AI Agents as Multipliers: What They Improve and What They Don’t
Alper Nebi Kanlı (Staff Software Engineer, AI Search at Optimizely) will talk about what actually changes when you hand work to AI agents, including:
why agents multiply rather than replace: they scale the strong parts of a system and the weak parts equally
which work gets better under delegation (focused tasks with clear signals, scoped workflows with verification) and which gets worse (vague goals, hidden state, busywork that looks like progress)
why many agent problems are architecture problems before they are model problems
one question to ask before delegating anything: what will this agent actually multiply?
Useful whether you’re building agents or deciding where to use them in your own work.
Bring your questions :)
Speaker bio
Alper Nebi Kanlı is a Staff Software Engineer working on AI search at Optimizely in London. Before that, worked on search ranking and computer vision at Hepsiburada and Turkcell. He writes and time-to-time gives talks about embeddings, retrieval and how language models work.
Allpaca is open to all genders. We don't discriminate based on gender or work status.
These sessions are led by Anny, who worked at an AI native startup and deployed AI features to production annyhe. She also writes about AI at AI Turtle on Substack https://aiturtle.substack.com/
Past topics: https://luma.com/allpaca?period=past
LinkedIn scraper agent
OpenClaw + Hermes for customizing job application
python agent + Claude skills
ollama + vs code setup
python notebooks
The schedule is:
intros
demos: usually 1 or 2
discussions
wrap-up
Learning is more fun with others. Bring your friends!