Introduction: Understanding AI Adoption Levels

Recently, through out my extensive travels, I read a thought-provoking article in the Harvard Business Review by Scott Cook, Andrei Hagiu, and Julian Wright, titled “Turn Generative AI from an Existential Threat into a Competitive Advantage”. They describe three levels of AI adoption, a framework that resonated deeply as I considered how our own construction industry is progressing with AI.

Here’s how they define the levels of AI adoption:

• Level 1 – Adopt Publicly Available AI Tools: Use general-purpose AI tools (e.g., ChatGPT, Midjourney) to enhance productivity and efficiency. A good starting point, but this soon becomes standard practice“table stakes”, as others catch up.
• Level 2 – Customize AI with Proprietary Data: Train AI models on your own company’s data and expertise to create tailored, high-value solutions competitors can’t easily replicate.
• Level 3 – Create Continuous, Automatic Improvement Loops: Develop AI systems that improve themselves continuously through real-time feedback, creating a self-reinforcing competitive advantage.

In construction, while the conversation about AI is gaining momentum, practical implementation remains largely superficial. Most construction firms linger at Level 1, cautiously experimenting with general-purpose tools without moving toward deeper integration. Few have reached Level 2. Level 3 is largely aspirational, but achievable.This hesitation isn’t without risk, as remaining at this entry-level adoption will inevitably become insufficient in maintaining competitiveness.

In this article, I will reflect on how these levels of adoption apply specifically to construction, what realistic steps firms can take to progress through them, and how our industry can methodically harness AI as a true strategic asset, transforming it from merely a trending buzzword into genuine business advantage.

Now imagine this: It’s 2028, and two major construction companies compete for a massive infrastructure project. One effortlessly integrates AI-driven solutions, precisely forecasting risks and streamlining operations, while the other struggles, still reliant on traditional tools. Guess who wins?

Level 1 – Starting Smart: Leveraging Existing AI Tools

If you’re not using widely available AI tools already, you’re behind. Here’s how to catch up, and why it’s essential.

Why Level 1 is Essential, But Only the Beginning

Adopting tools like ChatGPT, Midjourney, or Runway can streamline tasks such as proposal writing, project visualization, and communications. For many firms, these tools unlock early efficiency gains and reduce administrative overhead.

Quick Wins at Level 1:

• Faster document drafting (proposals, contracts, internal reports),
• Quicker design brainstorming and visualization;
• Improved client communications.

Yet, while useful, these gains are easy for competitors to replicate. There are also many hidden risks, as I would call them: Convenience at the Cost of Control.

While Level 1 tools offer convenience, they come with significant risks:

• Data privacy exposure: Public AI tools process data externally. Even when using “private” versions, data may travel across servers or jurisdictions not compliant with construction industry or regional privacy standards (e.g. GDPR).
• Security vulnerabilities: Sensitive information from contracts, claims, or project documents can inadvertently be exposed or stored in systems outside your control.
• Regulatory non-compliance: For construction firms involved in government projects or operating in regulated markets, using open AI tools without strict data control can breach compliance obligations.
• Lack of accountability: Open tools provide little visibility into how data is processed or how outputs are generated, a serious liability for high-stakes projects.

Why This Matters for Construction

In construction, where documentation, claims, and compliance are mission-critical, handing over data to public AI platforms could lead to:

1. Legal risks;
2. Reputational damage;
3. Potential contract violations.

This is why the path to Levels 2 and 3, customized AI with strict data sovereignty becomes not just a competitive strategy but a safety imperative.

Level 2 – Building Differentiation: Tailoring AI with Your Proprietary Data

Why Your Proprietary Data Holds the Key

At Level 2, companies stop relying solely on off-the-shelf solutions. Instead, they develop AI tools tailored to their specific data and workflows, leveraging accumulated knowledge like past project data, claim records, and operational best practices.

First Practical Steps Toward Level 2

1. Audit your data: Identify where your valuable informations/evidence resides (emails, RFIs, schedules, drawings).
2. Consolidate: Centralize and standardize data in a single environment.
3. Implement structuring tools: Use secure systems like Lupa’s Data Management platform to fetch and organize data from 50+ sources.
4. Deploy tailored AI tools: Start applying Prompt Libraries and Semantic Search to begin extracting value.

AI becomes truly powerful when it speaks your company’s unique language.

Key Capabilities at Level 2

Several LUPA Technology clients, including top-tier consultants, international developers and leading contractors, have successfully transitioned to Level 2. They now benefit from:

• Centralized, structured proprietary data;
• Semantic and conceptual search;
• Prompt-based reports and insights customized for user queries;
• Model accuracy and fairness maintained through human oversight;
• Static models, no unsupervised learning to ensure reliability.

These incremental innovations have driven measurable efficiency gains and risk reduction. With Lupa, no client data is ever used to train models, ensuring privacy and minimizing costs.

Our Experience at Lupa

When reflecting on how to help our client reach Level 2 of the AI adoption faster, here at Lupa, one of the things we did is to develop our proprietary Prompt Library, drawing on years of industry-specific insights. This customization has significantly improved accuracy and reliability for clients in construction and legal sectors.

To maintain accuracy and fairness, LUPA Technology gives you:

1. Transparent, explainable AI: No black-box outputs.
2. Controlled environments: Each client’s data stays private, no cross-client data sharing.
3. Industry-tuned models: Designed for construction-specific insights, not general internet data.
4. Human-in-the-loop options: Clients can review and verify outputs before decisions.

Customization isn’t plug-and-play. Tailoring AI to your data unlocks differentiation, but requires commitment and a clear strategy.

*** A Note for Smaller Firms

Even without massive proprietary datasets, small firms can:

• Use industry-trained AI models (such as Lupa’s) that don’t require extensive retraining.
• Tailor search parameters, reports, and tagging systems to fit their workflows.
• Avoid the expense and risk of large-scale AI development.

Level 3 – The AI Advantage Frontier: Dynamic, Intelligent Workflows

The true power of AI in construction isn’t autonomous learning, it’s automating complex workflows and delivering insights when and where they matter most.

What Sets Level 3 Apart

At Level 3, AI automates data retrieval and insight delivery, transforming from a reactive tool into a proactive partner.

In industries like e-commerce or social media, Level 3 might mean self-learning AI models that evolve automatically. But construction’s challenges and risks require a different approach.

At this level, your AI should be able to:

• ·Actively fetches data from multiple platforms (emails, RFIs, drawings, schedules).
• Structures and enriches data automatically, extracting summaries, key concepts, and categorizing documents.
• Generates AI-driven reports and alerts, providing early warnings on risks, claims, and project delays.
• Uses semantic and conceptual search, making critical information accessible without keyword expertise.
• No autonomous model learning, maintaining strict compliance and traceability.

Why This Is Smarter for Construction

Autonomous, evolving AI tool (as seen in consumer tech) present serious risks in construction, where data integrity, traceability, and compliance are non-negotiable.

Dynamic, intelligent workflows, is what Lupa delivers:

1. Consistency and accountability;
2. Data privacy and sovereignty;
3. Actionable insights without unpredictable system changes.

Our Commitment to Data Security and Trust

At Lupa, we never train our models on our clients’ data. Each client operates in an exclusive, secure environment, with strict compliance to GDPR, ISO 27000, SOC2 and other relevant standards. Your proprietary data stays under your control, always.

In construction, AI’s role isn’t to learn unchecked. It’s to empower human experts with the right insights at the right time, all while safeguarding data privacy and compliance.

Common Pitfalls in AI Adoption (and How to Avoid Them)

AI integration often fails, not because of technology, but because of strategy. Here’s what to watch out for:

• Disorganized data environments: Without centralized, structured data, AI efforts fail.
• Over-automation mindset: Trying to leap to Level 3 without mastering Levels 1 and 2.
• Neglecting employee buy-in: Skipping training leads to resistance.
• Data privacy missteps: Choosing solutions that compromise data control or sovereignty.

Lupa addresses these risks with an adoption framework, strong client control over data, and AI tools that integrate smoothly into existing workflows.

Conclusion: From Theory to Competitive Advantage

Adopting AI isn’t just a technological step, it’s a strategic evolution. Construction companies that methodically move from Level 1 to Level 2, and thoughtfully approach Level 3, will be the industry leaders of tomorrow.

We’re at a pivotal moment:

AI isn’t just coming, it’s already here. The choice is whether you’ll harness its power strategically, or risk watching competitors pass you by.