Hyppää pääsisältöön
Yleisötilaisuus

Stop Prompting. Start Architecting. How a Four-Agent Architecture Turns Human Intent into AI Agent Designs?

Tampereen ammattikorkeakoulu
SijaintiKuntokatu 3, 33520 Tampere
Ajankohta20.11.2026 10.00–11.30
Kielienglanti
PääsymaksuMaksuton tapahtuma
A person on stage presents AI agent design. In the background, a presentation slide displays the title “Stop Prompting. Start Architecting.” alongside a diagram illustrating the four-agent architecture. Logos are displayed at the bottom of the image.
Kuva: Photo generated in Adobe Express with AI-generating.
What if AI agents could be used to create entirely new AI agents? In this webinar, the Four-Agent Architecture will be introduced, presenting a human-centred approach to designing Microsoft 365 Copilot agents. Participants will discover how specialized AI agents can be brought together to understand challenges, develop methodological instructions, design new agents and create their visual identities.

The future of AI isn't just about writing better prompts. It's about understanding problems, designing intelligent processes and creating meaningful solutions.

Imagine an architecture where AI agents collaborate to understand complex challenges, develop methodological knowledge, configure new AI agents and create their visual identities.

Before building an agent, we need to understand:

  • What problem are we actually trying to solve?
  • What processes and human needs should the agent support?
  • What knowledge, methods and instructions does it need?
  • How can we improve reliability and reduce hallucinations?
  • How do we maintain human responsibility and oversight?

This workshop introduces a practical architecture designed to address these questions systematically.

Meet the Four-Agent Architecture

Developed by Apa Ali-Löytty at Tampere University of Applied Sciences (TAMK), the architecture brings together four specialized AI agents.

1. PROCESS AGENT - Understand Before You Automate

Explores the underlying problem, process or opportunity.
It investigates context, human needs, alternative approaches, systemic effects, risks and future possibilities.
Key question: Do we really understand what we are trying to improve?

2. INSTRUCTION AGENT - Build Deep Methodological Intelligence

Transforms process understanding into comprehensive, structured and evidence-aware instructions.
These may include research-informed methods, critical reflection, quality assurance, risk management and human decision points.
The result can be an extensive methodological document supporting the new agent's operation.
Key question: How can we give AI agents the methodological guidance needed for complex tasks?

3. STRUCTURE AGENT - Turn Knowledge into a Working Agent Design

Transforms extensive methodological instructions into a practical Microsoft 365 Copilot agent specification.
It defines the agent's purpose, instructions, knowledge sources, capabilities, suggested prompts, testing procedures and governance principles.
The approach combines:
Compact agent instructions + extensive supporting methodological documentation.
Key question: How can we translate deep methodological knowledge into a practical, testable AI agent?

4. ICON AGENT - Make the Agent's Purpose Visible

Creates a meaningful visual identity that communicates the agent's purpose and functionality.
It considers symbolism, accessibility, user experience and visual consistency.
Key question: How can visual design make AI agents more understandable and approachable?

This is not just a workshop about technology.
It's an invitation to rethink how we approach challenges, develop processes and create new possibilities with AI.

Whether you work in healthcare, education, engineering, business, social services, creative industries, research or administration, the same fundamental questions apply.
 

What Will We Do During the Workshop?

During our 90-minute session, we will:

  • Explore the four-agent architecture and its underlying philosophy.
  • See demonstrations of how specialized agents can work together.
  • Discover how extensive methodological instructions can support AI agent development.
  • Examine the practical structure of Microsoft 365 Copilot agents.
  • Explore how the architecture could support your own professional challenges.

Meet the speaker

Apa "singing careercoach" Ali-Löytty is a multidisciplinary professional at Tampere University of Applied Sciences (TAMK), combining human-centred development, coaching, education, creativity and artificial intelligence. His approach to AI development emphasizes systems thinking, critical reflection, practical experimentation and human responsibility.

The four-agent architecture presented in this workshop reflects a simple guiding idea:
AI should not replace human thinking. It should help us ask better questions, understand complex challenges and create new possibilities.

This webinar is connected to Next-Level Sales, an international project co-funded by the European Union through the Interreg Central Baltic Programme. The project brings together Tampere University of Applied Sciences (TAMK), Estonian Business School (Estonia), and Latvia University of Life Sciences and Technologies (Latvia) to strengthen young people's employability and prepare them for emerging opportunities in customer success management. The project combines practical training, career guidance, international collaboration and employer engagement. A key focus is developing future-oriented professional skills, including artificial intelligence, AI agent development, customer relationship management and digital communication.
Through cross-border learning and practical experimentation, Next-Level Sales aims to help young professionals navigate the changing world of work and discover new career opportunities.
 

Järjestäjä

Next-Level Sales -project & TAMK Career Services

Lisätiedot

asko.ali-loytty@tuni.fi