JAVIER OLIVIERI, home
ESEN
Book a call
IMPLEMENTED CASESTONE FABRICATION · SALES

From floor plans to quotes: an agent for D'Angelo

I designed and built a conversational agent that reads floor plans, applies the business's commercial rules, checks material stock and generates the quote as PDF and Excel, replacing a manual process.

kitchen_plan.pdf
AI · reads dimensions, pieces and material
Code · applies rules and calculates
QuotePDF · XLSXMaterialPieces and cutsFinishesTotal
CLIENT
D'Angelo Marmolería
SECTOR
Stone fabrication · sales
STATUS
Implemented
MY ROLE
Discovery, architecture, development and integration

Context and problem

Quoting depended on one person and a handful of spreadsheets

Every quote required reading the plan, calculating pieces and cuts, checking materials and putting the document together. Each new inquiry competed with the rest of the work.

Before

  • Someone read the plan and calculated pieces and cuts by hand
  • Prices and rules spread across the team and spreadsheets
  • The document was assembled manually

Now

  • The agent talks with the customer and gathers the data
  • AI reads the plan; code applies rules and checks stock
  • The quote comes out as PDF and Excel, ready to review and send

The sequence

From floor plan to quote, in five steps

  1. CODEPlan receivedArrives with the inquiry
  2. AIAI interpretsDimensions, pieces, material
  3. CODEBusiness rulesPricing and stock
  4. CODEValidationConsistency checks
  5. OUTCOMEQuotePDF and Excel

AI versus code

AI understands. Code validates.

AI

Understand

  • Reading plans with vision
  • Guiding the conversation and asking for what is missing
CODE

Calculate and validate

  • Business rules and pricing
  • Stock lookup
  • Consistency checks
OUTPUT

Deliver

  • Quote as PDF and Excel
  • Team review before sending

“The model reasons, the code validates.” No quote depends solely on what a model interprets.

Technology

What it's built with

Claude API VisionTool callingNext.jsTypeScriptGoogle SheetsPDF / Excel

My role

  • Discovery with the team and process mapping
  • Design of the architecture and the review points
  • Development of the agent, the rules and document generation
  • Integration and production rollout

Lessons

What I take away from this project

Technical lessons

  • Separating interpretation from calculation makes every error detectable and fixable.
  • Business rules written in code are easier to audit than ones buried in a prompt.
  • Human review at the end builds the confidence to adopt the system.

Business lessons

  • Start with the process that eats the most time, not the flashiest one.
  • Involving the person who quotes from the start avoids building something that goes unused.
  • Measuring before and after is what lets you decide whether to scale.

D'Angelo Marmolería and its brand belong to their owners. My role was as an engineer and developer. The illustration is my own schematic and does not show the real interface.

Does your company still quote by hand? Let's talk.

Tell me how you put quotes together today and we'll see which part is worth automating.

Book 30 min