Using AI in business: case studies, tools and integrations (APIs)
- Reworking software means taking over existing code to correct it and make it evolve.
- We start by accessing the code, then carry out a free maintainability analysis.
- Depending on the quality of the code, improvements are made using the same technology or part of the code (often the back-end) is overhauled.
- Changing service providers allows you to unlock fixes when the current team is no longer available.
- Updating a language or migrating a technology improves performance and security.
Using AI in business is no longer a “futuristic” subject: the aim of this presentation is to show concrete cases (marketing and operational), tools to try out, and how to integrate them into a company - with time set aside for questions and workshops.
AI in business: a simple definition and why it's “taking off” now
In the presentation, AI is explained simply as a software capable of learning and reproduce certain actions that a human would do. What's changing today is that tools like ChatGPT have made AI can be used immediately in companies, in a very accessible way.
The benefits are twofold:
- go faster (saving time on time-consuming tasks),
- acquire “skills” on demand (marketing, support, etc.), particularly for small organisations.
LLM (such as ChatGPT) vs specialised tools: how to choose
The presentation cites several ChatGPT equivalents (LLM): Copilot (Microsoft), Claude, Mistral... The idea: these models are global and can help with a wide range of needs.
Then, we take things “one step further” with specialised tools by use (content, video, analysis, support, etc.). The proposed logic :
- start with an LLM to explore,
- then choose targeted tools when a need becomes recurrent (marketing, support, HR, operations).
Marketing: posts, videos, blog, newsletter, white paper, Excel analysis
Marketing is presented as the most “tangible” area to start with, because the link between AI and results is immediate.
Examples cited:
- create social network posts with an LLM or dedicated tools (a “copy a” type tool is mentioned),
- generate videos for networks (with a workshop demonstration planned),
- produce long content blog articles, newsletters, white papers.
The speaker stressed an important point: AI helps to avoid blank page syndrome and speed up production, but it recommends do not publish as is. The idea : rapidly generate a content base using ia, then add a personal touch before publication.
Marketing data analysis (Excel) : it is also explained that ChatGPT (or equivalent) can be used to analyse internal data and produce analyses/diagrams, with an exercise planned for the workshop.
Customer support: 24/7 chatbot, redundant FAQs, lead capture
Customer support is presented as a major lever, because it can be very time-consuming. An example is given: a large number of questions can be redundant (the speaker mentioned up to “80%” in some cases).
What AI could bring, according to the presentation:
- a chatbot available 24/7 to answer frequently asked questions,
- a first level of support that can also capture leads and direct them to the right channel.
He also mentioned the possibility of’analyse what is being said on the internet about a company (opinions/comments) to understand perceptions and then adapt certain decisions (example given of feedback by country).
HR: job descriptions in minutes, language, attractiveness and recruitment
In terms of human resources, the main example concerns the writing job descriptions A task that can sometimes get in the way, because it requires time and precision.
The principle presented :
- AI makes a first draft in just a few minutes,
- then there is a short time left to adapt and customise.
Other benefits mentioned:
- reduce the barrier to language,
- make the offer more attractive,
- show that the company uses AI (which may attract certain profiles).
Operations: defects, photo stock, multilingual call centre
The presentation includes a number of “operational” cases:
- Fault detection For example, the example given of learning from photos to detect scratches/damage (case mentioned: Volvo), to avoid sending a product with a defect to the customer.
- Inventory / stock via photo The ability to take a photo and “count” items, for fast, regular stocktaking and better stock monitoring.
- AI call centre Setting up a multilingual reception area, with a demo example of how to book an appointment (for a dentist), to avoid the need for a dedicated full-time resource.
The key message: many of the building blocks already exist “in service”, which makes these solutions more accessible, especially when the cost is linked to usage (rather than a complete development).
Integrating AI via API: diary, email, software and workflows
An important passage concerns API and interconnection: the idea is not necessarily to “create AI”, but to integrate it to your existing tools.
Examples explained:
- connection to a agenda to propose available slots,
- management’emails The recommended approach is to produce a draft (draft) rather than sending it automatically in all cases,
- different workflows depending on the context : existing customer vs new customer, welcome message, internal routing,
- the ability to automatically send a relevant resource (e.g. a PDF of prices) depending on the request (e.g. “bathroom” mentioned).
For’integrating AI into your marketing, HR, administrative and operational processes.
To move towards “AI agents in companies”capable of carrying out actions
Risks & adoption: deepfake, confidentiality, Google/SEO, training
Deepfake / identity theft
One question concerns the risk associated with face and voice. The danger mentioned is the deepfake (putting someone's head on another image/video) and voice copying, with the idea that a very short sample may suffice (depending on the speaker).
AI-generated SEO content
An exchange mentions the detection of AI-generated content on Google's side. The recommendation in the discourse: use AI to speed up, but do not publish raw, and rework to add a personal touch and avoid “generic content” effects.
Confidentiality / shared data
A key point in the discussion: the risk of “feeding the beast” by giving sensitive content to a public AI. One alternative mentioned:
- host a model open source (e.g. Mistral is quoted) on its own servers,
- or use modes/solutions where there is no learning from your data (the speaker also mentioned the use of APIs as an option).
Internal adoption
Lastly, he pointed out that there can be a fear in employees (replaced by AI). The key: communicate, imply, and train so that AI and the automation of internal processes is a support tool, not an imposed project.
To remember
- Start with concrete examples: marketing, support, HR, operations.
- LLMs (such as ChatGPT) are a good entry point, followed by specialised tools take over.
- Above all, AI helps to save time on the blank page and tasks repetitive.
- The impact often comes from API connect AI + calendar + email + software + workflows.
- Don't publish the content “as is”: rework and personalise it.
- Anticipate risks: deepfake, confidentiality, choice of tools.
- Success = also training and change management.
The next stage
Would you like to identify 2-3 priority use cases and see how they can be properly integrated (APIs, workflows, security)?
- Take meeting with an AI integration expert
- Or contact us via the form for AI integration projects
Tell us about your project
One exchange, a thousand possibilities.
Describe your vision to us using this form: we'll analyse your request and get back to you within 24 hours with personalised advice and a concrete action plan.
We have the team and the resources to help you with your projects. Give us the details in this form and we'll get back to you as soon as possible to discuss them together.
This is the use of tools capable of assisting or automating certain tasks (writing, support, analysis, workflows) in a very accessible way. Setting up these automation tools requires a AI training for in-house teams.
Marketing (content, posts, newsletters) and customer support (chatbot for recurring questions) are presented as the most “tangible” to start with. Then you can extend to HR and operations. So you can start with add AI to your marketing, HR, financial and operational tools.
Look for a “use case + integration” approach: workflows, APIs (calendar/email/software), security and internal adoption. The aim is to transform an idea into a measurable operational process.
Yes, the approach often consists of connecting existing components via API and defining workflows (frequently asked questions, routing, leads). The most effective approach is to define 2-3 priority scenarios during a meeting. Take meeting with an AI expert, chatbot and automation
Reduce the sharing of sensitive data and favour architectures where your content is not used to “drive” a public model (depending on the options chosen). You can also frame an enterprise solution (governance, access, workflows, security) by contacting our AI and automation team.