Most estimating tools guess the number.
This one shows the reasoning behind it.

Paste an RFP. Watch the system extract scope, surface load-bearing assumptions, retrieve historical comparables from the firm's archive, and generate a structured estimate with the logic visible at every step. The estimator's judgment is the spec — not a black box.

01Input · RFP / Scope

Ready

02Reasoning Trace

Idle

◇ Waiting for input.
Every reasoning step appears here in sequence — extraction, assumptions, comparables, line-item build.

Generated Estimate · Draft 01

Project

Estimated Total

Range: $0 – $0

Line Item	Qty	Unit	Rate	Total

Confidence

— —

Risk Register

Coverage

Why the demo is structured this way

The agent is the visible layer. Decision Architecture is the spine underneath.

Estimating doesn't fail because the math is hard. It fails because the senior estimator's judgment — the rules about when to escalate, what assumption to make under ambiguity, which historical job is a true comparable — lives in one person's head. Before any AI ships, that judgment gets encoded. The five layers below are how each recurring decision in the estimating workflow gets specified before a single line of agent code is written.

LAYER 02

Trigger

What event causes the estimating decision to fire — an RFP arrival, a scope change, a field note from the foreman.

LAYER 03

Frame

How the project is understood — building type, trade scope, complexity bands, geographic factors.

LAYER 04

Logic

The criteria and weights — unit rates, productivity factors, overhead, the estimator's adjustment rules.

LAYER 05

Move

The committed action — generate estimate, escalate to senior, request clarification, decline to bid.

LAYER 06

Echo

What the decision leaves behind — won/lost outcome, variance from actuals, comparable for the next bid.

For a partner conversation

Same team encodes the estimator's logic, then ships the working agent.

What you just watched is the engagement model — Discovery to Build to Enablement, no spec drift between rooms. The Decision Audit surfaces every recurring estimating decision in the firm. Agent Engineering turns the encoded ones into working systems. The Academy trains the team to use them. Happy to walk through scope, sequencing, and what the first 30 days look like.

Ryan Rezeika · ResonAi · ryan@resonai.ai · resonai.ai