In the ever-evolving landscape of construction technology, one topic has dominated conversations across jobsites, boardrooms, and industry conferences: Artificial Intelligence. But is AI just another buzzword, or is it truly transforming how we build? In a recent podcast conversation between Niran Shrestha, CEO and Co-Founder, and Jayesh Khushalani, Senior Product Manager at Kwant, we got some fascinating insights into how AI is reshaping construction management—and where the technology still has room to grow.
Beyond the Buzzword: Is AI Delivering Real Value?

When asked if AI is just a buzzword or something genuinely valuable, Jayesh was quick to clarify: "AI isn't just about technology. It's about amplifying human potential, making decisions safer and faster, specifically in construction where it's a high-stake environment with things changing dynamically."
This perspective cuts through the hype. While many companies are still figuring out where AI fits into their operations, the technology isn't just theoretical anymore. It's already enhancing capabilities in real construction environments.
What's particularly interesting is how many construction companies are approaching AI implementation. As Niran pointed out, even large general contractors are still defining the problems they want to solve with AI. They're avoiding the common pitfall of finding a solution before clearly understanding the problem—a refreshing approach in an industry sometimes eager to adopt new technologies without clear use cases.
Finding the Right Problems for AI to Solve

The podcast highlighted a critical insight about implementing new technologies: customers don't always know what they want. Referencing the famous Henry Ford quote about "faster horses," Jayesh explained: "When a customer said they want a faster horse, it actually means they want to reach that place faster. It doesn't necessarily mean you have to build a faster horse."
This philosophy guides how construction technology companies should approach AI development. When customers request features, the real skill lies in identifying the underlying problem rather than building exactly what was requested. It's about translating "I need a faster horse" into "you need more efficient transportation."
For construction professionals evaluating AI tools, this means focusing conversations on your core challenges rather than specific technologies. The best technology partners will help you define those problems clearly before proposing solutions.
Fail Fast: The Right Approach to AI Implementation

One of the most practical insights from the conversation was about implementation strategy. Rather than spending years perfecting an AI solution before release—like the cautionary tale of the Human AI Pin that took six years and $250 million before ultimately failing—Jayesh advocated for a "build fast, fail fast" approach:
"In a startup, the main thing has always been build fast, fail fast. Sometimes it's better to build a prototype, test it out, fail, revamp the prototype, build it and test it out and fail again so that you always have the option to revisit that cycle and incorporate customer feedback."
This iterative approach has proven successful for Kwant's "Ask Bob" feature, which allows users to verify wage and payroll compliance through simple queries. Users adopted it naturally without extensive training because the interface felt intuitive—more like asking a question to an assistant than learning a new software tool.
The Three Layers of AI Development
The podcast broke down AI development into three distinct layers, all of which are advancing simultaneously:
- Infrastructure Layer: GPU processing power, data centers, and the physical components that make AI possible
- Model Layer: Large language models like those from OpenAI, Anthropic, and Meta that provide the core intelligence
- Application Layer: The specialized tools built on top of these models for specific industry uses
For construction companies, the application layer is where the most immediate value lies. However, advancements in all three layers contribute to better construction-specific applications. The distributed infrastructure allows smaller companies to access powerful AI without massive investment, while improvements in base models enhance the capabilities of specialized applications.
Domain Expertise: The Missing Piece

A critical challenge highlighted in the conversation is that general AI models lack construction-specific knowledge. As Niran pointed out, "All these OpenAI, Anthropic, all these models, they do not have construction-specific domain expertise."
Overcoming this requires several approaches:
- Fine-tuning models with construction-specific datasets
- Creating specialized small language models focused on construction knowledge
- Using retrieval-augmented generation to pull relevant construction information before processing
Jayesh explained that this is similar to how professionals specialize: "Consider a large language model as a human with human-level intelligence. When you or anyone in the industry goes for their masters or PhD, they're basically fine-tuning themselves to that construction or domain-specific knowledge."
Companies with extensive historical construction data have a significant advantage here. By using proprietary datasets collected over years—like Kwant's safety, productivity, and compliance data—AI tools can provide more accurate and relevant insights for construction environments.
How AI Will Transform Construction Roles
Rather than replacing workers, AI is poised to augment various construction roles:
For Field Workers and Trades
AI can enhance safety and precision without replacing skilled labor. For painters, for example, AI can help match Pantone colors or alert supervisors when someone is working at an unsafe height without proper training.
For Superintendents
AI can serve as a decision support system, providing real-time alerts about scheduled tasks, resource allocation, and safety concerns. Imagine getting an automatic notification if a subcontractor hasn't arrived at a designated zone by a specific time—without having to physically check.
For Project Managers
This is where AI might have the most significant impact. Project managers spend considerable time creating progress reports, compliance documents, and risk assessments. AI can transform these tasks from hours-long processes to 20-minute reviews. As Jayesh explained: "The idea is that the two hours of work can now be cut down to 20 minutes. The project manager has to just read the document, correct the things they don't like... and they have a well-drafted document to share with the team."
The key takeaway? As Sam Altman from OpenAI famously said, "AI won't take your job—someone who knows how to use AI probably will." Construction professionals who embrace these tools will likely outperform those who resist them.
Facing the Challenges: Data Quality and Hallucinations
Despite the promise, AI in construction faces significant challenges. Two of the most pressing are data quality and hallucinations (when AI confidently produces incorrect information).
Construction data is notoriously unstructured and often siloed across different systems. Before AI can provide reliable insights, this data needs cleaning and integration. As Jayesh put it, "The better the data, the better the answers... It doesn't need to be 100% clean, but it still needs to be clean enough to avoid hallucinations."
The industry is making progress in reducing hallucinations, with some systems achieving 95-98% accuracy. However, that remaining 2-5% can be problematic in high-stakes construction environments where decisions impact safety and costs.
Learning from Other Industries
The podcast referenced Duolingo as an example of how established companies can thrive alongside general AI advances. When GPT-4 was released, many predicted it would disrupt Duolingo's language-learning business. Instead, Duolingo integrated AI features like role-playing conversations into their platform, enhancing their existing expertise rather than being replaced by it.
This pattern applies to construction technology as well. Companies with deep domain knowledge and specialized datasets will likely maintain advantages over general AI tools, even as those tools become more powerful. The combination of industry expertise and AI capabilities creates more value than either component alone.
What's Next for AI in Construction?
Looking ahead six months, Jayesh predicted we won't see drastic industry changes but rather increasing openness to AI adoption. Many general contractors are already building tech teams specifically focused on integrating AI into workflows.
We can also expect more emphasis on the ethical dimensions of AI use in construction: data protection, privacy, and responsible implementation. As these technologies become more embedded in daily operations, governance frameworks will become increasingly important.
Finding the Right Balance
The podcast title—"Humans, AI and Jobsite: Finding the Right Balance"—captures the central challenge facing construction professionals. AI isn't replacing human judgment and expertise; it's enhancing our capabilities and freeing up time for higher-value activities.
The most successful implementations will be those that:
- Start with clear problem definitions rather than technology-first approaches
- Take an iterative approach, testing and refining rather than pursuing perfection
- Combine general AI capabilities with construction-specific domain knowledge
- Augment human roles rather than attempting to replace them
- Maintain high standards for data quality and ethical use
For subcontractors and project managers navigating this changing landscape, the message is clear: AI tools are becoming essential components of efficient construction management. Those who embrace them thoughtfully—understanding both their capabilities and limitations—will gain significant competitive advantages.
The construction jobsite of tomorrow won't be fully automated, but it will be enhanced by AI co-pilots that help humans work smarter, safer, and more efficiently.
Want to hear more insights on AI in construction? Listen to the full podcast conversation between Niran Shrestha and Jayesh from Kwant on their discussion about artificial intelligence applications in the construction industry.