Useful for explaining AI agents, MCP connections, and DataBird training in practical terms.
Best for
A B2B AI training company wanting a course or workshop on agents and enterprise integration.
Not for
An English-language developer-tools vendor wanting a paid video series judged on reach.
Key takeaways
An AI agent can sort customer emails and alert the right team.
MCP could connect AI to tools such as HubSpot or Salesforce.
New model capabilities need reliable data, clear rules, focused use cases, trained teams.
Check first: Confirm language and market fit, terms, and comparable training results before contacting them.
Main ideas
Build agents around real workplace tasks · Make connections between AI and business tools easier to maintain · New model capabilities need organisational foundations
Alexandre Miny de Tornaco writes about how businesses can put AI to work, especially through agents connected to everyday tools.
Their posts explain technical ideas in practical terms: one describes an AI agent builder by showing how it can sort customer emails and alert the right team; another explains how MCP could connect AI to tools such as HubSpot or Salesforce.
They also consider how new AI capabilities might change organisations, stressing the need for reliable data, clear rules, focused use cases and trained teams.
These examples may interest readers exploring business AI and agent-building.
Ask for details of any signature book, show or earlier body of work.
Miny de Tornaco translates AI developments into workplace tasks and decisions.
Rather than leave ideas such as AI agents, MCP or a new model as technical labels, they explain how people might use them: to handle customer requests, connect business systems, analyse sales or identify at-risk customers.
They also temper the promise of new tools with practical requirements: sound data, clear rules, targeted uses and team training.
Their MCP post is exploratory rather than conclusive: it outlines the protocol’s potential and asks whether it is reality or fantasy.
Readers get a question to examine, not proof that the proposed integration is already in place.
Across these posts, Alexandre Miny de Tornaco presents business AI as a practical integration and training challenge, not just a matter of choosing a powerful model.
Read more
The examples span agents handling routine tasks, connecting AI with company tools, and preparing teams and data to use new capabilities well.
Build agents around real workplace tasks
Alexandre describes an AI agent as able to understand a request, follow several steps, use tools and trigger actions.
Read more
The builder’s job is to connect it to the systems a team already uses. The focus is on handing off part of a concrete workflow, not treating chatbot conversation as the whole of AI.
An agent could handle part of the work involved in reading customer emails, finding relevant information and alerting the right team. Alexandre also lists tasks such as qualifying a lead, summarising a meeting and filling a spreadsheet.
Make connections between AI and business tools easier to maintain
Alexandre frames integration as a recurring obstacle: building a separate connector for each tool takes work, and connectors may need attention when a tool or AI changes.
Read more
They present MCP, or Model Context Protocol, as a possible way to connect AI with multiple tools through one connection. This remains a prospect, not a proven result; DataBird plans to try this kind of technology.
The post contrasts building individual links to systems such as HubSpot or Salesforce with an MCP connection that could let AI retrieve information across tools in one query. It closes by asking whether this promise is reality or fantasy.
New model capabilities need organisational foundations
Discussing an OpenAI product announcement, Alexandre sorts possible effects by how work might change: actions gathered in one interface, more tasks accessible to non-specialists and AI help with existing code.
Read more
Their conclusion: access to new tools is not enough; organisations also need reliable data, clear rules, focused use cases and trained teams.
The post suggests workplace requests such as analysing quarterly sales, identifying at-risk customers and drafting a follow-up email. It names data, rules, targeted uses and training as prerequisites for getting value from new tools.
This is a clear starting point for anyone asking what the job title means.
Read more
It answers “what is an AI agent builder?” directly, shows what the work involves and notes that the author runs training on the subject.
The post opens with the question in its title and answers it step by step.
AI once meant talking to ChatGPT; now it can handle more complex needs, such as processing a customer request, analysing a document, summarising a meeting, qualifying a lead or filling a spreadsheet.
An AI agent is described as an assistant that understands a request, follows several steps, uses tools and triggers actions.
An AI Agent Builder designs those agents and connects AI to the company’s existing tools: email, internal documents, tracking sheets, CRM and communication platforms.
The worked example is concrete: instead of reading every customer email, finding information by hand and alerting the right team, staff hand part of the work to an agent.
The post ends by announcing that DataBird has started AI Agent Builder training sessions for a new group of learners.
For brands planning a campaign
For a training provider, this post offers a ready-made course outline.
It identifies three skills to cover: defining the role, connecting an agent to existing business tools, and handing off a routine workflow, from inbox triage to team notification.
The email-to-CRM example makes a practical workshop exercise, starting and ending in tools sales or support teams already use.
This post addresses a practical integration question brands face first: how do you connect AI to a CRM and other systems without rebuilding the connection every time something updates?
Read more
It explains MCP in plain terms.
The post says the author first encountered MCP, or Model Context Protocol, at a conference run by Anthropic and Dust at the Builder Summit in February.
It lays out the problem plainly: today, you have to build a specific connector between AI and each tool, such as HubSpot or Salesforce, then repeat the process for every tool.
When either side updates, the connection breaks and the work starts again.
With MCP, once the connection is established, AI can pull data from all the tools in a single query; the post describes it as a universal connector.
The argument is prospective, not settled: the post says this could change things, make teams more autonomous in retrieving information to train their own models, and that DataBird plans to put this technology in place.
It closes by asking whether this is reality or fantasy, and what it adds compared with Apple Intelligence.
For brands planning a campaign
Two points matter for a training programme.
First, connector maintenance makes a strong case study: building a custom connector for each tool, only for it to break with every update, is a problem a workshop can tackle.
Second, the closing question shows how the author evaluates technology, treating a new protocol as a claim to test rather than a settled answer.
This post answers the question companies ask after every model launch: what does this actually change for us?
Read more
Rather than repeat the announcement, it sorts the implications into categories and closes with what organisations still need in place.
The post reports that OpenAI has launched GPT-5.5, described as its most capable model to date, and quotes Greg Brockman on a “super app” that unifies ChatGPT, Codex and an AI browser.
It then lays out the implications in three blocks.
First, where daily work happens: people move between browsers, spreadsheets, CRM, messaging, project management and internal documentation.
An AI interface could centralise tasks such as analysing a quarter’s sales, preparing a committee summary, identifying at-risk customers or drafting a follow-up email.
Second, which tasks become accessible: work once reserved for technical profiles, such as analysing a complex file, comparing sources, structuring monitoring, building a prototype or automating a simple task.
Third, what a code agent can do: understand the context of an existing codebase and help fix a bug, write tests or document a feature.
The closing argument is that new tools appear every day and everyone wants access, but getting value from them requires reliable data, clear rules, targeted use cases and trained teams.
Without those, even the best model is badly used.
For brands planning a campaign
This offers a template for a “what this release changes for you” segment at the start of a course: sort the announcement into daily work, newly accessible tasks and technical assistance, then close with readiness.
Before using the launch details in course materials, verify them against the announcement.
The final paragraph makes the commercial case for the programme: teams, rules and data come before models.
Explains the role: designing AI agents that understand a request, follow several steps, use tools and trigger actions.
Read more
Describes connecting AI to tools already in use — email, internal documents, tracking spreadsheets, CRM and communication tools. Gives a client-email example: instead of having someone read each message, find the relevant information and alert the right team, an AI agent can take on part of that work.
For brands planning a campaign
A ready-made module for a course or workshop, covering role definition, the tool landscape and a reproducible task-level example. The post reads as introductory teaching material rather than advanced technical instruction.
Enterprise integration: MCP as a universal connector
Sets out the connector problem in enterprise AI: linking an AI to a CRM such as HubSpot or Salesforce means building a bespoke connector, repeating the work for every tool and rebuilding when an update breaks the link.
Read more
Presents MCP (Model Context Protocol) as a single connection through which AI can pull data across all tools in one query. Says they first saw MCP at a conference run by Anthropic and Dust, and plan to deploy this type of technology at DataBird.
A clear contrast: the current integration pain, followed by the single-connection outcome and a first-person conference anecdote.
The DataBird rollout is described as an intention, not a documented live deployment.
For brands planning a campaign
Addresses enterprise integration with a current, named protocol. It offers a credible scenario for a workshop exercise, though DataBird’s rollout is described as a plan, not a finished implementation.
From tool to workspace, and the foundations it requires
Takes OpenAI's GPT-5.5 launch and the idea of a 'super app' unifying ChatGPT, Codex and an AI browser, and argues that an AI interface will centralise actions: analyse quarterly sales, prepare a committee synthesis, identify at-risk clients and draft a follow-up email.
Read more
Says tasks once reserved for technical profiles become easier to take on — analysing a complex file, comparing sources, structuring monitoring and creating a prototype.
Adds agentic coding: AI reading an existing codebase to fix a bug, write tests or document a feature.
Closes by arguing that new tools need reliable data, clear rules, targeted uses and trained teams; without those, even the best model is poorly used.
Turns a product announcement into organisational consequences, then lists the prerequisites for making good use of it.
No release date is stated in the posts here.
For brands planning a campaign
Useful for positioning a course around readiness — reliable data, clear rules, targeted uses and trained teams — rather than tool access alone. Ask for the post date before tying a campaign to specific release timing.
Says DataBird launched AI Agent Builder training sessions and welcomes the new learners.
Read more
In the MCP post, says they plan to deploy MCP-type technology at DataBird and report how it changes the team's daily routines.
For brands planning a campaign
Shows an existing training practice and a willingness to pilot the tools being taught, supporting a workshop pitch. Ask for learner outcomes, completion numbers and programme length before assessing the training’s track record.
The posts point to one business connection: DataBird, where Miny de Tornaco says Miny de Tornaco helped launch AI Agent Builder training sessions and refers to "our sessions," "our learners," and "our teams."
Read more
The posts do not say whether Miny de Tornaco owns, works for, or contracts with DataBird, so the relationship is commercial, but its type is unstated. Other brands appearing in the same posts — OpenAI, Anthropic, Dust, HubSpot, Salesforce and Apple — show up as news topics, a conference Miny de Tornaco attended, or hypothetical examples, not as relationships with Miny de Tornaco.
DataBird
Disclosed partnership
In a post explaining the AI Agent Builder role, Miny de Tornaco says they and colleagues at DataBird launched AI Agent Builder training sessions and wishes “our new learners” well.
Read more
A second post says DataBird plans to adopt MCP-style connections and refers to “our teams.” Ask whether they own, work for or contract with DataBird; the posts don’t specify.
Original disclosure
"En début de mois, on a eu le plaisir chez @Databird de lancer nos sessions de formation AI Agent Builder." / "On compte mettre en place ce genre de techno chez DataBird, je vous dirai comment ça transformera le quotidien de nos équipes."
Alexandre Miny de Tornaco writes about putting AI agents to work inside companies.
Read more
Start with the post explaining what an AI Agent Builder does: designing assistants that understand a request, follow several steps, use tools and trigger actions, then wiring them into the email, documents, spreadsheets and CRM a team already uses.
The MCP post explains why connecting AI to a tool such as HubSpot or Salesforce today means building a separate connector for every system and rebuilding it when an update breaks the link.
It presents Model Context Protocol as a single connection that survives those changes.
The GPT-5.5 post turns to readiness, arguing that reliable data, clear rules, focused use cases and trained teams matter more than access to the newest model.
Together, the posts offer a short, consistent look at enterprise AI, agent building and AI training.
For other voices on adjacent ground, look at Cédric Bidet, Marie Robin and Matthieu Rousseau.
Bidet tackles the same friction point: time lost opening a CRM, adding filters, exporting a spreadsheet and cross-referencing it with another tool, when an agent could fetch the selection or launch a campaign in natural language.
Read more
Both treat MCP and similar protocols as the plumbing that makes this possible, explaining them in plain rather than technical terms.
Bidet writes from inside Firmin.ai, a product business, and positions lightweight agents for small and mid-sized firms at a stated monthly price.
The posts make a sales and positioning argument.
By contrast, Alexandre Miny de Tornaco’s posts explain the subject through training sessions rather than offering a priced product.
Robin offers practical French-language analysis of how AI changes what agencies sell and how they win work.
Read more
Like Alexandre Miny de Tornaco, she writes for business leaders deciding where AI fits, not engineers, and grounds her argument in named examples rather than broad claims about transformation.
Robin focuses on agency growth: prospecting response rates, positioning, pricing models and unconventional agency strategies. Alexandre Miny de Tornaco focuses on the agent itself — what a builder does, how it connects to existing tools and what an organisation needs to benefit from it.
Rousseau makes the same foundational argument: AI depends on the data underneath it.
Read more
He names MCP and agent use as reasons semantic layers and trustworthy data now matter. Both write for French-speaking business and technology leaders weighing a purchase.
Rousseau covers the data side — platforms such as Snowflake and Databricks, data engineering and cost optimisation — from the perspective of consultancy Modeo. Alexandre Miny de Tornaco focuses on agent design and training, using everyday office tasks such as reading customer email or qualifying a lead as examples rather than platform architecture.
For a B2B AI training company selling a course or workshop on building AI agents, Alexandre Miny de Tornaco's French-language posts speak directly to people working on enterprise AI adoption.
Read more
They explain what an AI agent builder does, how tools like an MCP connector link AI to a company's CRM and internal systems, and what a new model release changes for teams and code.
The posts suggest a fit with business and technical decision-makers in a French-speaking market, but don’t establish their job roles or buying authority.
The only measured audience figure published here is 9,239 LinkedIn followers (observed 9 October 2026).
No post-level reach, impressions, engagement, geography, age, gender, seniority, occupation, or employer data are provided.
Treat the follower count as reach on one channel and the posts as a signal of topic-language fit.
Ask Alexandre for recent results on comparable topics and formats—posts about agent building and enterprise tool integration—and any audience survey covering seniority, function, and industry.
Alexandre Miny de Tornaco’s three LinkedIn posts do not present a published partnership or sponsorship offer: they include no rate card, sponsored format, pricing, disclosure policy, or booking or inquiry route.
Read more
They describe an ongoing company relationship, including launching AI Agent Builder training sessions at DataBird, planning to roll out Model Context Protocol-style connectors across DataBird’s tools, and connecting AI to workplace systems such as CRM, email, documents, and internal tracking.
This is a training and company relationship, not a purchasable package.
An AI training company planning a workshop on agent building and enterprise integration should ask directly about commercial terms.
Yes, this fits a B2B AI training company looking for a course or workshop on building AI agents and integrating them into enterprise systems.
Read more
The three posts read like teaching material in miniature.
One defines the AI Agent Builder role and walks through a client-email process in which an agent reads the message, finds the information and notifies the right team instead of having a person do each step by hand.
Another tackles the connector problem: a bespoke link for each tool, such as HubSpot or Salesforce, must be rebuilt whenever either side updates.
MCP, by contrast, is presented as a single connection that lets AI pull data from all tools in one query.
A third turns a model release into an organisational checklist: what changes, which tasks become possible, and what reliable data, clear rules, targeted use cases and trained teams must be in place first.
They also announce that DataBird has launched AI Agent Builder training sessions and plan to implement MCP-style technology there, then report how it changes the team's routines.
That puts them at the intersection of agent design and enterprise integration.
The posts are French-language text for a French-speaking business and technical audience, on one channel with 9,239 followers (observed 9 October 2026).
Agree on language and market rather than assuming they will suit another one.
No rate card, format list, pricing or booking route is published, so discuss terms directly.
Ask for comparable training results before scaling: the posts give no learner outcomes, completion figures or programme lengths.
The MCP rollout at DataBird is a plan, not a finished deployment case study, so use it as an in-progress workshop scenario.
When to choose a specialist
A weak fit: an English-language developer-tools vendor wants an always-on paid video series — two vertical videos a week for three months, judged on reach and view-through rate, with brand scripts.
Read more
Every post is French text; none includes video, a transcript, a short-form clip or a visual asset, so the required format and its performance are unverified.
The only measured audience figure is 9,239 LinkedIn followers; there are no impressions, engagement rates, geography, seniority or job-function data to establish a baseline for reach or view-through targets.
The posts also describe no paid format, disclosure practice, pricing or booking route.
The volume expectation is a poor match for the work shown: the posts explain one idea at length, and one closes by asking the audience to test a claim, rather than offering repeated brand-scripted placements.
This does not rule out paid work generally, but a buyer would be testing an unverified format, language, cadence and measurement plan, not buying a proven package.
A worked campaign example
The posts support a practical workshop on moving from an AI assistant to an AI agent that uses your own tools.
Open with the role definition and client-email walkthrough.
Have participants map a routine that starts with a message and ends with the right team informed, then sketch the agent's steps.
Next, tackle integration, where the posts are strongest: explain the cost of bespoke connectors and how updates can break them, then present MCP as a single connection that enables one query across tools.
For the exercise, participants map the stack — CRM, documents, tracking sheets and internal communication — choose the first tool to connect, and note what breaks if either side updates.
Close with the release post's prerequisites: reliable data, clear rules, a targeted use case and trained teams must come before model choice.
Participants leave with one mapped process and a first-agent specification for their own stack.
Agree on learner outcomes, session length and reporting in the contract, including what the buyer receives afterwards and by when; the posts give no comparable programme results.
Settle the terms, delivery language and use of the DataBird training example as an in-progress scenario, not a completed deployment.
Choose Cédric for a workshop helping smaller businesses find practical AI-agent opportunities in everyday operations.
Read more
In a post about Firmin.ai, they describe lightweight agents for sorting email attachments, turning meetings into tailored task lists, and combining e-commerce data with CRM and sales goals. These examples offer buyers concrete processes and a pragmatic approach to fitting tools into existing workflows.
Alexandre is the closer match if the course must teach AI Agent Builder skills and enterprise integration.
They define the role, explain how to connect agents to business tools, and discuss MCP.
Cédric’s reviewed examples focus more on applying agents to operational problems than on teaching integration protocols or a course format.
Choose Matthieu when the workshop’s main goal is preparing an organisation’s data platform for AI and self-service use.
Read more
Their posts identify practical obstacles such as siloed metrics, poor data quality, manual work, scaling limits, and teams’ dependence on data specialists. They also connect semantic layers with MCP and AI agents, making this a stronger choice for a data-foundations session.
Alexandre is better suited to a session centred on what an agent does and how it connects to business tools. Matthieu’s reviewed work concentrates on data-platform challenges and technology choices, rather than directly explaining agent building or offering a training session.
Choose Marie if the buyer wants a workshop for agencies on applying AI to growth and client operations rather than a course on building enterprise agents.
Read more
Their post proposes a mix of clearer positioning, proof, customer experience, CRM use, and AI-assisted prospecting; another surveys alternative agency business models and ways to use AI in internal processes.
Alexandre is the stronger match for teaching agent concepts and enterprise tool integration. Marie’s posts focus on agency strategy and operations, not an AI Agent Builder curriculum or a demonstrated integration-training offer.
Three LinkedIn posts by Alexandre Miny de Tornaco are available here: an explainer on the AI Agent Builder role and workplace tasks; a post on MCP as a single connector for business tools; and a post on a model release and the organisational foundations it requires.
The posts are in French, and no publication dates are listed here.
Similar creators Cédric Bidet, Marie Robin and Matthieu Rousseau each have two posts available here.
Missing information that affects decisions includes no video or transcript material, no learner outcomes or programme details, no published partnership formats, pricing or booking route, and no audience data beyond a LinkedIn follower count of 9,239 observed 9 October 2026.
The DataBird MCP-style rollout is described as planned, not as a documented live implementation.