L'article
In episode 2 of this series, we outlined the barriers preventing B2B companies from implementing a content strategy that meets their ambitions, particularly the lack of specialized internal resources and the difficulty in offering unique content.
When judiciously integrated into the content creation process, generative Artificial Intelligence can help marketing departments overcome certain obstacles, as long as the human element remains at the helm. Here’s how…
Content marketing : What AI can do…
AI has a place in the content creation process. This is undeniable. However, to adopt and capitalize on this disruptive technology, it is essential to understand its true scope of action and remain aware of certain limitations.
1. Summarizing articles and studies
In their sourcing work, marketing teams can use AI to summarize articles or studies into bullet points, speeding up the documentation phase and providing a broader overview of the state of the art.
This application benefits technical and complex topics, allowing for better contextual work and giving content more depth, even a scientific foundation.
Limitations : AI cannot always extract the most significant insights from studies, especially when the original file is very large. It tends to analyze the first pages and invent (or “hallucinate”) the rest. Therefore, human verification is essential to avoid going off track. Additionally, the sources to be analyzed must always be verified as reliable and authoritative in their field.
2. Analyzing a database to extract insights
Do you have the results of a study in Excel format or a screenshot of a data table? You can submit these elements to an AI chatbot to identify trends, patterns and lessons that can be used in your content. AI supports marketers who are not necessarily comfortable with data analysis.
Limitations : Caution is required. Even advanced AI chatbots like ChatGPT-4 and Claude 3 Opus often make calculation errors, even with simple additions. Therefore, all outputs must be verified by a human.
3. Understanding concepts through practical applications
This is probably the most relevant use case for ChatGPT, Gemini, or Claude AI. These chatbots can serve as response engines, replacing Google for queries that require understanding beyond theoretical definitions.
Let’s take a concrete example: you want to understand the benefits of Sales Enablement in a practical, hands-on way.
- Google will likely offer you dozens of web pages addressing the question in an encyclopedic way, with definitions, advantages, disadvantages, best practices, etc.
- Your AI chatbot (with the right prompt and on a good day) can provide a concrete scenario of implementing Sales Enablement in a fictitious company, where you can define the industry, number of employees, and other details.
This ability to explore concepts in an ultra-personalized manner enhances the relevance of content and goes beyond typical perspectives.
Limites : AI excels at generating disconnected answers that appear very convincing. The examples provided may not always align with real-world scenarios. Human operators must remain vigilant and watch for inconsistencies and illogical assertions in the responses. They need a deep understanding of the subject and its implications to assess the answers’ relevance. Moreover, the quality of the prompt is crucial; posing a good prompt is essential for obtaining appropriate responses. Prompting becomes a skill for marketers
4. Translating content for documentation
AI can act as an automatic translator, but this application was already widely used before the arrival of ChatGPT, notably through tools like DeepL or Google Translate. However, AI enables choosing the tone of the translation.
Limitations: There is a tendency to translate literally or even word-for-word. Careful proofreading and rephrasing work are always essential.
5. Adding emojis to a social media post
This is a feature that allows the integration of relevant emojis into already written posts.
The results are generally good, although it’s necessary to filter out emojis that are interpreted too literally (for example: using a loaf of bread to illustrate the expression “a lot on one’s plate”).
6. Completing a bullet point list from existing elements
AI can potentially use the first three or four items of a bullet point list to suggest the continuation. For example, if you have identified four relevant benefits of Brand Content, you can submit them to your chatbot and ask it to continue the list. Generally, having a relevant base enhances the likelihood of less generic results
Limitations: You may need multiple exchanges with the chatbot to arrive at two or three interesting ideas. Therefore, the usefulness of this application is limited as the expected time-saving benefit is not guaranteed.
Content marketing: what AI cannot do
As we’ve seen, AI serves more as a co-pilot than a pilot when it comes to creating high-quality marketing content.
It accelerates certain aspects of sourcing and documentation, but it is truly humans who bring the stamp of originality and differentiation. Here’s what AI cannot do in content creation.
1. Creating original content that stands out from the competition
By definition, AI does not create content from a blank slate. It relies exclusively on the data it has absorbed during its training phase. As a result, the originality of AI-generated content is limited to what already exists in its dataset.
Any “creation” is essentially rephrasing existing content… at best. Sometimes, the generated text already exists verbatim elsewhere on the web, which can harm SEO and brand perception.
Such content cannot differentiate itself from the competition or the roughly 600 million blogs on the internet (According to HubSpot). Savvy professionals have developed a knack for identifying AI-generated content on LinkedIn and blogs. There are indeed telltale signs that give it away:
- A monotonous pace that is consistent throughout the text;
- The same idea is generally repeated over two or three sentences, with paraphrases;
- Vague and general information that does not demonstrate a deep understanding of the subject;
- No strong, potentially controversial or debatable viewpoint;
- Capitalization at the beginning of each word in titles;
- Capitalization after explanatory colons;
- The repetitive use of certain words or expressions such as “crucial”, “it is essential to”, “it is important to”, “specific characteristics”, “distinguished by” etc.
#2 Write in natural and idiomatic French
Historically, even before AI (with techniques like Spin Content), text generation tools have always performed better in English. This remains true with AI, which has been trained on predominantly English-language datasets.
In French, the standard tone is generally cold, impersonal, robotic, and neutral. If you ask AI for a more engaging style, 9 times out of 10 you end up with a literal and word-for-word translation from English, resulting in awkward expressions.
3. Infusing an emotional component into the contents
Marketing content isn’t always 100% educational and rational. In B2B relationships between companies, human decisions are still paramount.
To engage the audience, content must also appeal to feelings, subjectivity, and emotions. This is precisely where AI typically struggles. It often falls into a caricatured or overly emphatic advertising tone.
4. Integrating viewpoints or professional anecdotes
AI cannot conduct interviews with experts because this process requires an immediate understanding of the provided responses and the ability to adjust subsequent questions based on the conversation’s flow.
This level of interaction is crucial for capturing nuances, emotions, and subtle perspectives.
By design, AI cannot enrich content with personal anecdotes and strong viewpoints. It won’t be of much help in creating subjective content like opinion pieces, columns, and experiential feedback.
And it is precisely these elements of authenticity that engage audiences in the age of information overload.