When co-construction gives birth to the augmented sales engineer
At DARYL, we're convinced of one thing: industrial AI only has value when it's built with operational teams, as close to the ground as possible. The pilot run with Terres du Sud — and more specifically with Delta Sud, its agricultural-irrigation subsidiary — is a concrete demonstration of that.
This project was never thought of as a simple “AI integration”, but as a co-construction effort aimed at strengthening the sales engineer without distorting their domain expertise.
Goal of the pilot
Avoid spending time on technical sizing until the sale is secured. In agricultural irrigation, every project is specific: topography, water needs, regulatory constraints, equipment choices… The result: a lot of time spent on technical studies very early on, sometimes on projects that never close.
The shared ambition of Delta Sud and DARYL was clear: enable reliable pre-sizing and a realistic budgetary offer from the very first visit, to speed up the decision both for the customer and for the sales team.
An assumed co-construction approach
The project began with an on-site framing meeting, gathering the Delta Sud and DARYL teams around a whiteboard. The goal: map the real process for sizing an agricultural-irrigation project.
This phase connected the engineering method used by Delta Sud, existing tools (Irricad CAD, internal GIS data), supplier product catalogues and the prices actually applied in the field. Nothing theoretical: only concrete material drawn from the teams' experience.
From generic assistant to business assistant
Alpha phase — laying the foundations
Delta Sud shared its source data: product sheets, technical catalogues, internal Excel calculation models. These elements allowed us to build a first Alpha version of the DARYL assistant. We quickly hit the limits of a standard generative AI system: context-window depth, long-term memory in a conversation, calculation reliability.
Beta phase — making AI explainable and controllable
To avoid any “black box” effect, we quickly evolved to a Beta version where every technical parameter is visible as a chip in the application, every assumption is editable, and the sales engineer stays in control.
The real turning point came when the Delta Sud engineering method was described in natural language. That transfer of human know-how to the machine is what turned DARYL into a true business assistant rather than a generic tool.
DARYL: an expert assistant
At the end of the project, the DARYL platform now lets teams analyse customer needs faster from the very first visit, propose reliable hydraulic pre-sizing, and produce a realistic budgetary offer close to a full study.
The result: a clearer, more credible pitch and significant time savings for the teams. This pilot demonstrated a fundamental point: generative AI, when framed by human expertise, accelerates work without ever distorting it.
What's next?
Our platform can now be applied to any industrial domain that demands complex technical pricing, long sales cycles and heavy reliance on internal expertise.
At DARYL, we're convinced: the future of technical sales lies with augmented sales engineers, not replaced ones.
