.svg){kind=small}
Webinar Recap: Mapping the Agentic AI Era
On December 11th, Pointr hosted a webinar titled “Mapping the Agentic AI Era” featuring Pointr CEO Ege Akpinar, Mapbox Solution Architect Moritz Forster, and Pointr Director of Software Delivery Serhat Önal. The discussion explored how Agentic AI is emerging as the next evolution of artificial intelligence. If you missed the webinar, here’s a recap.
Why Did Chatbots Fail?
Human language is deeply contextual—it depends on conversation history, intent, and ambiguity. Early chatbots didn’t fail because of limited vocabulary, but because they couldn’t understand context or identify what mattered in a conversation. With limited context awareness, they often misunderstood intent and produced shallow, dead-end interactions.
What Changed: Attention and Scale
Traditional language models processed text sequentially, causing important information to fade over time. In 2017, Google’s “Attention Is All You Need” paper introduced “Attention Mechanisms”, enabling Large Language Models (LLMs) to focus on relevant context regardless of position. Combined with scale, this unlocked reliable language understanding.
In simple terms: Context + Attention + Scale = Reliable Understanding
From Prediction to Reasoning
LLMs generate language by predicting the next token. Adjusting the temperature introduces controlled randomness, allowing the model to consider multiple plausible responses instead of always choosing the single most likely one. This makes interactions feel more natural and helpful.
In simple terms: Context + Attention + Scale + Choice = Reasoning
Why Agents?
Agents are not only better chatbots, but they can act. While chatbots have low autonomy and limited actions, agents are goal-driven systems that can plan, decide, and act.
Agentic AI = Context + Attention + Choice + Action
Agents become more capable by accessing external tools (including other agents) via an MCP server. This allows agents to connect to 3rd-party systems, call APIs, and take action within real-world workflows.
Agentic AI in Action
During the webinar, live demos showed how AI agents can connect to Pointr and other 3rd-party systems via Pointr MCP servers to complete real tasks, including:
- Scheduling meetings using Google Calendar and finding the meeting room location
- Providing indoor wayfinding and locating the nearest place on a map
- Recalling past context, such as where a car was parked
These demos showed that agents don’t just respond to questions. They can complete end-to-end workflows.
We also demonstrated a Mapbox Location Agent, which can:
- Find location data
- Plan a bike route
- Add restaurant stops along the way
The demo showed how this can be built using Mapbox’s MCP server, with integration guided by Mapbox’s documentation.
The Road Ahead
The session concluded with how Agentic AI can support a seamless airport journey. A demo video, filmed over 6 years ago, showed our early vision for using AI to help passengers navigate airports.
Today, that vision is becoming reality. Agentic AI is no longer a future concept; it is already here, and its adoption is happening in real time. The shift from chatbots to agents marks a fundamental change in how AI systems understand, decide, and act.
Gina works in the Pointr marketing team assisting with ABM, blog posts, customer slide decks and attending local conferences. She recently graduated from Bucknell University.