Harnessing Artificial Intelligence for Project Management Efficiency

Abstract

This thesis investigates and analyzes how to implement Artificial Intelligence (AI) into project management, addressing the process through a two-step approach. Initially, a framework is developed to identify and prioritize project management areas where AI can enhance operations. This framework was constructed based on a comprehensive literature review, adapting existing frameworks to the specific requirements of project management. It assesses tasks currently performed by project managers and ranks these tasks according to their potential for AI implementation. Subsequently, the thesis investigates the practical implementation of AI within these identified high-potential areas. Two AI solutions were developed as demonstrations; the first, a risk register utilizing a retrieval augmented generation (RAG) architecture, was evaluated to offer limited value. In contrast, the second demonstration, a budget tool designed to automate information extraction from PDF files, demonstrated significant potential. This tool successfully automated the extraction of information from various PDF structures provided by different suppliers, achieving a 98% accuracy rate for readable PDFs through techniques such as prompt engineering and fine-tuning. Furthermore, a business case analysis for the budget tool suggested a potential payback period of 0.36 years for deploying a fully functional application. The findings suggest that effective AI implementation in project management should begin with identification of tasks suitable for AI implementation. These tasks should then be prioritized based on financial, time, and risk implications, alongside the effort required for AI integration. The implementation process should foster collaboration between technical experts and domain specialists, embrace rapid iterative feedback, and initiate pilot demos for stakeholder evaluation prior to full-scale production. The thesis also concludes that successful AI deployment in organizations demands robust data management, data protection measures, comprehensive AI education for the workforce, and a culture that trusts but also critically evaluates AI solutions.

Abstract (Swedish)

Förevarande uppsats undersöker och analyserar hur artificiell intelligens (AI) kan implementeras i projektledning genom en process i två steg. Inledningsvis utvecklas ett ramverk för att identifiera och prioritera områden inom projektledning där AI kan förbättra det dagliga arbetet. Detta ramverk skapades baserat på en omfattande litteraturstudie, där befintliga ramverk anpassades till de specifika kraven inom projektledning. Ramverket bedömer de uppgifter som för närvarande utförs av projektledare, och rankar dem efter deras potential för att kunna förbättras genom AI. Vidare analyseras den praktiska implementeringen av AI inom dessa identifierade områden med hög potential. Två AI-lösningar utvecklades som demonstrationer; den första, ett riskverktyg som använder en arkitektur för retrieval augmented generation (RAG). Denna bedömdes erbjuda begränsat värde. I motsats visade den andra demonstrationen, ett budgetverktyg utformat för att automatisera informationsutvinning från olika PDF-format, betydande potential. Detta verktyg automatiserade extraktionen av information från olika PDF-strukturer som tillhandahålls av olika leverantörer. Den uppnådde en noggrannhet på 98% för läsbara PDF:er genom tekniker som prompt engineering och fine tuning. Vidare gjordes en investeringsbedömning för budgetverktyget som visade en potentiell återbetalningstid på 0,36 år för att bygga en fullt fungerande applikation. Resultaten tyder på att effektiv AI-implementering i projektledning bör börja med identifieringen av uppgifter med hög potential för AI. Dessa uppgifter bör sedan prioriteras baserat på ekonomiska, tidsmässiga och riskrelaterade implikationer, tillsammans med den tid och kompetens som krävs för att skapa AI-lösningen. Implementeringsprocessen kräver samarbete mellan tekniska experter och domänexperter, användning av snabb iterativ feedback och pilot-demonstrationer för utvärdering av lösningens potential. Utredningen resulterade i slutsatsen att framgångsrik AI-implementering i organisationer kräver robusta processer för datahantering, åtgärder för att skydda data, omfattande AI-utbildning för personalen och en kultur som litar på men också kritiskt utvärderar AI-lösningar.

Popular Abstract

Will project managers be replaced by Artificial Intelligence in the future? This is a question asked within many professions today. While the answer for project managers is most likely no, there are a lot of things a project manager can do to be more effective in day-to-day work. Building on this, project-oriented organizations have the potential to improve their processes with the help of Artificial Intelligence. We propose a framework that can be applied to identify and prioritize what parts of project managers' daily work can be replaced or improved with the help of Artificial Intelligence. We then apply this framework and develop a demo of a solution using the latest technology within Artificial Intelligence that has the potential to save hundreds of hours of work for project managers.

Artificial Intelligence (AI) is probably the most discussed topic in the business world today, largely due to the introduction of large language model applications such as Chat-GPT. Many industries already use AI or machine learning (which is a form of AI where computers learn by analyzing large amounts of data). However, project management has traditionally been a bad fit for AI due to its constant variation in project scopes. Since AI is built on top of historical data, it is hard for an AI model to find patterns among historical data if the data is constantly changing along with projects. There has also historically been a lack of knowledge about the benefits and drawbacks of AI among project managers due to the focus on human interaction and leadership in the work of a project manager. While project management will likely remain human-centered, our hypothesis was that there is a large potential to automate parts of their processes, thus freeing up more time off the project manager to be spent on leadership and development.

To decide where AI is best suited to be implemented, we developed a framework based on previous research that helps to identify and prioritize where it is suitable to apply AI in the role of a project manager. The framework has three steps. First, the job of a project manager is split up into tasks, ideally covering all aspects of the job. Second, the tasks that (for now) completely lack potential for AI (examples of this could be negotiating prices in an in-person meeting) are removed from further analysis. Finally, the remaining tasks are weighted and ranked based on their impact on the end result of the project and how long they would take to implement. The result is a prioritized list of opportunities to implement AI.

To test our framework, we applied it to the processes of a project management division of Tetra Pak, a global leader in packaging and processing. Based on our framework, the highest priority task to implement was a solution that helps project managers input supplier quotation information into the budget of each project. Currently, this is done manually for the hundreds of quotations that belong to each project, resulting in project managers spending hundreds of hours on this task in a year. The demo of a solution we built is an AI model based on GPT-4 (the underlying model powering Chat-GPT), that automatically extracts information from the different quotations in a project and outputs that into an Excel file containing the budget of the project. The demo was improved over several iterations, starting out with only 50% accuracy but ending up with a whopping 98% accuracy for a full project!

Although the model shows great potential, there are several steps left to put the model into full production mode. First, it needs to be able to handle all different types of quotations, even those that were not included in the training material used for the demo. Secondly, the solution needs to be integrated into the organization to be utilized fully, something that is a common challenge when implementing AI in companies. This thesis shows that there is huge potential for AI to impact project management but that there are also many challenges that can stop AI from driving real value in the organization. With the help of the approach in this thesis, project-driven organizations are hopefully better equipped to meet the AI revolution.

This popular scientific article is derived from the master´s thesis “Harnessing Artificial Intelligence for Project Management Efficiency” authored by Oscar Tyrberg and Tim Adenmark.

• author: Tyrberg, Oscar and Adenmark, Tim
• supervisor: Lars Bengtsson
• organization: Innovation Engineering
• alternative title: Utnyttjande av Artificiell Intelligens för Ökad Effektivitet i Projektledning
• course: INTM01 20241
• year: 2024
• type: H2 - Master's Degree (Two Years)
• subject: Technology and Engineering
• keywords: Artificial intelligence, AI in project management, Generative AI, Prompt Engineering, Retrieval Augmented Generation, Fine Tuning.
• language: English
• id: 9156223
• date added to LUP: 2024-06-17 11:03:39
• date last changed: 2024-06-17 11:03:39