AGENCE IA : VENTE, STRATÉGIE MARKETING, SERVICES IA

The speaker begins by addressing a technical issue with their live stream, ensuring the audio and video are working correctly. They then introduce the main topic, prompted by a message received: the potential of Artificial Intelligence (AI) in both the operational and psychological dimensions of business. The sender of the message, who used AI to draft it, highlighted that while AI's role in execution (marketing, sales, accounting, product creation) is widely discussed, its application in the pre-execution phase (clarifying vision, structuring strategy) and its psychological dimension are often overlooked, especially in the French-speaking market. The speaker agrees with the potential of AI in the psychological dimension, sharing a personal anecdote. They describe how they spent an entire Saturday using ChatGPT to improve their personal brand, focusing on self-discovery to articulate their values and singularity. This involved detailing their life story from age 18 onwards, providing information about their preferred books, meals, vacation spots, and other personal interests. For each element, they asked the AI what it represented about them, thereby using AI for deep self-reflection. Returning to the message, the speaker questions what the sender meant by "transformation" in business, clarifying that AI can indeed help diagnose and transform a business, but this often involves generative AI systems rather than a directly sellable product. They acknowledge the sender's point about AI assisting in augmenting, automating, and orchestrating via AI agents, calling it a significant, underexplored area with enormous potential. The discussion then shifts to questions from the audience. Nabil asks about selling vocal AI agents and the most profitable uses. The speaker explains that vocal AI agents are best used as receptionists for incoming calls, handling information processing and reception. While some might use them for direct sales, this can be less effective as potential clients may not be ready to buy. Regarding pricing, the speaker mentions a student who is a leader in vocal AI in France, selling installations for more than four figures, with recurring fees also close to four figures. The price depends on various factors such as call volume, functionalities (reporting, notifications, CRM integration), and the agent's tasks (taking orders, appointments, answering questions). Arnaud, a program participant, shares his background in IT project management, auditing, and delivery management, noting that with the right approach, interesting things can be achieved with AI. The speaker then promotes a podcast they recorded with Hugo Pana, an individual earning over €100,000 with AI, encouraging viewers to watch it. The conversation returns to audience questions. Someone asks about Yomi Denzel's net worth, which he estimates to be between 60 and 70 million euros, including investments, real estate, businesses, and company stakes. He highlights the difficulty in precisely valuing assets when money is "everywhere." A key question arises: "How to sign your first client when you start, as everyone talks about social proof but at the beginning we don't have it." The speaker emphasizes that proof is created and built, and social proof is just one type among many (mathematical, political, journalistic, authority, sociological). They cite an example of someone who, lacking personal social proof for an AI training, leveraged the success stories of others in the AI agency niche. The speaker suggests that if one is in a specific niche like lawyers, they can find existing cases, articles, or podcasts about AI use in that field to build credibility. Alternatively, one can offer free trials or advice to clients, then ask for feedback. A positive testimonial, even from a single piece of advice, can serve as initial social proof. This leads to a broader suggestion for a future live topic: "How to increase your credibility on the market when you start." The speaker then challenges the audience, noting a lack of questions and suggesting three reasons: either they are not taking action and thus not encountering problems, they believe they know everything, or they are passively waiting for information. They stress that action creates information, not the other way around. Another question concerns the pertinence of Meta/Google ads for B2B with medium to high-ticket clients. The speaker confirms its relevance, citing an example of a company that generated 90% of its acquisition through ads. They clarify that the effectiveness depends on the strategy, emphasizing that ads should not directly sell products to an unaware audience. High-ticket sales via Meta ads are possible (e.g., €3,000-€50,000), but the return on investment will be longer. The focus should be on cost per lead, call, and sale, rather than just cost per click, which is an optimization metric. Joa asks about using statistics combined with ROI simulations to sell services. The speaker fully endorses this, suggesting it can be part of a lead magnet or a key element in sales presentations. They propose showing a client how a manual process (e.g., handling 15 emails a day, taking 5 minutes per email) can be automated by AI, reducing time to seconds, and then presenting a clear ROI table detailing the financial savings and gains. This method, they believe, can significantly boost closing rates. Nabil asks whether the priority is to learn how to sell or how to implement automation tech. The speaker unequivocally states that the priority is to know how to sell. Without sales, there's nothing to automate. Regarding web development post-Covid, with the rise of no-code and AI-generated sites, David asks about the best approach. The speaker advises augmenting the offering beyond just a website. This means integrating marketing strategy, differentiating elements (like simulators or tools), and focusing on SEO (including AI-generated content). Crucially, they recommend niching down, citing the vast potential even within specific sectors like construction (BTP), where many can still thrive. They dismiss "TPE/PME" (small and medium-sized businesses) as a niche, arguing it's too broad to effectively target with limited resources. Christopher asks about the profitability of an AI agency versus training in the AI niche. The speaker views them as two distinct businesses. While B2B (agency) can potentially generate more revenue, B2C (training/support) can offer higher margins and competitive positioning. They mention Wolf Collective as an initiative to combine both. The discussion then touches on hosting solutions (preferring self-hosting on one's server for 90% of cases) and AI models (not using DeepSeek, but acknowledging its potentially cheaper tokens). Joa asks about combining horizontal and vertical niche strategies. The speaker recommends combining both in 90% of cases, for example, targeting the sales process for lawyers. The only caveat is to ensure the market isn't too small, suggesting at least 5,000 actors. The speaker then addresses the common objection of GDPR and data security. They admit to often "defying" this objection by pointing out that most people use everyday tools (Gmail, WhatsApp, Instagram) where data is already exposed. They assert that personal and company data are likely already compromised globally. However, they acknowledge the importance of GDPR for specific niches or sensitive data (e.g., health data requiring HDS certification for hospitals/clinics). The program does cover GDPR, with cybersecurity specialists involved for such cases. Finally, the speaker offers insights into their own prospecting methods: their newsletter (very effective), YouTube channel (especially for technical videos), and cold prospecting. They emphasize that while automation tools like LemeList can be used, initial cold prospecting is crucial for beginners. They also note that Instagram and LinkedIn can generate leads, though their Instagram and YouTube tend to be more B2C-oriented. Podcasts are a long-term strategy, generating fewer immediate leads. The speaker concludes by reiterating their availability for questions on Instagram or Telegram, and briefly mentions wanting to demonstrate how easy it is to use AI tools like CloudCode and N8N by showing a "ping-pong" interaction between the tool and AI for guidance, but notes that technical discussions often attract developers who may not convert into paying clients for their support programs.