AI agent for machinery quote data collection
Objective
Speed up response time for quote requests and automate the information-gathering work, replacing a manually verified sales checklist of more than 20 questions related to request type, delivery location and customer data.
The challenge
Build a system that uses AI to interpret unstructured input, such as email requests, while remaining reliable enough for a high-value sales cycle. It must compare multiple data sources, including the customer's location, existing CRM records, information provided in the request and product-specific requirements.
The solution
We developed an AI agent integrated into CRM with a precise, high-ROI task: activate on each new quote request, evaluate its completeness by comparing the text of the request received by email with any information about the customer and delivery location already present in the CRM record, and collect what is missing.
The project includes:
- Automatic activation via webhook on each new CRM request, without any manual triage
- Semantic analysis of the request that compares existing data with the email content and evaluates completeness field by field, distinguishing truly usable data from evasive answers
- Dynamic form that asks the customer only for missing data, generated from a code-versioned field catalog reusable across different clients
- Direct write of collected data into CRM when the form is submitted, without personal information transiting through or remaining in third-party services
- Automatic assignment of complete requests to salespeople and routing of incomplete ones into a dedicated waiting state, always visible in CRM
- Automatic follow-ups and recategorization of unanswered requests
- Per-step observability for each agent run: inputs, outputs and retries visible for monitoring and debugging
- Only the completeness check uses AI; all downstream logic is deterministic, making the system predictable and verifiable while keeping API costs under control