Omid Hosseinian Engineering Group

The Right Path for Introducing Artificial Intelligence into Engineering Offices

The expansion of artificial intelligence-based systems has gradually transformed the workflows of architecture offices, engineering firms, and organizations operating in the construction industry. However, the starting point of a successful AI project is not necessarily the selection or purchase of a new tool; rather, it begins with the precise identification of an operational, repetitive, and measurable problem. The first episode of Radio Mohandes examines why defining the problem, establishing the objective, and understanding organizational constraints must take place before selecting any technology.

In many engineering companies, the main challenge is not the lack of intelligent tools; rather, it is the weakness in information management and document control. Scattered files, uncertainty regarding the latest version of documents, the absence of an integrated structure for classifying correspondence, and low data quality or limited traceability can significantly affect the reliability of artificial intelligence outputs.

A system may provide a coherent and seemingly credible response; however, if its answer is based on an outdated, incomplete, or irrelevant document, the likelihood of making an incorrect decision increases. Therefore, the accuracy, freshness, consistency, and traceability of data are more important than the sheer volume of information.

In this episode, three common use cases in engineering environments are compared: searching and retrieving information from reports, meeting minutes, and correspondence; preparing management summaries from daily reports; and classifying incoming correspondence and suggesting appropriate routing paths.

This comparison demonstrates that the first AI pilot project should have a clearly defined scope, sufficient repeatability, limited risk, measurable outputs, and the ability to return to the previous workflow if necessary.

Within this framework, artificial intelligence acts as a decision-support system rather than the ultimate decision-making authority. The system’s outputs must be reviewed and, when necessary, corrected by the technical office manager, document control team, project manager, or a qualified engineering specialist.

This approach, known as Human-in-the-Loop, enables organizations to benefit from the processing speed of artificial intelligence without transferring professional responsibility, technical control, and expert judgment to the system.

The practical outcome of this episode is the presentation of a four-week framework for implementing a limited AI pilot project. This framework includes selecting a specific process, defining the baseline condition, preparing a controlled dataset, conducting a pilot test with a limited number of users, and comparing the results with the existing method.

Indicators such as reduction in task completion time, output accuracy, human correction rate, number of critical errors, traceability of responses, and user satisfaction will serve as the basis for decisions regarding project continuation, modification, or termination.

One of the most common mistakes in digital transformation projects is adopting a tool-oriented approach instead of a problem-oriented approach. In this approach, organizations move directly toward purchasing or deploying an AI tool before properly defining the problem, process, data requirements, and decision-making responsibilities.

The consequences of such a decision may include unnecessary cost increases, the use of unreliable data, dependence on untraceable outputs, excessive user reliance on the system, and ambiguity regarding final decision responsibility.

This episode demonstrates that before initiating any artificial intelligence project, organizations must clearly define:

• Business objectives

• Data quality and ownership

• Risk levels

• The role of human reviewers

• Acceptance criteria

• Project termination conditions

Applications of This Episode

• Evaluating the organizational and data readiness of an engineering office for adopting artificial intelligence

• Identifying and prioritizing appropriate use cases based on value, risk, and feasibility

• Assessing the quality, accuracy, freshness, and traceability of data and documents

• Designing decision workflows with effective human involvement and defining final responsibilities

• Establishing key performance indicators, acceptance criteria, and project continuation or termination conditions

• Using artificial intelligence to identify formatting inconsistencies, coordination issues between departments, discrepancies between tables and text, and missing references in technical reports

• Maintaining professional responsibility and engineering expertise in technical and structural decisions

Five Key Questions Before Starting

1. What specific problem, decision, or process needs improvement, and how is its current performance measured?

2. Do the required data meet the necessary conditions regarding quality, freshness, ownership, accessibility, and usage permissions?

3. What exactly is the role of artificial intelligence within the decision-making chain, and who will be responsible for final review and approval?

4. What is the most critical potential error of the system, what consequences could it create, and what mechanisms are planned for control, interruption, and returning to the previous method?

5. Based on which indicators will project success, failure, acceptance criteria, and termination points be evaluated?

Sources Used

1. HBR Guide to AI Basics for Managers 2023

Chapter 4: AI Doesn’t Have to Be Too Complicated or Expensive for Your Business

Focuses on data quality and consistency rather than relying solely on the volume of information.

Chapter 8: How to Spot a Machine Learning Opportunity

Focuses on identifying suitable problems and processes for applying machine learning.

Chapter 9: A Simple Tool to Start Making Decisions with the Help of AI

Introduces the use of an AI decision canvas or AI Canvas for defining decision components.

Chapter 10: How to Pick the Right Automation Project

Focuses on evaluating and selecting appropriate automation projects and developing organizational capabilities.

Chapter 14: Managing AI Decision-Making Tools

Addresses determining the level of human intervention and designing oversight mechanisms for AI-based decisions.

Chapter 15: Your Company’s Algorithms Will Go Wrong. Have a Plan in Place

Focuses on anticipating algorithmic failures and designing response, control, and shutdown strategies.

2. CIOB Artificial Intelligence Playbook

Chapter 1: Practical Implementation of AI in Construction

Covers practical principles for implementing artificial intelligence in the construction industry and starting with familiar, limited, and controllable processes.

Chapter 3: Machine Learning in Construction: Opportunities and Uses

Discusses machine learning opportunities in construction and the twelve-step strategy for implementing AI within organizations.

Chapter 5: The Foundations of Quality Data

Explains the role of reliable, structured, manageable, and traceable data in the performance of artificial intelligence systems.

Section: Resources — AI Use Case Template

Provides a framework for defining business objectives, user needs, data resources, proposed solutions, expected benefits, risks, and evaluation criteria.