BASTION CAPITAL PROJECT MODULE
Part of Forge Bastion IWMS · Built in Rust

80% of capital projects exceed budget. Yours don’t have to

Bastion Citadel brings predictive cost intelligence, AI-driven risk detection, and real-time project controls to every capital project in your portfolio — from tenant fit-outs to ground-up construction. One platform replaces spreadsheets, monthly reports, and the optimism bias that costs the industry billions.

80%
Of capital projects exceed their original budget. The construction industry loses an estimated $1.6 trillion annually to poor productivity, rework, and cost overruns. AI-driven project controls are projected to reduce project costs by 20% while maintaining or improving quality — yet fewer than 5% of firms have achieved their AI implementation objectives.
8
Intelligence engines
22%
Closer to budget
20%
Cost reduction (AI-driven)
Real-time
vs. monthly reporting
The Capital Project Crisis

Capital projects fail in a predictable pattern: optimism at approval, blindness during execution, shock at completion. The budget was based on incomplete scope. The schedule assumed no surprises. The monthly report arrived 30 days after the problems it described. Change orders accumulated without cumulative impact analysis. And by the time leadership learned the project was over budget, the overrun was already baked in.

This pattern is not unique to poorly managed organizations. Industry data shows that budgeting and cost management is the top challenge facing project managers across the construction industry, driven by rising material costs, labor shortages, and the fundamental inability of monthly reporting cycles to detect variance in time to correct it. The firms pulling ahead in 2026 are replacing reactive monthly reports with continuous, real-time project controls that track costs daily, predict overruns before they materialize, and flag risks when they are still cheap to address.

The Overrun Anatomy

How projects fail in slow motion

Phase 1 — Optimism Bias
The project is approved at $24M based on conceptual estimates. Historical data shows similar projects averaged $29M. Nobody references the data because the approval process rewards confidence, not accuracy. The baseline is already 20% below reality before a shovel hits dirt.
↓ Citadel corrects: AI cost estimation analyzes historical data from comparable projects, adjusting for materials, labor market, and regional conditions.
Phase 2 — Scope Drift
Fourteen change orders in the first four months. Each is individually justified. Nobody is tracking cumulative impact. The project manager approves each change against the original budget line — not against the already-depleted contingency. By month five, the contingency is exhausted but nobody has declared an overrun.
↓ Citadel corrects: Change order AI predicts cost and schedule impact before approval, with cumulative exposure tracking against remaining contingency.
Phase 3 — Reporting Latency
The monthly report shows the project is “on track” because it measures planned activities completed — not cost trajectory. The actual spend-to-date vs. earned value diverged 6 weeks ago. By the time the next report captures it, the variance has compounded into a structural overrun.
↓ Citadel corrects: Daily cost tracking with predictive burn-rate analysis. Variance detected in days, not months.
Phase 4 — Schedule Compression
The project is behind schedule. To recover, crews are doubled. Overtime is authorized. Acceleration costs 2–3× normal labor rates and introduces rework risk. The schedule recovery costs more than the original delay would have. Nobody modeled the tradeoff because the tools don’t support what-if analysis.
↓ Citadel corrects: AI schedule optimization tests millions of task sequences. Models cost of acceleration vs. delay before committing.
Phase 5 — Post-Completion Reckoning
The project closes at $31.2M — 30% over the approved $24M. The post-mortem identifies the same failures as the last five post-mortems: inadequate initial estimates, uncontrolled change orders, latent reporting, and no predictive risk management. The lessons are documented. They will be ignored on the next project.
↓ Citadel corrects: Cross-project learning engine. Every completed project trains the AI to improve estimates, risk detection, and schedule optimization for the next one.
Intelligence Engines

Eight engines. One project command center.

01
Predictive Cost Estimation
Historical analysis · Market-adjusted pricing · Confidence-banded forecasts
▼
The most expensive mistake in capital project management happens before construction begins — the estimate. Traditional estimates rely on conceptual design, past experience, and institutional optimism. Citadel’s AI cost engine analyzes historical cost data from thousands of comparable projects, adjusted for current material prices, regional labor rates, supply chain conditions, and regulatory requirements. The output is not a single number but a confidence-banded forecast: a most likely cost, an optimistic bound, and a pessimistic bound with quantified probability for each.
Comparable project analysis — AI matches proposed projects to historical completions by type, size, region, complexity, and building system scope. Identifies relevant cost benchmarks from real delivery data
Market-adjusted pricing — integrates live material cost indices, labor availability data, and supply chain lead times. Adjusts estimates for current market conditions, not last year’s averages
Confidence-banded forecasts — delivers P50 (most likely), P80 (conservative), and P95 (worst case) cost projections. Gives leadership a range to underwrite, not a false-precision single number
Optimism bias detection — flags estimates that fall below historical norms for comparable scope. Quantifies the statistical probability of delivering at the proposed budget
22%
Closer to final cost (vs. traditional)
P50/P80/P95
Confidence-banded output
Live
Material price integration
1,000+
Comparable projects in model
02
Real-Time Cost Controls
Daily spend tracking · Burn-rate analysis · Earned value intelligence
▼
Monthly cost reports are autopsies. By the time they arrive, the money is spent and the variance is structural. Citadel replaces monthly reporting with continuous, daily cost tracking that compares actual expenditure against earned value, budget baseline, and projected cost at completion — updated as invoices are processed, labor is logged, and materials are received. The AI analyzes burn rates and projects likely cost at completion using current spending patterns, flagging divergence from plan within days rather than months.
Daily cost capture — field teams log labor, materials, and equipment through mobile tools. Invoices processed through AI-driven OCR. Cost data flows into dashboards as activity occurs, not when reports are compiled
Predictive cost-at-completion — AI projects final project cost using current burn rates, remaining scope, and historical completion patterns. Updated daily with confidence intervals
Earned value intelligence — tracks CPI (Cost Performance Index) and SPI (Schedule Performance Index) continuously. Alerts when indices cross warning thresholds with recommended corrective actions
Threshold-based alerts — configurable alerts fire when spending nears approved thresholds by cost code, trade, or project phase. Alerts trigger before overruns, not after
Daily
Cost tracking (vs. monthly)
94%
Cost-at-completion accuracy
3–5d
Variance detection (vs. 30–45d)
75%
Faster reporting generation
03
AI Schedule Optimization
Neural-network sequencing · What-if simulation · Critical path intelligence
▼
Construction schedules are complex dependency networks where a single delayed trade can cascade through the entire critical path. Traditional CPM scheduling calculates one path. Citadel’s neural-network scheduler tests millions of task sequences to find the optimal construction path — considering trade dependencies, resource availability, weather windows, material lead times, and site logistics constraints. When delays occur, the AI dynamically reschedules and models the cost implications of acceleration vs. extension before the team commits to either.
Neural-network sequencing — tests millions of possible task orderings to identify construction paths that reduce total duration without adding labor or cost. Treats scheduling as an optimization problem, not a linear plan
What-if scenario engine — when a delay occurs, models the cost and schedule impact of multiple recovery strategies: acceleration, resequencing, scope deferral, or managed extension. Quantifies tradeoffs before decisions are made
Weather and supply chain integration — ingests weather forecasts, material lead time data, and subcontractor availability to identify scheduling risks 4–6 weeks in advance
Cross-project learning — duration estimates calibrated against actual delivery data from completed projects. Eliminates the optimism bias in schedule assumptions that plagues manual planning
Millions
Task sequences tested
18%
Average schedule compression
4–6wk
Risk anticipation horizon
Real-time
Dynamic rescheduling
04
Change Order Intelligence
Scope impact prediction · Cumulative exposure tracking · Approval governance
▼
Change orders are the silent killer of capital project budgets. Each one is individually reasonable. Collectively, they consume contingency, extend schedules, and transform a well-planned project into a cost overrun. Citadel’s Change Order Intelligence engine does what project managers cannot do mentally: track the cumulative impact of every approved and pending change against remaining contingency, schedule float, and final cost projection — in real time, before each approval decision.
Pre-approval impact analysis — before a change order is approved, AI predicts its cost, schedule, and downstream scope impact. Models interactions with previously approved changes for cumulative exposure
Contingency burn tracking — real-time visualization of contingency depletion rate. Alerts when cumulative changes consume contingency faster than project progress, indicating hidden overrun risk
Scope creep detection — AI scans change order descriptions, meeting minutes, and correspondence to identify patterns of incremental scope expansion that individually seem minor but collectively redefine the project
Governance workflow — configurable approval routing by change order value, type, and cumulative exposure. Escalates to senior leadership when cumulative changes exceed defined thresholds
38%
Change order cost reduction
Real-time
Cumulative exposure tracking
AI
Scope creep pattern detection
100%
Pre-approval impact analysis
05
Predictive Risk Detection
Correspondence scanning · Early warning signals · Risk register automation
▼
Project risks do not emerge without warning. They leave traces in meeting minutes, RFI responses, submittal delays, inspection failures, and contractor correspondence weeks before they become schedule impacts or cost overruns. Citadel’s Predictive Risk engine uses natural language processing to scan all project documentation — emails, meeting notes, daily logs, RFIs, and submittals — for early warning patterns that signal emerging risks. When the AI detects repeated mentions of “supplier delay” or “pending approval,” it flags the risk before the delay materializes on the schedule.
NLP correspondence scanning — analyzes emails, meeting minutes, RFIs, and daily reports for risk signal patterns: supplier delays, inspection concerns, subcontractor capacity issues, material availability warnings
Dynamic risk register — risk register updates continuously from live project data rather than quarterly manual reviews. Probability and impact scores adjust as new signals emerge or risks are mitigated
Cross-project risk patterns — identifies risks that recur across the portfolio. When one project experiences a subcontractor performance issue, the AI flags the same subcontractor on other active projects
Mitigation recommendation — for each detected risk, the AI proposes mitigation strategies based on responses that worked on similar risks in past projects. Converts risk detection into risk resolution
85%
Risks detected before impact
3–6wk
Average early warning lead time
NLP
Document scanning engine
Continuous
Risk register updates
06
Draw & Payment Management
Invoice automation · Lien waiver tracking · Multi-source capital allocation
▼
Capital project funding rarely comes from a single source. Equity, mezzanine debt, senior debt, grants, and tax incentives each have their own disbursement rules, documentation requirements, and drawdown sequences. Citadel’s Draw Management engine enforces capital waterfall logic at the transaction level — ensuring funds are drawn in the correct sequence, lien waivers are collected before payment, and inspection requirements are met before disbursement. AI-powered invoice processing reviews draw requests in minutes with 99%+ accuracy, routing exceptions to human reviewers rather than requiring manual processing of every line item.
AI invoice processing — autonomous review of draw requests with AI classification of line items against budget codes, flagging discrepancies, duplicate charges, and scope-inconsistent costs. 99%+ accuracy on routine draws
Capital waterfall enforcement — tracks multiple funding sources per project with defined drawdown sequences and percentage allocations. Prevents out-of-order disbursement across equity, mezzanine, and senior debt
Lien waiver automation — collection, tracking, and verification of conditional and unconditional lien waivers for every payment. No disbursement without complete waiver documentation
Cash flow forecasting — projects draw schedules and payment timing against available capital, upcoming commitments, and interest reserve balances. Prevents funding gaps before they create project delays
99%+
AI draw review accuracy
3min
Routine draw processing time
100%
Lien waiver compliance
Multi-source
Capital waterfall tracking
07
BIM & Constructability Intelligence
Clash detection · MEP coordination · Design-to-field alignment
▼
Design errors caught in the field cost 10–100× more to resolve than errors caught in the model. Citadel’s BIM Intelligence engine integrates with Revit, AutoCAD, and IFC models to perform AI-enhanced clash detection and constructability analysis that goes beyond geometric collision checking. The AI identifies conflicts across mechanical, electrical, plumbing, and structural systems, then prioritizes them by cost impact and schedule criticality — so the design team resolves the $400,000 ductwork conflict before the $4,000 bracket misalignment.
AI-enhanced clash detection — identifies MEP/structural conflicts with cost and schedule impact scoring. Prioritizes resolution by downstream consequence, not just geometric overlap
Constructability analysis — evaluates design models for field execution feasibility. Flags sequences that require impossible installation order, insufficient clearances, or access constraints
Design-to-field alignment — compares BIM model against site progress captured by 360° cameras and drone surveys. Identifies deviations from design intent before they become rework
Carbon impact modeling — integrates with Bastion Meridian to calculate embodied carbon of design choices. Recommends lower-carbon material alternatives that meet structural and code requirements
10–100×
Field vs. model error cost ratio
92%
Clash detection rate (pre-field)
Revit/IFC
Native BIM integration
45%
Rework reduction
08
Portfolio & Cross-Project Learning
Multi-project dashboards · Resource optimization · Institutional memory engine
▼
Most organizations repeat the same capital project mistakes because there is no mechanism to transfer learning from one project to the next. The post-mortem is written, filed, and forgotten. Citadel’s Cross-Project Learning engine captures every estimate, variance, risk event, change order, schedule deviation, and final cost from every completed project and feeds it back into the AI models that drive estimation, scheduling, and risk detection for future projects. Every project makes the next one more accurate.
Institutional memory engine — every cost variance, schedule deviation, risk event, and change order from completed projects is captured, categorized, and available to inform future project planning
Portfolio-level dashboards — unified view of all active capital projects with comparative cost performance, schedule status, risk exposure, and resource utilization across the entire program
Resource optimization — identifies shared resources (project managers, inspectors, specialized trades) across concurrent projects and optimizes allocation to minimize conflicts and downtime
Contractor performance scoring — tracks contractor, subcontractor, and vendor performance across all projects: schedule adherence, change order frequency, quality ratings, and safety record. Informs future bid evaluations
Every
Completed project trains the AI
Portfolio
Unified project command center
15%
Estimation accuracy improvement (YoY)
0
Lessons lost between projects
Deployments

The projects that finished on budget

Healthcare System · $180M Campus Expansion · 3 Buildings
$180M campus expansion delivered at $183M — 1.7% variance on a 3-year program where the industry average is 30%
A regional healthcare system undertook a $180M campus expansion comprising a new patient tower, surgical center, and central energy plant. Citadel’s Predictive Cost Estimation engine flagged the original $168M budget as statistically unlikely based on comparable healthcare construction data — recommending a $180M baseline with $12M contingency. During construction, real-time cost controls detected a concrete cost escalation 18 days after subcontractor mobilization, enabling renegotiation before the variance compounded. The Change Order Intelligence engine tracked 47 change orders with cumulative impact analysis, keeping contingency consumption visible to leadership at all times. Final cost: $183M, 1.7% over the corrected baseline.
1.7%
Final cost variance (vs. 30% avg)
$183M
Final cost (from $180M baseline)
47
Change orders with impact analysis
18d
Cost escalation detection speed
University · 12 Concurrent Renovations · $94M Program
12 simultaneous building renovations delivered with zero schedule conflicts and $8.2M in cross-project savings
A major university needed to renovate 12 academic buildings during a single 14-month construction window between academic years. The complexity was not any single project but their simultaneity — shared contractors, competing material deliveries, overlapping campus disruption zones, and a single facilities team overseeing all 12. Citadel’s Portfolio engine sequenced all 12 projects simultaneously, optimizing shared resources, staggering deliveries to prevent site congestion, and scheduling noisy work around the three buildings that remained occupied. The AI identified $8.2M in bulk purchasing savings by consolidating material orders across projects and detected 6 subcontractor scheduling conflicts 4–8 weeks before they would have caused delays.
12
Concurrent projects managed
$8.2M
Cross-project savings
0
Schedule conflicts at execution
14mo
All 12 completed on time
Commercial REIT · Tenant Fit-Out Program · 340 Projects/Year
From 22-day average tenant fit-out cycle to 14 days — while reducing cost variance from 18% to 4%
A national commercial REIT executing 340+ tenant fit-out projects annually was losing tenants during the gap between lease signing and occupancy. Average fit-out took 22 days with 18% cost variance — driven by inconsistent estimating, manual change order processing, and no institutional learning between projects despite executing hundreds of similar scope renovations each year. Citadel’s cross-project learning engine trained on 2 years of historical fit-out data and produced standardized cost models by space type, square footage, and finish level. Real-time cost controls eliminated the 30-day reporting lag. The result: 14-day average fit-out cycle, 4% cost variance, and tenant satisfaction scores that improved lease renewal rates 8 percentage points.
22d→14d
Fit-out cycle reduction
18%→4%
Cost variance improvement
340+
Annual projects optimized
8pt
Lease renewal rate increase
From the Field

Our original budget was $168 million. Citadel told us — based on historical data from comparable healthcare construction — that the statistical probability of delivering at that number was 12%. We rebased to $180 million. We finished at $183 million. In healthcare construction, where 30% overruns are normal, a 1.7% variance is not just good project management. It’s a different category of outcome.

VP, Capital Projects & Facilities
Campus Development & Construction
Regional Healthcare System

We execute 340 tenant fit-outs a year and we were learning nothing from one to the next. Every project started from scratch. Citadel trained on two years of our data and now every new estimate benefits from every project we’ve ever completed. Our cost variance went from 18% to 4%. Our cycle time dropped 36%. The AI didn’t just improve our process — it created institutional memory we never had.

Director, Tenant Improvements
Construction & Project Delivery
National Commercial REIT

Twelve simultaneous renovations on an active campus. If any one project slipped, it would cascade into the academic year. Citadel detected six subcontractor conflicts 4–8 weeks before execution. We resolved every one without a single day of schedule impact. The bulk purchasing optimization alone saved $8.2 million. That is what portfolio-level project intelligence looks like.

Associate VP, Facilities Planning
Capital Program Management
Major Research University
80%
Projects that exceed budget
22%
Closer to final cost
1.7%
Best deployment variance
$8.2M
Cross-project savings
Build with Certainty

Your next project deserves a different outcome

Schedule a demonstration of Bastion Citadel — configured for your capital program, your project types, and your delivery challenges.

Or contact our capital project team at citadel@brindwell.com