انتقل إلى المحتوى الرئيسي

AI Cost Estimation

The AI Cost Estimation module helps estimators produce defensible, fast cost estimates for tenders and feasibility studies — combining historical project benchmarks, parametric models, and Claude-powered narrative generation.

Tip. The module is mockable end-to-end for tenants without an AI API key — it returns reasonable estimates from the historical bench when Claude is unavailable.

Overview​

Three modes:

  • Parametric — area × unit rate by space type, calibrated from your historical projects.
  • BOQ-driven — line-by-line estimation using current rates from the catalog.
  • AI hybrid — combines parametric + BOQ + an AI narrative explaining assumptions and risks.
Screenshot: Cost estimation workspace — inputs, parametric output, AI assumptions panel

Key concepts​

Estimate. A versioned estimate record — EST-2026-0042. Has inputs (project type, area, location, complexity), outputs (cost range, breakdown), and assumptions.

Benchmark. A historical project's actual cost per m² (or per appropriate unit) tagged by project type, era, location, complexity. The system seeds 50+ UAE benchmarks; tenants can add their own.

Parametric model. A formula cost = f(area, type, location, complexity) calibrated from benchmarks.

Confidence interval. Every estimate produces a low / mid / high range based on benchmark variance.

AI assumptions. When Claude is enabled, the AI generates a structured assumptions narrative — what's included, what's excluded, key risks. The narrative is reviewed by the estimator before being attached to the tender.

Step-by-step: produce a parametric estimate​

  1. Estimation → New estimate.
  2. Pick the project type (residential, commercial, infrastructure, etc.).
  3. Enter area / quantity, location (emirate / city), complexity tier (1-5).
  4. Click Estimate — output shows low / mid / high in your tenant's base currency.
  5. Save the estimate.

Step-by-step: produce a hybrid AI estimate​

  1. From a parametric estimate, click Run AI hybrid.
  2. The AI pulls comparable benchmarks and produces a narrative explaining the estimate.
  3. Review the narrative and edit before attaching.
  4. Save — the hybrid estimate is now available for tender submission.

Step-by-step: convert estimate to BOQ​

  1. From an estimate, click Generate BOQ skeleton.
  2. The system creates a BOQ with sections sized to the estimate's breakdown.
  3. The estimator fills line-by-line rates.
  4. Final BOQ is locked into the tender.

Common tasks​

  • Compare two estimates: Estimation → Compare side-by-side.
  • Add a benchmark: Estimation → Benchmarks → New — your tenant's project actuals.
  • Recalibrate the parametric model: Estimation → Settings → Recalibrate runs nightly automatically.
  • Export estimate to PDF: standard PDF layout with assumptions, breakdown, charts.

Troubleshooting​

  • "Estimate range is suspiciously wide." — Limited benchmarks for that project type. Add more historical benchmarks.
  • "AI hybrid returned an empty narrative." — Claude API key missing or quota exceeded. Falls back to parametric only.
  • "BOQ skeleton missing sections." — Estimate breakdown is too high-level. Pick a finer-grained project template.
  • "My historical project doesn't show in benchmarks." — Project must be marked Closed with actual cost recorded.

Permissions reference​

PermissionWho needs it
ESTIMATION_VIEWBid team, PMs
ESTIMATION_CREATEEstimators
ESTIMATION_AI_RUNSenior estimators
ESTIMATION_BENCHMARK_MANAGEEstimating manager

See also: QTO, BOQ, Tenders, Predictive Analytics.