kth.sePublications KTH
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
AI och prompt engineering i arbetsflöden: En jämförande studie av strukturerade promptlösningar, stora språkmodeller och etablerade arbetsmetodiker
KTH, School of Engineering Sciences in Chemistry, Biotechnology and Health (CBH), Biomedical Engineering and Health Systems, Health Informatics and Logistics.
KTH, School of Engineering Sciences in Chemistry, Biotechnology and Health (CBH), Biomedical Engineering and Health Systems, Health Informatics and Logistics.
2025 (Swedish)Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE creditsStudent thesisAlternative title
AI and Prompt Engineering in Workflows : A Comparative Study of Structured Prompt Solutions, Large Language Models and Established Work Methods (English)
Abstract [sv]

Utformningen av en tydlig och effektiv prompt som säkerställer korrekt förståelse hos stora språkmodeller kräver både tid och mental ansträngning. Detta beror på att processen innefattar noggrann formulering och flera överväganden. Vid fri textinmatning föreligger dessutom en risk för oavsiktlig delning av konfidentiell information. För att adressera dessa utmaningar utvecklades en applikation för kommunikation med en stor språkmodell som ersatte fri textinmatning med fördefinierade svarsalternativ. Med enbart knapptryckningar kunde användare enkelt navigera mellan frågor, medan optimerade prompter genererades i bakgrunden baserat på deras val och metodiker för prompt engineering.

Studien genomfördes med deltagare från en redovisningsbyrå som testade tre olika arbetsmetoder: den utvecklade applikationen med fördefinierade svarsalternativ, en AI-baserad metod med fri textinmatning till en språkmodell samt en etablerad metod utan AI-stöd. Resultaten visade att den utvecklade applikationen presterade bäst utifrån fem centrala aspekter: tidsåtgång, relevans, användartillit, informationssäkerhet och mental belastning. Studien visar att en strukturerad AI-applikation som är verksamhetsanpassad inte bara effektiviserar arbetsflöden, utan också frigör arbetsresurser och minskar risken för att konfidentiell information läcks. Detta möjliggör för arbetare att istället fokusera på uppgifter där deras expertis kan användas.

Abstract [en]

The design of a clear and effective prompt that ensures accurate understanding by large language models requires both time and mental effort. This is due to the fact that the process involves careful formulation and multiple considerations. In cases of free text input, there is also a risk of unintentional disclosure of confidential information. To address these challenges, an application for communication with a large language model was developed, in which free-text input was replaced by predefined response options. Through the use of button presses alone, users were able to easily navigate between questions, while optimized prompts were generated in the background based on their selections and methods for prompt engineering.

The study was conducted with participants from an accounting firm who tested three different working methods: the developed application with predefined response options, an AI-based method with free text input to a language model, and an established method without AI support. The results showed that the developed application performed best across five central aspects: time expenditure, relevance, user trust, information security, and mental load. The study demonstrates that a structured AI application tailored to specific business operations not only streamlines workflows, but also frees up work resources and reduces the risk of confidential information being leaked. This enables workers to instead focus on tasks where their expertise can be utilized.

Place, publisher, year, edition, pages
2025.
Series
TRITA-CBH-GRU ; 092
Keywords [en]
Large Language Models, Artificial Intelligence, Prompt Engineering, AI-Human-Interaction, Information Security, Business Adaptation.
Keywords [sv]
Stora språkmodeller, artificiell intelligens, prompt engineering, AI-människa-interaktion, informationssäkerhet, verksamhetsanpassning.
National Category
Computer Sciences Human Computer Interaction Artificial Intelligence Software Engineering Information Systems
Identifiers
URN: urn:nbn:se:kth:diva-364203OAI: oai:DiVA.org:kth-364203DiVA, id: diva2:1964826
Subject / course
Computer Technology, Program- and System Development
Educational program
Bachelor of Science in Engineering - Computer Engineering
Supervisors
Examiners
Available from: 2025-06-09 Created: 2025-06-05 Last updated: 2025-06-09Bibliographically approved

Open Access in DiVA

AI och prompt engineering i arbetsflöden(1698 kB)163 downloads
File information
File name FULLTEXT01.pdfFile size 1698 kBChecksum SHA-512
dc5cbeb7da7e2b70e8d83521c680460e650ee3a77758b5fe299b4304461b62128c0a97adbe16846639693f60714f56bf6267e545c5b8997ff3a104bc4b6e7901
Type fulltextMimetype application/pdf

By organisation
Health Informatics and Logistics
Computer SciencesHuman Computer InteractionArtificial IntelligenceSoftware EngineeringInformation Systems

Search outside of DiVA

GoogleGoogle Scholar
Total: 164 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 190 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf