ChatGPT or Perplexity, under the name of space, making it possible to specialise and historicise a discussion thread with specific instructions and files.
The search switchover...
On-line research is no exception to this revolution. For a long time, LLMS were very poor tools for finding information, but things are changing.
Examples of response engines: Perplexity, Open AI's Search GPT and, since January 2025, Google Deep Research.
Use case: ChatGTP and Perplexity respond to the same query "What is a foundation model in generative AI?".
These response engines are redefining the rules of online search.
Sourced response engines are gradually replacing search engines. They produce automatic summaries of the sites deemed most relevant in relation to the question formulated (prompt). Users will then be able to check the information if they wish, but will no longer be obliged, as is the case today, to go from site to site and summarise the information collected themselves.
According to Gartner, the Google search engine could lose up to 25 % of its traffic by 2026. In short, a change in search logic!
Will there be a real reshuffling of the cards? Google has no intention of losing its hegemony.
Permissive AI in 2025
2025 is the year of the massive deployment of AI in OSes (Apple, etc.), browsers (Google, Safari) and reference software solutions (Adobe, Salesforce)..
Microsoft is already illustrating this dynamic with its CoPilot 365 but all of them intend to integrate AI at the very heart of their solutions.
Training
Key points:
- Mastering techniques for formulating effective prompts
- Apply the skills acquired to real-life use cases
- Acquire a working knowledge of Copilot's functionalities
- Understanding the ethical and security implications of using Copilot
Example of practical work: creating a summary of a document, writing an e-mail, creating formulas in Excel and creating a presentation with Copilot.
AI will be absolutely everywhere by 2025.
The widespread adoption of AI will make it easier for even the least technophile companies to adopt it.
What technological advances are expected?
Technological advances will not stop here. We can expect further sensational announcements from all the players in the AI ecosystem.
This is already the case for :
- Nvidia at the Consumer Show
- Open AI: the ChatGPT o3 version obtained a score of 85 % on the ARC-AGI test on 20 December 2024, announcing the possibility of reaching thegeneral artificial intelligence (AGI) in the near future.
What is AGI?
The AGI is an AI capable of understanding and learning autonomously, similar to a human, in various fields. According to researchers, it will be operational by 2030. Before then, super-intelligent artificial intelligence (ASI) could see the light of day. This is a hypothetical AI that would surpass human intelligence in all areas, including creativity and social skills.
The major trend of the moment, however, remains the rise of the intelligent agentsThis is an essential component of the rise of AI in the enterprise in 2025. For example: on 14 January 2025, OpenAI launches the Tasks functionality for ChatPGT.
These LLMS-based solutions, capable of interacting with third-party applications, offer a vast array of new uses.
The new uses are varied: from automated intelligence to customer relations, from automatic pre-processing of administrative tasks to the automation of certain commercial tasks, from data analysis to finance. Directories of intelligent agents are already beginning to appear: AI Agents Directory and AI Agents List.
The obstacles to be overcome
Compliance with the IA act and the RGPDHowever, data security and confidentiality remain a key concern and will continue to hold back the deployment of AI.
But companies cannot remain on the sidelines of this revolution without running the risk of eroding their competitiveness in the medium term.
They will also be subject to the development of shadow IA with employees increasingly using tools without the approval or control of IT departments. The arrival of new generations of employees trained in AI is reinforcing this trend.
What are the solutions?
The only answer must be employee awareness and the introduction of generative AI usage charter clear and precise. A well-designed charter will ensure that AI is used responsibly and effectively within companies.
Establishing a dedicated charter is essential for guarantee the controlled use of artificial intelligence in business. This must include several key points:
- Define a clear framework: put in place an approval and control procedure for AI systems, specify good practice and prohibit certain uses.
- Raising awareness and providing training: informing employees about the risks associated with AI and offering them appropriate training.
- Structuring the governance Appoint a person or team to oversee application of the charter.
- Adopt a pragmatic approach: ensure that the charter is accessible, understandable and easy to apply on a day-to-day basis.
"There are many advantages to using generative AI today, just as we have moved on to the calculator or the Internet. Today, we are convinced that it will be an indispensable working tool in the years to come."
Ilhem Alleaume, Director of People Development & Learning at L'Oréal (January 2025)
Companies need to accelerate their transition to AI, by placing the human factor at the heart of this transformation. To succeed, they need to put in place appropriate training and clear governance (IA charters for example) in order to strengthen confidence and maximise the effectiveness of the tools. Support for this transformation must be gradual and voluntary in order to bring about a lasting change in working habits, while demystifying AI and professionalising its use. AI in 2025 will continue to be a revolution that cannot be ignored. Investing intelligently in AI, while putting employees at the centre, is the key to reaping the benefits of this revolution. To remain on the sidelines of this phenomenon would be to take a major risk in the long term.
