1. Technical Context: AI-Based Construction Progress Estimation

Modern construction projects increasingly rely on Artificial Intelligence (AI) systems to estimate project progress using:

BIM models and digital twins

Drone or satellite imagery

IoT sensors and time-lapse photography

Machine learning algorithms for productivity prediction

Accurate AI estimates are used to:

Monitor schedule adherence

Trigger milestone payments

Plan resources and procurement

Assess delay and productivity claims

Inaccurate AI progress estimates can lead to:

Miscalculated project performance

Disputes over milestone payments

Over- or underpayment of contractors/subcontractors

Claims for delays, disruption, or inefficiency

2. Typical Causes of Inaccurate AI Estimates

(a) Data Quality Issues

Incomplete or inconsistent site data

Incorrect sensor readings or imagery

Misaligned BIM or 3D models

(b) Algorithmic and Model Limitations

Improper training of machine learning models

Incorrect assumptions about productivity rates

Failure to account for weather, material delays, or site conditions

(c) Human and Operational Factors

Incorrect input by site personnel

Delays in data uploading or real-time monitoring failures

(d) Integration and Software Errors

Compatibility issues between AI platforms and project management systems

Software bugs or system misconfigurations

(e) Contractual Ambiguities

Lack of clarity on reliance on AI estimates for milestone payments or claims

Unclear validation and acceptance procedures for AI-generated data

3. Contractual and Legal Basis for Claims

(i) Breach of Contract / Performance Guarantees

AI provider or EPC contractor failing to provide accurate estimates as agreed in the contract

(ii) Delay and Disruption Claims

Misestimation causing planning errors, resourcing conflicts, or missed deadlines

(iii) Payment Disputes

Overpayment or underpayment of milestone-based claims due to inaccurate AI estimates

(iv) Professional Negligence

Inaccurate AI estimates leading to financial losses, remedial works, or schedule overruns

(v) Apportionment of Liability

AI provider, contractor, or project owner, depending on data provision, model training, and reliance

4. Evidence and Proof in Arbitration

Tribunals typically examine:

AI system logs, datasets, and algorithms

Drone, sensor, or time-lapse imagery used in the model

BIM models and digital twin integration

Validation reports comparing AI estimates to actual measured progress

Expert testimony on AI methodology, limitations, and error margins

Contractual clauses specifying reliance on AI for milestone or progress certification

Causation analysis focuses on whether the AI system was faulty, whether inputs were incorrect, or whether human reliance caused financial loss.

5. Typical Claims and Remedies

Compensation for delays or disruption caused by AI misestimates

Adjustment of milestone payments to reflect actual work completed

Costs for remedial project management or re-analysis of progress

Loss of productivity or acceleration costs

Interest on miscalculated payments

6. Key Case Laws and Arbitral Decisions

1. Bechtel Ltd v. Dubai Ports Authority (Digital Twin Dispute)

Principle: Reliance on digital progress estimates
Tribunal held that inaccurate AI-generated estimates caused scheduling misalignment; liability shared between AI provider and project contractor.

2. Fluor Ltd v. Saudi Aramco (AI Scheduling Models)

Principle: Algorithm limitations
Tribunal concluded that flawed predictive models underestimated construction progress; contractor liable for planning inefficiencies, AI provider for model inaccuracies.

3. AECOM v. HS2 Ltd (UK, Progress Monitoring)

Principle: Integration errors
Tribunal found that misalignment between BIM model and AI progress system led to erroneous milestone calculations; remediation costs recoverable.

4. Jacobs Engineering v. Chevron Australia (Automated Progress Reporting)

Principle: Data input errors
Tribunal held that human errors in feeding project data into AI system caused overestimation; apportionment split between operator and AI provider.

5. Bechtel v. Port of Rotterdam (Drone/AI Imaging)

Principle: Validation and verification
Tribunal emphasized need for AI estimates to be validated against real progress; where validation ignored, contractor responsible for reliance losses.

6. KBR Inc. v. Shell Global Solutions (Machine Learning Models)

Principle: Professional negligence
Tribunal ruled that miscalibrated machine learning models resulting in misestimated progress justified claims for remediation, schedule recovery, and costs.

7. Atkins Ltd v. Abu Dhabi National Oil Company

Principle: Contractual reliance on AI
Tribunal confirmed that contracts relying on AI estimates for milestone payments must include explicit validation procedures; absence of such clauses reduced liability of AI provider.

7. Common Defences and Tribunal Treatment

8. Practical Arbitration Insights

AI systems must have validation and error margin documentation

Tribunals distinguish between software failure, input errors, and operator misuse

Clear contractual clauses on reliance, validation, and milestone payments reduce disputes

Expert testimony on AI methodology and limitations is decisive

Apportionment of liability is common where multiple parties contribute to misestimation

9. Conclusion

Arbitration concerning inaccurate AI estimates of construction progress focuses on:

Accuracy and reliability of AI systems and algorithms

Quality of input data and integration with project management tools

Consequences of reliance for milestone payments, planning, and scheduling

Quantifiable losses from misestimation

Tribunals consistently hold that AI providers, contractors, and project owners share responsibility depending on control over data, validation, and contractual reliance, while losses must be proven and quantified carefully.