AI agent for machinery quote data collection
The challenge
The system must use AI to interpret unformatted input, such as emails and free-form requests, while remaining reliable enough for a high-ticket sales cycle. It needs to compare data from multiple sources: customer location, CRM records, email content and the specifics of the requested product.
The solution
We developed an AI agent with a precise job: activate on each new quote request, evaluate completeness and collect only the missing information.
Activation and analysis
- Automatic activation via webhook on each new CRM request
- Semantic analysis of the email request
- Comparison between existing data and newly received information
- Field-by-field evaluation of real completeness
Data collection
- Dynamic form generated from a code-versioned field catalog
- Customer is asked only for missing data
- Direct CRM write on form submission
- Personal data does not transit through or remain in third-party services
Operational workflow
- Automatic assignment of complete requests to salespeople
- Dedicated waiting state for incomplete requests
- Automatic follow-ups and recategorization of unanswered requests
- Per-step observability on input, output and retries
Control and predictability
Deterministic logic remains downstream of the only AI decision point. This makes the system more predictable, verifiable and helps keep API call costs under control.