Beyond the Verdict: How AI is Reshaping Dispute Resolution in Construction

The construction industry stands at a fascinating crossroads where artificial intelligence is beginning to transform everything from project planning to dispute resolution. We recently had the pleasure of hosting Robert Dean, Associate Director at Dialis based in Abu Dhabi, on the Project Flux podcast. You can listen to the full episode here . With his extensive experience in dispute resolution across arbitration, expert witness work, and project claims, Robert offered invaluable insights into how AI is reshaping conflict management in construction.

As someone actively experimenting with AI tools and even developing AI agents for dispute resolution, Robert brings a unique perspective that balances technological enthusiasm with practical realism. His thoughts on everything from model selection to blockchain-based dispute systems paint a compelling picture of what's possible today—and what's coming tomorrow.

The AI Model Landscape: Finding the Right Tool for the Job

One of the most striking aspects of our conversation with Robert was his nuanced approach to AI model selection. Rather than pledging allegiance to a single platform, he moves fluidly between different models based on their strengths.

"I tend to find myself sort of jumping between the different models for different tasks," Robert explained, noting his recent shift from primarily using OpenAI's models to finding Anthropic's Claude (particularly the recent Sonic 3.7 release) more consistent for many of his use cases.

This "model hopping" approach reflects a sophisticated understanding that's increasingly necessary in today's rapidly evolving AI landscape. As Robert pointed out, "a week will go by and there'll be a new model that's launched that has varying capabilities to previous generations." This pace of change creates both opportunities and challenges for professionals trying to leverage these tools effectively.

While some may find this overwhelming, Robert sees it as essential to maintaining an edge: "If you want to have the edge in what you're doing, you basically need to be able to understand where the strongest one is gonna be." Whether it's coding assistance, document analysis, or web browsing, different models excel in different domains—and these strengths continue to shift with each new release.

Data Quality: The Foundation of Effective AI in Dispute Resolution

Perhaps the most crucial insight Robert shared was about the importance of data preparation—what happens before you even engage with AI tools. The principle of "rubbish in, rubbish out" applies more than ever when dealing with complex construction disputes.

"I think a lot of the time, the work that you do upfront is almost the most important component to it," Robert emphasised. This is particularly relevant in construction disputes, where documentation can be voluminous and often of varying quality. When dealing with context windows that limit how much information an AI model can process at once, the quality of that information becomes paramount.

Robert painted a vivid picture of the challenges: scanned documents from construction sites with poor image quality that generate useless characters, wasting precious context window space. The solution? "It's important that you basically spend the time upfront and understand how much you would improve the output of these processes from spending the time and sort of cleaning your data properly."

This insight resonates strongly with James's experience: "The people who come to me and say, 'Well, I've tried these models and they don't work'—it's because of the prompt, right? And it's because they're not putting the work in upfront." Effective AI use requires thoughtful preparation, clear context setting, and specific guidance on the desired output.

The Current State of AI in Construction Disputes

When asked about AI's potential to reduce conflict in the construction industry, Robert was cautiously optimistic: "It definitely could be a force for good." He envisions AI capabilities eventually being embedded into the key tools used on construction sites, though he acknowledges this integration is still in its early stages.

"I think inevitably, these forms of intelligence are going to be embedded into the key tools that are used on site and they're only going to get better," Robert predicted. While he doesn't expect contractors to train their own AI models anytime soon, he anticipates that software providers will increasingly incorporate AI capabilities into their existing platforms.

Interestingly, Robert noted that AI could cut both ways in dispute resolution: "It might help avoid them potentially or it might help create them because certain things are sped up by the process." This dual potential reflects the transformative nature of the technology—it doesn't simply automate existing processes but can fundamentally change how disputes arise and are resolved.

James shared a personal example of using AI to successfully contest a parking ticket, highlighting how AI is democratising access to legal reasoning that previously required expensive professional advice. While this example might seem trivial, it illustrates how AI is already changing the power dynamics in dispute resolution across all scales.

Reasoning Capabilities: Are We There Yet?

When Yoshi asked about using newer reasoning models to analyze disputes, Robert offered a measured assessment: "I think in their present form, they're not probably there yet." He estimated that current AI reasoning capabilities are "about 50% of the way there to human reasoning, depending on which benchmarks that you sort of prescribe to."

The challenge is particularly acute in construction contracts, where understanding requires the ability to jump back and forth between interconnected clauses and provisions. "For any sort of complex scenario... if it was a big complicated contract, I think you would almost certainly want a model that could basically read one thing and then be able to jump back and forth between other things," Robert explained.

While lawyers can "rest easy for the time being," the rapid advancement in reasoning capabilities suggests this gap is closing quickly. As benchmarks continue to improve, we may soon see AI tools that can navigate the complexity of construction contracts with increasing sophistication.

Blockchain and Dispute Resolution: A Glimpse into the Future

One of the most fascinating segments of our conversation explored Robert's volunteer work with Kleros, an Ethereum-based project focused on blockchain dispute resolution. This system uses game theory principles to create a decentralised approach to settling disputes—something Robert described as "incredibly smart stuff."

Robert and the president of Kleros even built a mediator agent (before GPT-4 was released) that could funnel disputes into their system. "It was using a system whereby if you engage in a mediation and the mediation wasn't successful, it would present options to the parties that would give them the ability to potentially push the dispute to Kleros for jurors to vote on the outcome," he explained.

This represents a true AI agent with autonomous action—not just a conversational interface but a system that can take meaningful steps in a dispute resolution process. While Robert doesn't see this approach replacing traditional dispute resolution entirely, he believes it fills an important niche, particularly for "low value, high volume disputes" that don't require extensive documentation review or specialised legal expertise.

The conversation naturally extended to the potential for smart contracts in construction. While Robert doesn't see cryptocurrency-based payment systems coming to construction sites anytime soon, he acknowledged it's a possibility in the longer term. James painted an intriguing vision of LIDAR scanners measuring site progress, with data recorded immutably on the blockchain, potentially transforming how payment evaluations are conducted.

The Rise of AI Agents in Construction

Robert's experience building a mediator agent places him at the cutting edge of what many consider 2025's defining AI trend: agentic systems. These autonomous tools that can take actions beyond just conversation represent the next frontier in AI application.

The barrier to entry for creating such agents remains relatively high, requiring API knowledge and programming skills. However, Robert highlighted how AI itself is lowering this barrier: "One of the great things with these language models is you can ask them how to do things like this. They're great tutors." His own coding abilities have improved dramatically through this iterative learning process with AI assistance.

As agents become more prevalent, Robert emphasised the need for robust "proof of humanity" systems: "I think inevitably we're going to need really robust proof of humanity systems... Otherwise, we're going to lose trust in what's real and what's been created by robots." This concern about authenticity and trust will only grow as AI capabilities expand.

The Human Element: Why People Will Remain Central

Despite his enthusiasm for AI's potential, Robert remains firmly convinced that humans will stay central to construction processes: "I don't think anything technical that involves multiple steps, that you're dealing with sort of complexity, I don't really see that dropping away anytime soon."

He grounded this perspective in the physical nature of construction: "Building is a very physical thing and it's not so much an IT thing, you know, it's the real world... It's inherently a collaborative process that involves humans to humans." While AI and eventually robotics will transform aspects of construction, the fundamentally collaborative nature of the industry ensures humans will remain essential.

This human-centric view extends to dispute resolution as well. While AI can help distill complexity and provide initial analyses, Robert sees humans remaining "in the loop" for the foreseeable future, particularly for high-stakes decisions and complex judgments.

Looking Ahead: Digital Twins and Beyond

Looking to the near future, Robert believes we'll see AI-powered digital replicas transforming construction management before physical robots appear on sites. "One thing that might come before that is digital replicas being made a lot quicker and slicker that track progress on site," he suggested.

The current bottlenecks to widespread BIM adoption—the personnel requirements and costs of maintaining these models—could be overcome by AI systems that build and update digital twins automatically based on site data. Robert drew a parallel to developments in gaming and video technology: "When you look at the technology that exists that Google are doing with video and computer games... how they're essentially able to generate their own worlds now that AI can build and actually engage in. It's incredible and that almost certainly will come to construction."

These AI-powered digital twins could dramatically improve efficiency while also providing an objective record of project progress—potentially reducing disputes before they even arise.

Navigating the AI Revolution in Construction Disputes

Our conversation with Robert Dean revealed a nuanced picture of AI's role in construction dispute resolution—neither a panacea that will eliminate all conflicts nor a distant future technology with little practical relevance today. Instead, we see a rapidly evolving landscape where thoughtful professionals are already finding ways to leverage AI's strengths while acknowledging its limitations.

The key insights for construction professionals are clear: invest time in data preparation, understand the strengths and weaknesses of different AI models, maintain human oversight of critical decisions, and stay curious about emerging technologies like blockchain and agentic systems.

As Robert put it when recommending the YouTube channel Machine Learning Street Talk: staying ahead of the curve requires continuous learning. The construction industry may be built on physical foundations, but its future will increasingly be shaped by digital intelligence—and those who understand how to harness it effectively.

Connect with Robert Dean on LinkedIn to follow his insights on AI and dispute resolution.

And as always, subscribe to the Project Flux newsletter for weekly updates on how AI is transforming the future of projects.