Thomas Aglassinger<p>Yesterday I gave a talk about the lessons learned from prototyping using <a href="https://graz.social/tags/AI" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>AI</span></a> with an <a href="https://graz.social/tags/LLM" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>LLM</span></a> and #<a href="https://graz.social/tags/MCP" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>MCP</span></a> to support <a href="https://graz.social/tags/ProjectManagement" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>ProjectManagement</span></a>.</p><p>The approach taken attempts to address common concerns with AI usage:</p><p>- Lack of access restrictions: Have both permission data and AI data in the same database and join accordingly.<br>- Lack of data sovereignty: Have all software components and data located on-premise.<br>- Confabulations: Make it easy to check suspicious answers by having the AI use the same data as the application and business intelligence tools. It is still up to the user to actually get suspicious, though.</p><p>Technically, this can be achieved with <a href="https://graz.social/tags/ollama" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>ollama</span></a> for embedding and prompt processing, <a href="https://graz.social/tags/pgvector" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>pgvector</span></a> for data storage and retrieval, the <a href="https://graz.social/tags/ollmcp" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>ollmcp</span></a> MCP client, the <a href="https://graz.social/tags/Django" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>Django</span></a> application framework and the <a href="https://graz.social/tags/DjangoMcpServer" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>DjangoMcpServer</span></a>.</p><p>However, further work is needed to improve the robustness of results.</p><p><a href="https://github.com/siisurit/events/tree/main/2025-09-24%20AI%20support%20for%20project%20management" rel="nofollow noopener" translate="no" target="_blank"><span class="invisible">https://</span><span class="ellipsis">github.com/siisurit/events/tre</span><span class="invisible">e/main/2025-09-24%20AI%20support%20for%20project%20management</span></a></p>