Malt Logo

Artificial intelligence is no longer a thing of fiction. In just two months, ChatGPT has attracted 100 million users, compared to 4.5 years for Facebook. This rapid adoption marks the dawn of a new era, that of widespread AI. This leap forward actually hides a long process that began several decades ago. Just as mechanization revolutionized industrial production lines in the 20th century, starting with blue-collar and then extending to white-collar jobs, AI automation today aims for three objectives: improving working conditions, production quality, and operational gains.

Beyond the promises of optimization, the rise of widespread AI poses immense technical, societal, and ethical challenges. This disruption, which shakes up traditional customer journeys, paves the way for hyper-personalization but also poses significant threats to employment. Like any technological revolution, its responsible adoption will require a decade of efforts to train employees, define a robust ethical framework, and address security and data privacy challenges. In this ongoing transformation, everyone, from companies to freelancers, will need to embrace this incredible opportunity to apply it best to their activity. A path fraught with obstacles, but also promises for those who can tame this new technology.

The first Smalt Talk event took place at Malt on February 27, 2024, gathering experts to address these questions with the following guests: Natacha Agafonov (Freelance Project Manager at Moët Hennessy), Mathieu Caron (Global Consumer Care & Experience Director at L'Oréal), Alexandra El Amari Cunin (Regional Sales Director at Saleforce), Victor Kessler (General Manager at ISDI), Claire Lebarz (Chief Data & AI Officer at Malt) et Nicolas Marchais (Co-founder of m.ai club).

1. Definitions and key terminologies

Before delving deeper into the numerous implications of general artificial intelligence on businesses and workers, it's essential to establish some definitions and key concepts to properly understand the various technological components at play.

  • Artificial Intelligence (AI): A field of study aimed at enabling machines to imitate or perform tasks that would normally require human intelligence, such as vision, voice recognition, decision-making, or translation. AI encompasses both simple algorithms and complex systems capable of learning and reasoning. AI exists in our daily lives, for example, through the use of automatic translation or personalized recommendations pushed to users based on their purchase history or online behavior.

  • Machine Learning (ML): A sub-category of AI dedicated to designing systems capable of learning from human feedback on so-called structured databases. Machine learning algorithms identify patterns in data to make predictions or decisions. According to Dan Miklovic, founder and principal analyst at Lean Manufacturing Research LLC: "Machine Learning does not replace humans; it helps them work better and become more efficient." Thus, Machine Learning will assist professionals in making decisions, whether in the healthcare sector for diagnosis or in economics to anticipate market trends, for example.

  • Deep Learning: Deep Learning is a branch of Machine Learning based on unstructured data. It is a machine learning technique based on deep artificial neural networks that share similarities with the human brain. These networks, composed of multiple hierarchical layers, allow for learning complex features from vast volumes of raw data without human assistance. It is particularly effective for image and speech recognition or content creation.

  • Foundation Models: A recent advancement, these models are massively trained on unstructured data to acquire vast general knowledge. They can then be fine-tuned to perform specific tasks without complete retraining. GPT is an example of such a model.

  • General AI: The ultimate, yet distant goal, General AI refers to systems endowed with comprehensive and versatile intelligence equivalent to human intelligence. This could push the boundaries of innovation but also raises immense challenges.

Find an AI Expert

null

2. The importance of data in the construction of customer journeys

For Parker Harris, co-founder of Salesforce, "quality AI relies on quality data." Currently, 90% of the data held by companies is said to be unstructured. Unstructured data has not been processed, meaning it exists in various formats such as photos, audio, PDFs, etc. Conversely, structured data is more like a database: it's information that has been prepared and stored to be accessible. AI precisely allows for the transformation of this unstructured data into structured and usable information, or into structured data.

Why is it important for this data to be structured? Let's consider a few examples to illustrate this:

  • Schneider Electric integrated its customer service with AI to improve its response time for customers. How? By creating a database of recurring queries, or FAQs, and adding a bot that can redirect users to these pre-built answers. This automates "low added value" tasks, freeing up customer teams to handle more complex issues that require more attention and thus more personalization. The result? A 30% improvement in time spent and a better customer experience.

  • LVMH faced an unprecedented crisis in its "champagne" vertical due to Covid-19. Before 2020, a majority of sales went through wholesalers, but after the health crisis, the demand for direct sales exploded. The customer team had to implement a CRM adapted to a vast amount of unstructured data. The result? A direct link with its customers and unprecedented personalization.

3. AI to redefine customer experience

"AI is for developers." This was a common sentiment not too long ago. However, the application of AI within companies is starting to become democratized and is moving beyond the technical department. How? By offering two major fields of application for customer experience. In the "front" with conversational assistants, chatbots, and virtual agents (already democratized for about a decade). But also in the "back" with automation, personalization, and the generation of custom content through the exploitation of unstructured data. We are moving from the initial very productivity-oriented use cases, which responded to a logic of optimization, to a very personalized customer experience with a high level of advice. As mentioned above, we now talk about an augmented agent or agent 2.0 to meet these demands. The agent automates repetitive tasks to focus on tasks that require personalization.

But it's easier said than done. From a technical standpoint, the success of AI projects within a company mainly depends on the company's ability to be customer-centric and thus correctly model customer journeys to anticipate all possible consumer responses. Being customer-centric is primarily about having a desiloed approach among the different departments of the same company. Indeed, AI is no longer "just the technical team's problem." What will this look like tomorrow? Will all departments have control over AI within their service? We will know very soon.

Here are some examples of current enhanced customer journeys:

  • L'Oréal offers the Modiface service, a digital skin diagnostic tool built on 15 years of scientific research conducted by its teams. Users can share a photo or video of their skin to receive tailored product recommendations. The result? Years of research enhancing the customer experience.

4. Impacts on employment: threats and opportunities

While there are strong fears of massive job losses related to AI, partly due to sensational newspaper headlines over the past year, experts urge caution. In short: it remains difficult to make a definitive statement on this issue at the moment. The introduction of AI in businesses is the logical continuation of digital transformation: some jobs do not yet exist, while others will simply evolve. Because while the adoption of AI will indeed evolve existing professions, this will primarily involve training rather than job eliminations. As with the arrival of chatbots a few years ago, professionals will be repositioned on activities with higher added value in contact with the end customer. Salesforce even anticipates the creation of 11 million new jobs by 2028 thanks to AI, enhancing humans and thus acting as an assistant that allows them to focus on hyper-personalization and advisory services.

And what about freelancers? On the freelancer side, the demand for AI skills jumped by 250% in 2022 at Malt alone, proof of the creation of new professions. The most sought-after skills in 2023 were ChatGPT (30%), NLP (20%), Chatbot (18%), and Midjourney (5%). It's also noted that the number of freelancers listing GenAI skills on their profiles jumped by +120%. But although the majority of projects were in tech at 34% and data at 38%, it will be interesting to follow the rise of AI within other business verticals in the coming years.

5. Essential ethical and legal challenges

Trust and ethics remain a major concern for businesses, with over 50% of professionals surveyed still wary of AI. Data security and its processing will thus be the main investment focus in the coming years. With the AI Act passed in 2023 being developed at the European level, it is certain that iterations will not be lacking.

It will also be necessary to take into account and correct potential biases and discriminations generated by these human-derived data. AI could, in fact, help reveal and correct these biases. Companies will also have to deal with an evolving legal framework, as exemplified by L'Oréal's ban on the use of AI-generated images of human beings. A significant effort in education and awareness will therefore remain indispensable in the coming years.

If artificial intelligence is revolutionizing the customer experience by exploiting previously untapped data sources, its adoption also raises significant societal, legal, and trust issues. Only a gradual and reasoned integration, along with the training of employees and a robust ethical framework, will enable a sustainable transformation and create lasting value rather than accumulating risks.

Find an AI Expert