Skip to content
Type
Civil Engineer blog

AI can help you with routine engineering tasks — here’s how

Date
31 January 2025

Introducing AI agents, advanced software components that can handle unpredictable questions and adapt on the fly. Dr Mike Rustell, Brunel University lecturer and director of Inframatic.ai, explains.

AI can help you with routine engineering tasks — here’s how
AI agents can help engineers with some of their work- and time-intensive tasks. Image credit: Shutterstock

Let’s face it—as civil engineers we spend a large part of our time on unglamorous tasks.

From scanning PDF files for relevant details to manually pulling out key parameters from historical design records. These repetitive tasks can be labour-intensive and tedious.

Civil engineers are increasingly turning to artificial intelligence (AI) to help.

But today’s AI tools can do far more than routine document checks or quick calculations.

Enter AI agents

What are AI agents?

Broadly speaking, an AI agent is an advanced software component built around a large language model (LLM)—like GPT-4o.

It can reason over data, plan sub-tasks, and use various tools achieve a goal.

It can also use APIs (application programme interfaces), which allow software apps to communicate to one another.

Traditional automation follows a fixed, predefined path. Instead, AI agents are designed to handle unpredictable questions and adapt on the fly.

Types of AI agents

Although the word “agent” can evoke images of futuristic robots, AI agents exist as software systems.

Some agents use predefined workflows, navigating each stage of a specific task in a linear fashion.

Others are more autonomous. They decide dynamically which subtask needs doing, which tool to use, and how to verify results.

Imagine, for instance, an AI agent scanning a set of structural drawings to detect missing sections. It could reference external design codes and then notify you only if it finds an inconsistency.

Automating drudgery: why civil engineers should care

AI agents can help engineers with some of their work- and time-intensive tasks:

1. Complex data management:

Engineering projects generate a wide array of information—from CAD files and geotechnical logs to environmental impact studies.

AI agents excel at parsing, summarising, and correlating these files in seconds.

This changes the game from searching for hours to asking an AI agent a question and getting an immediate, context-rich answer.

2. Automated workflows:

AI agents can call on many “tools”, like design software APIs, web-search or knowledge graphs.

This means agents can handle tasks ranging from real-time sensor analysis to producing quick predictions.

They can even generate proposals for project phasing or design alternatives.

3. Continuous improvement:

AI agents can incorporate user feedback to refine their processes.

They “remember” how prior tasks were solved and learn to avoid past mistakes, which is helpful for iterative design and project reviews.

How AI agents fit into the work of a civil engineer

Below is a conceptual overview showing how AI agents fit into a civil engineering workflow.

AI agents: beyond basic automation. Image credit: Dr Mike Rustell
AI agents: beyond basic automation. Image credit: Dr Mike Rustell

While each organisation might have unique arrangements, the essential parts are the same:

  1. User request

    A query such as “Analyse soil stability at the proposed tunnel site” or “Summarise last year’s maintenance logs” is submitted.

  2. Agents: planning and tools

    An AI agent receives the query. Acting as a hub, it figures out which data or tools it needs.

    Perhaps it calls a geotechnical data API or references an engineering code library.

    It can also delegate work to sub-agents (e.g., one agent checking design codes, another sifting through last year’s site inspections).

  3. Data types

    Agents can pull from diverse data—structured (like a database of soil properties), unstructured (PDF design documents, site pictures), or programmatic (APIs for scheduling, cost estimation, etc.).

    The agent merges relevant information and ensures tasks are completed smoothly.

  4. LLMs and feedback loop

    The LLM (e.g., GPT-4o) underpins the agent’s reasoning.

    It processes the user query, interprets real-world constraints, and proposes actions.

    As it calls tools or sub-agents, it “learns” from results in real time, refining subsequent steps until a final answer is ready.

  5. Response

    The engineer receives a short summary, recommended actions, or visual output from software tools.

  6. Iterate

    Crucially, this loop can continue with follow-up queries. This ensures the agent’s output remains aligned with evolving project requirements.

AI that can handle all types of data

One of the most exciting frontiers is multimodal AI. These systems can handle not only text but also voice commands, images, videos, and sensor data.

Imagine verbally asking your AI agent to “Review last week’s site inspection drone footage and highlight any cracks in the retaining walls.”

The agent:

  1. Processes your voice command to understand the request.
  2. Analyses the video footage using computer vision.
  3. Cross-references the results with previous inspection reports to see if the cracks are expanding.
  4. Produces a short briefing note and flags any need for deeper structural assessment.

This type of workflow unlocks a whole new level of speed and thoroughness. Especially for large-scale projects that generate massive datasets spanning different formats.

Autonomous design and beyond

While much of the hype focuses on how AI handles routine tasks, its potential for autonomous design will revolutionise the industry.

As agents get better at understanding context, constraints, and objectives, and working in large organised groups, they will undertake increasingly complex tasks, such as:

  • Laying out an optimal route for a rail line through complex terrain, taking into account engineering and socio-economic factors.
  • Dealing with non-conformance reports, reviewing the design documents, site photos and preparing the design modifications.
  • Configuring the geometry for a wind farm to maximise energy capture and preparing the planning and permitting documents.

Their output will need human oversight, particularly for safety and regulatory compliance, but agents will evolve to be more like team members than tools.

Looking ahead

AI agents can offer more than just convenience. They have the potential to fundamentally reshape our working methods.

By shifting the burden of routine tasks onto machines, we can focus on the innovative, human-centred aspects of our profession.

As the technology evolves, we’ll look back on this transition as the moment civil engineering stepped into a new era of collaboration.

Not just between humans, but between humans and the autonomous systems that empower them.

  • Dr Mike Rustell, lecturer in structural engineering at Brunel University