From Expert-Dependent Quoting to AI-Assisted Order Preparation

The business delivered customized projects that required experienced staff to translate customer requests into practical equipment and resource requirements.

A seemingly straightforward customer enquiry could involve several decisions about technical specifications, equipment combinations, project scope and inventory selection.

Those decisions were essential to delivering the service correctly, but not every part of the process required the same level of expert judgment.

Every new enquiry started with substantial manual preparation. Customer requirements often arrived as descriptions of what the customer wanted to achieve rather than as a complete technical specification. Staff had to interpret those requests, determine the relevant equipment and construct detailed orders with numerous inventory line items. Even when projects shared common requirements, much of the equipment-selection process was repeated manually. This made preparing a new order time-consuming before the team could focus on its more specialized elements.

Experienced staff had become the operational bottleneck. Senior employees understood which equipment combinations worked for different situations, how to translate vague requirements and which exceptions needed special consideration. But that expertise was difficult to reuse consistently. Much of the decision-making knowledge existed in individual employees' experience rather than in a structured process that other team members could readily apply. As the volume of enquiries increased, the business remained heavily dependent on the availability of those specialists.

Routine equipment selection competed with creative work. The company differentiated itself through customized service delivery, technical problem-solving and its ability to translate customer expectations into practical solutions. Those activities required human creativity and judgment. However, the same employees were also responsible for selecting standard equipment, entering inventory items and assembling the repetitive portions of orders. Time spent repeating predictable selections was time not available for more valuable client-facing and project-design work.

The company needed greater capacity without removing expert control. Fully automating complex project decisions would have introduced unnecessary risk. The business needed a more selective approach: automate recurring preparation, make use of existing inventory information and reserve the final design and approval decisions for experienced people. The opportunity was to change where specialists spent their time, not remove them from the process.

We implemented an AI-assisted pre-sales workflow designed to turn customer requirements into a structured starting point for technical order preparation.

The system combines contextual information, multi-step AI reasoning, inventory validation and human review.

Context-aware customer requirement processing: The workflow brings together relevant information from customer communications and supported project documents, including structured files, PDFs and email correspondence. Instead of requiring employees to manually reconstruct all the requirements before starting an order, the AI agent uses the available project context to identify relevant needs and prepare the next stages of processing. This provides a more consistent starting point for equipment selection.

Multi-step agentic order preparation: We implemented an AI workflow that can work through several interconnected decisions rather than generating an order in a single response. The agent breaks down the available requirements, processes relevant details and selects appropriate actions for developing the draft. When additional evaluation is needed, the workflow can revisit earlier decisions or follow an alternative processing path. This iterative approach reflects the fact that technical orders often require several dependent decisions.

Inventory-grounded equipment selection: The system connects to the business's existing inventory and order-management environment. Equipment recommendations are grounded in the company's actual inventory records rather than freely generated product descriptions. The workflow maps proposed items to valid inventory references, supporting the preparation of draft orders that staff can review against the operational system. This keeps the AI's recommendations connected to the equipment catalog the business uses.

Validation of generated order details: The workflow includes defined validation checks to identify unsuitable or inconsistent outputs before they reach the final approval stage. Inventory references and proposed order details are checked against configured requirements. Items requiring further judgment remain visible for specialist review. These checks help constrain AI-generated suggestions, although they do not eliminate the need for professional verification.

Human-in-the-loop order finalization: The AI agent prepares the recurring portions of an order, but experienced employees retain responsibility for the final result. Specialists review the proposed equipment, adjust configurations, handle unusual requirements and apply their technical and creative judgment. The prepared order is a starting point, not an automatically approved commitment to the customer. This separation allows the business to automate predictable preparation while preserving the expertise required for customized service delivery.