Why did we design our Custom GPTs differently?
At Albert, integrating generative AI has never been about following trends. We didn't add AI assistants to embellish our product or tick a marketing box. Our starting point was simple: what are the real pain points in the daily life of an HR team in charge of Strategic Workforce Planning (SWP)?
And how can we reduce or even eliminate them with AI?
The first temptation would have been to integrate a generic model such as ChatGPT into a text interface. This works, to a certain extent. You ask a question and you get a well-formulated, often relevant, sometimes brilliant answer. But in the context of SWP, this quickly reveals its limitations:
- The answers are not always consistent with the planning methodology used.
- The data handled is often unstructured and difficult to use directly.
- The user remains solely responsible for the prompt, which introduces variability, noise, and even errors.
So we made a different choice. A choice to specialize. At Albert, our AI assistants are designed as business agents. Their role is not to answer everything. Their role is to perform a specific task, in a clear context, with structured data.
A use case approach, not a technology one
We started with the use cases. What is the recurring need? What are the long, repetitive, or blocking tasks? When do HR teams waste time or go round in circles?
This is how we identified six very concrete cases:
- Create a skills repository: start with the job architecture to build a clear and consistent model.
- Extract a job architecture from existing data.
- Harmonize and clean up an existing but heterogeneous job architecture.
- Automatically generate job descriptions from available data.
- Identify HR performance drivers and structure them into actionable drivers.
- Generate a report from the SWP plan, ready for presentation to management.
For each of these pain points, we defined a business logic, a typical workflow, and the expected inputs/outputs. AI then steps in as an engine to accelerate and structure the process.
AI agents, not black boxes
Each AI assistant at Albert is designed as an agent. It has a mission, a framework, well-defined input data, and clear expected results. The user retains control over the data and parameters and can review the results. The agent does not replace the decision. It prepares the ground so that the decision can be made more quickly and with more context.
For example, the GPT Job Archi Cleaner does not rebuild an entire job architecture from scratch. It identifies duplicates, proposes harmonization, and suggests groupings. But it is the user who validates. The same goes for the SWP Report Generator: it organizes the data, reformulates it, and generates a distributable version. But each section can be reread, revised, and adjusted.
Seamless integration into the product
Our AI agents are not widgets added as an afterthought. They are integrated directly into the key stages of the Albert product. Their role is to shorten the path between data and action. No copy-and-paste. No switching to another tool. No friction.
It's this integration that makes the difference. When you generate a job description with Job Builder, it can be used immediately in the plan. When you identify drivers with Driver Builder, they are inserted into the simulation model.
Less general AI, more useful AI
Generating free text, without a framework or control, creates as many problems as it solves. At Albert, we have chosen a different path: guided, parameterized, contextualized AI. It doesn't answer everything. But it answers well, what it is asked, at the right time.
And that's exactly what our users want: not magic, but efficiency. Less surprise, more structure. Less generalities, more results.
Conclusion: AI that really saves time
Our Custom GPTs are the result of a strong belief: AI should not add complexity, but clarity. It should not create content for the sake of content, but produce what really helps move things forward.
In SWP, every hour counts. Every inconsistency is costly. Every gray area slows down decision-making. By providing structure, saving time, and improving readability, our AI assistants deliver what really matters to HR: useful decisions, made on time, with the right information.
Find out more about our custom GPTs in this article.
