Envision your future job site: autonomous bots doing layout, material staging and installations; drones conducting inspections; and office staff using virtual reality to interact with a digital twin comparing the plans to the as-built. For leaders of most construction firms, that future state for the field is not something they expect to see this decade.
In the office, AI agents are no longer a vision of the future. Agents are already being used to improve the user experience for construction business applications such as Procore, Salesforce and Microsoft PowerBI. And new developments have made it accessible for almost any business to tailor AI tools to specific trades, projects and the unique knowledge held by their organization. Adding a GenAI workstream to your technology road map is a crucial step in keeping your firm ahead in the face of a changing industry, tight margins, and growing competition.
Meet the New AI
Behind the layer that users interact with, there are a number of deeper technical concepts that executives need to be familiar with; the most important of those topics being retrieval augmented generation (RAG) and agent frameworks.
Early large language models (LLMs) were trained on public data and had knowledge cutoffs that limited their relevance in business use. Hallucinations were commonly observed and were cited as the reason for companies’ slow adoption of GenAI. Large companies could create, pre-train, or fine-tune a model with data more relevant to their use cases, but doing so required considerable investment and technology knowledge. And historically, construction firms haven’t been the first to invest in innovations where ROI is far from assured.
RAG solves these challenges and makes organization-specific tuning accessible by introducing two critical capabilities:
- RAG allows for LLMs to inference over both public and private data.
- RAG can inference over both structured and unstructured data.
Now, implementing the right solutions means that AI can generate output based on data specific to your business, and it does more than inference over basic text. It can draw from all your organization’s data, including emails, chats, PDFs and recorded meetings — even your construction management, ERP, BIM, and VDC datasets.
Additionally, organizations can use agentic AI to complete more complex tasks with less human intervention. Agentic AI helps reduce the burden on users who are counting on the promise of GenAI tools that can do more than simple one-off tasks. Unlike their non-agentic counterparts, LLM-enabled tools paired with agent frameworks can find information and use tools to complete difficult tasks without needing extra instructions from a user.
Initial experiences with ChatGPT were saddled by the limitations of early chat bots — users provided a message to it, and it generated an immediate response. When using the tools emerging in 2025, there is a notable difference in the agentic user experience.
Now, rather than a simple, immediate response, the agents can build a plan based on a prompt. Following that planning step, they can trigger a series of downstream actions that call other agents with access to internal organizational data. Those insights from an organization’s private data can then be combined with insights derived from public data that the model was trained on or that the model is retrieving from the public web. Another downstream agent might also prepare a list of questions for you to answer that will provide clarity to the agents building the output.
After building up a robust research context, often over the course of minutes or hours, the agent can deliver an output to the user that not only has greater relevance and accuracy but also has length, density, and citations with links to the source data used to inform the generated output.
With these capabilities, AI agents are nearing the point where they can output a completed deliverable such as a strategic plan, an RFP response or a legally sound contract that is ready to execute.
RAG & AI Agents in Action
The latest AI developments can help you go beyond helpful automations to transform the way you do business.
Previously, if you asked an AI tool to help build up budgetary estimates for a new data center, it would be able to inference public data to provide you with a framework for estimating. And while that framework would be a helpful starting point, it would also likely be outdated, too general to fully apply to your business or even contain inaccurate information.
Example of prompting ChatGPT without reference materials guiding the output. Information includes square footage, power consumption, direction on size and type of response (output), plus a note as to which guidelines/codes to follow. The prompt asks for a calculation of costs for plus a summarized total estimation in a table.Screenshot by Wipfli, captured February 2025
Without reference materials guiding the output, the result estimation of costs. As this example shows, it took ChatGPT 12 seconds to "reason." Small and easily missed, there's a note at the bottom stating that the program can make mistakes and reminding users to "check important info."Screenshot by Wipfli, captured February 2025
Now, with RAG and AI agents, your company can provide the same prompt and get a more functional output, especially when it can seek clarifications from members of the team.
As an example, AI could help estimate a data center with specifications such as:
- 1,000,000 gross square feet
- High-voltage electrical system
- Evaporative cooling system
The AI tooling can inference over your company’s historical data, including prior estimates for buildings that had a similar purpose, size or requirements, giving you a more accurate output that’s relevant to your specific project needs. It may also ask the team to provide additional details related to site conditions, zoning requirements, sustainability goals or timeline constraints.
After churning through the downstream reasoning steps, the output can be formatted based on the style guidelines you’ve established for previous estimates. Lastly, learnings from this estimation collaboration can be captured in the model’s memory to help ensure even smoother collaboration on estimates in the future.
That same construction project prompt with Copilot where the AI is tasked with searching a specific source for relevant files to better inform the calculation and the output.Screenshot by Wipfli, captured February 2025
Estimated cost results using a retrieval augmented generation agent. Note the cost differences between the basic vs these outputs. Note that the agent still includes the reminder that "AI-generated content may be incorrect".Screenshot by Wipfli, captured February 2025
AI Use Cases
RAG provides an easier, cost-efficient way to custom-train the LLM that powers your AI agents. By configuring the agents to interact with your firm’s data, you can leverage it to improve your workforce’s effectiveness, retain talent, win better projects and protect against margin compression as peers and competitors inevitably invest in the same.
Potential AI use cases for your construction firm include:
- Project management: AI-driven planning makes complex scheduling easier by helping automate the process and factoring in critical information such as availability, skillsets, location, subcontractor relationships, and even team preferences. You can also use AI to identify patterns in your historical project data to help you be proactive about potential issues and opportunities.
- Talent management: AI can help improve talent retention and engagement — critical metrics in a competitive labor pool. With AI, you can use employee engagement data to identify turnover risk so that your HR staff can address it proactively. It can also be used to provide real-time performance analytics and personalized training, making leadership development easier.
- Design: Reduce time to design by using AI for easier rendering, editing, animations, background replacement, and other design capabilities. You can even use generative AI to create multiple design alternatives based on specific project parameters.
- Contracts: AI tools can help you identify and mitigate contract risk, standardize contract review,w and give your team the insights they need to maintain contract compliance.
- Decision making: Leverage AI to surface critical, data-driven insights for decision-making, including how you can optimize marketing spend, make targeted business development decisions, and develop an early warning system for profit fade.
