Tell us about your project

Contact us Close
Industry Manufacturing

Accelerate Growth: AI-Powered Estimating

A Quebec leader in the manufacturing industry, this company set out to modernize an estimating process that was holding back its growth.

Challenge

Growth held back by a lengthy, complex estimating process

At this manufacturer, estimating a project is complex work that requires expertise. With more than 30 products—almost always custom-made—every project is unique. Producing an estimate takes dozens of hours: digging through catalogs hundreds of pages long, copying and pasting from multiple data sources, setting a price, validating a specification or a technical detail.

Goals

  • Speed up estimate production time
  • Support sales efforts
  • Centralize expertise
  • Reduce manual operations
  • Improve estimate accuracy
  • Maximize project profitability
Solution

AI-powered estimates

597-1-1-1

Centralized web-based estimating platform

Estimating expertise standardized from end to end. The estimator enters the project information, then the platform runs the calculations, generates an editable quote, and automatically produces a client-ready PDF contract. What used to take hours of copying and pasting now happens in a few clicks—consistently, no matter who does the estimating.

prompt-bar-en

Conversational AI agent

Instead of digging through files for hours to find a similar project, the estimator asks an AI agent, the same way they'd ask a colleague. The tool finds the right references, suggests the right calculations, and lets the expert focus on what truly requires their judgment.

ai-hub-en

Secure MCP server

This is where the real long-term value lies. All the formulas, pricing, technical specifications, and processes are centralized in a secure system. The tool can only answer based on these validated rules: no hallucinations, no exposed data, no undocumented expertise.

How we made it happen

Step 1

Target what will really pay off

Once you know you need to tackle the estimation process, where do you start? Instead of trying to transform everything at once, focus on a strategic case: a single piece of equipment, the most complex one. Why? Because if it works for that piece of equipment, it will work for everything else.

Step 2

Deliver measurable value in weeks, not months

Focusing on an initial scope allows us to get a working solution to users quickly. In this manufacturer’s case, estimators were testing the solution in just one month. This first step validates business rules, builds the foundation for the broader solution, and sets a clear roadmap for future equipment lines.

Step 3

Build fast, build smart

While Nexapp rolls out new equipment types one by one, the client documents business rules in parallel, gradually standardizing the entire process. The platform automates 80% of standard cases while leaving fields fully editable for exceptions. The goal is simple: eliminate repetitive work so human expertise is applied where it drives real value.

We deliver results

2000 h
recovered per year
12 X
aster at producing estimates
3 M$
in potential additional revenue
1.21
years payback period

AI, grounded in software engineering

Image pleine largeur cas dutilisation

Strict model guardrails

Structure and expose the estimation logic through an MCP server to isolate the mathematical formulas, preventing unauthorized changes and avoiding the context limitations inherent to large language models.
Image pleine largeur cas dutilisation (2)

Ensuring reliable outputs

The training data comes from real-world projects, which strengthens the model's accuracy and relevance.

Use Case Full Width Image (1)

Mixed-format data ingestion

Process historical project data across multiple sources and formats using a Python extractor.

Image pleine largeur cas dutilisation (1)

Scalable architecture

Optimizing storage in a vector-enabled Postgres database and design TypeScript services to allow new equipment types to be added in the future without a major rework.

What if AI did the same for you?
Right this way.

bg_shape bg_shape