The company becomes a learning organization
A company accumulates knowledge with every decision, but this memory often remains scattered across documents and employees. Here is how decision-making heritage changes that, and what conversational AI allows you to do with it.
Temps de lecture estimé : X min
This article continues the "Why Harnest" series we are publishing this summer with Fabien and Sabine. After drawing a parallel between financial and non-financial assets, I am returning here to what these assets actually change once they are in place.
Companies accumulate knowledge throughout their development. Every project, every trade-off, and every lesson learned enriches their understanding of their environment. Yet, a large portion of this knowledge remains scattered across documents or within the memories of the employees who participated in the decisions.
A CSR team changing managers is a common example of this. The new person finds the figures, the indicators, and the published reports. What they do not find is why an issue deemed material the previous year is no longer so this year, or why a particular calculation method was preferred over another. This part of the decision lived in their predecessor's head, not in a system. It disappears with them.
Decision-making assets: a structured memory
Non-financial decision-making assets provide a different answer. They do not just store data, but also the relationships between that data, the decisions they informed, and the developments observed over time.
A decision is no longer an endpoint: it becomes a piece of knowledge that gradually enriches the company's assets, allowing organizations to learn from their own operations.
We created Harnest with this goal in mind as well: ensuring that a company's knowledge no longer leaves with the people who hold it.
Engaging with information rather than searching for it
Artificial intelligence is also transforming how employees access this information. Interfaces are evolving toward a more conversational approach: users can ask questions in their own professional language, request summaries, or compare different situations without needing to know exactly where the information is stored.
To provide a relevant answer, artificial intelligence must understand the relationships between pieces of information, identify their context, and retrieve their history.
With Companion, the LLM agent we introduced this summer, a user can ask, for example: "How has our exposure to this risk evolved since our last materiality assessment, and what explains this change?" Answering this question requires connecting several elements over time: the original materiality score, the decisions made in the interim, and the data that justified each adjustment. A keyword search in a spreadsheet cannot easily reconstruct this narrative.
Artificial intelligence does not replace human analysis. It facilitates the exploration of a knowledge base that has become too vast to navigate manually. Users no longer query a series of applications; they engage in a dialogue with their company's decision-making assets.
The risk of homogenization
A response to my previous post raised an interesting point: the risk that AI might homogenize everything, smoothing out organizations just as it smooths out language models themselves.
At the time, I replied that the reasoning could go further than expected: AI could also apply Taylorism to executive and management committees, making them more efficient but also more replaceable. If intelligence and experience become shared and commoditized by AI, the value of a leader or director can no longer rely on what they know. What makes an individual unique—who they are and what they bring beyond analysis—is what makes the difference.
Decision-making heritage does not protect against this shift; it facilitates it. It makes an organization's intelligence and experience more accessible to everyone, which shifts value elsewhere: toward those who know what to do with it, rather than those who simply possess it.