AI Agent-Based Multilingual Content Automation for Creative SMEs
A Design Science Research Project
Project Overview
This research explores an AI-assisted workflow for creating promotional content in Persian, English, and Finnish, combining automation with mandatory human cultural and editorial review.
The Problem in Creative SMEs
Small and medium-sized creative organizations struggle with the substantial time and cognitive investment required to generate culturally authentic, trilingual promotional copy. Standard translation tools produce literal, flat phrasing that fails to connect across disparate cultural contexts.
The Research Prototype Solution
By orchestrating specialized generative AI agents guided by structured event briefs, the system drafts tailored promotional content across three language streams simultaneously—strictly governed by a non-negotiable human cultural and editorial gate.
Design Science Research Foundation: Developed according to the DSR methodology (Peffers et al.), this project creates and iteratively evaluates an artifact addressing a genuine operational challenge faced by multicultural publishing and creative enterprises.
Research Objectives
Multi-Agent Workflow Formulation
Design and implement an agentic pipeline capable of transforming a single structured event brief into comprehensive promotional assets across multiple digital formats.
Cross-Cultural Trilingual Alignment
Address linguistic nuances between high-context (Persian) and low-context (Finnish and English) registers, preserving cultural integrity without reliance on literal translation.
Human Governance Gate Integration
Establish and validate a mandatory human review protocol that enforces cultural sensitivity, factual accuracy, and editorial sovereignty before publication.
Empirical Multi-Stakeholder Evaluation
Assess usability, task efficiency, and perceived quality with 12 expert evaluators across 5 distinct stakeholder cohorts under rigorous DSR evaluation standards.
How the AI Workflow Works
End-to-End Execution Pipeline with Human Governance Gate
Event Brief
Input parameters: Title, theme, target audience, schedule, key messaging.
AI Content Generation
Multi-agent drafting, structural expansion, and contextual adaptation.
Persian / English / Finnish
Trilingual generation respecting separate cultural rhetoric and register.
Poetic cadence, formal honorifics, cultural resonance for Iranian & Middle Eastern diaspora.
International clarity, direct promotional rhythm, universal engagement across global audiences.
Concise, authentic phrasing, culturally tailored tone for the local Nordic community.
Human Cultural & Editorial Review
Mandatory Human Gate: All automated content is halted until certified by human cultural editors.
Approval and Publication
Distribution across Parsuma digital channels, social media, and newsletters.
Prototype Evaluation
Verified Findings from the Master's Thesis Research
Overall Satisfaction
Aggregated evaluation rating across usability, content quality, and workflow clarity.
Evaluation Participants
Expert evaluators participating across 5 specialized stakeholder groups.
Participant Groups
Diverse evaluation cohorts representing SMEs, linguistics, and software engineering.
Automated Execution
Approximately 18 seconds for full trilingual promotional draft generation.
Manual Comparison
Estimated baseline for manual multilingual drafting, proofreading, and cross-team iteration.
Processing Time Reduction
Based on automated execution comparison against estimated traditional drafting.
The approximately 99.98% processing-time reduction is based on this comparison and does not necessarily include human review and approval time. The system is a research prototype. Do not describe it as a fully deployed commercial product.
Evaluation Cohort Composition (5 Groups)
Key Research Insights
Cultural Nuance Requires Dedicated Agent Prompts Over Literal Translation
Generic machine translation fails to capture culturally steeped metaphors, honorific forms in Persian, or idiomatic phrasing in Finnish. Coordinated agent prompts that include cultural framing preserve emotional and rhetorical depth.
Drastic Cognitive Load Reduction Enables Better Editorial Focus
By accelerating initial draft generation from ~24 hours to ~18 seconds, editors are relieved from blank-canvas fatigue. Their effort shifts entirely to nuanced curation, style polishing, and cultural verification.
Human Governance Gates are Essential for Trust and Ethical Responsible AI
All 12 evaluation participants across all 5 groups concurred that automated pipelines without an explicit human review gate pose unacceptable cultural and brand risks. Human agency remains non-negotiable.
Accessibility for Resource-Constrained Creative SMEs
The prototype proves that sophisticated multi-agent pipelines can be designed with lightweight, cost-effective architectures accessible to small cultural enterprises with limited budgets.
Research Media & Demonstration
Supporting Media, Demonstrations & Documentation
Project demonstrations and supporting materials will be added here.
This editable section is configured to host demonstration videos, workflow screenshots, research poster graphics, multilingual content samples, and downloadable academic PDF documentation.
Screen recording showcasing event brief input, agent execution, and human review gate.
[Video Slot Pending]Interface captures of the multi-agent console, trilingual output views, and editorial dashboard.
[Screenshots Slot Pending]High-resolution digital version of the academic conference poster.
[Poster Slot Pending]Side-by-side promotional text generated across Persian, English, and Finnish.
[Examples Slot Pending]Executive summaries, thesis chapters, and evaluation questionnaires available upon thesis publication.
[PDF Files Slot Pending]Thesis Publication Status
Full thesis — Publication pending
The full Master's thesis manuscript has been submitted to South-Eastern Finland University of Applied Sciences (Xamk) and is undergoing institutional archiving. Once formal review is concluded, the permanent open-access publication record will be made accessible.
Khosravi, E. (2026). AI Agent-Based Multilingual Content Automation for Creative SMEs. Master's Thesis, Master of Business Administration (YAMK), Digital Industry and AI, South-Eastern Finland University of Applied Sciences (Xamk). Commissioned by Parsuma Oy.