Most “AI agents” today are passive.
On building autonomous agents (as a product, not coding agents): I’m partnering with Bia Affonso and Neu21 on Euda — a human-centred AI-powered wellbeing intelligence platform that helps people thrive and transforms workplace culture through continuous conversation while paying attention to wellbeing drivers and psychosocial hazards. A lot of “agents” today are passive — they only engage when the user engages, and they just respond to what has been prompted.
2 min read
Originally posted on LinkedIn · 13 reactions · 11 comments · View original →
Most “AI agents” today are passive. We’re building one that isn’t. 🤖
On building autonomous agents (as a product, not coding agents):
I’m partnering with Bia Affonso and Neu21 on Euda — a human-centred AI-powered wellbeing intelligence platform that helps people thrive and transforms workplace culture through continuous conversation while paying attention to wellbeing drivers and psychosocial hazards.
A lot of “agents” today are passive — they only engage when the user engages, and they just respond to what has been prompted.
Euda is different: she’s autonomous and has agency. Every half hour, she “wakes up” and "reflects" (for each employee), whether to proactively start a conversation and what the next best interaction should be: a follow‑up, a gentle nudge, sharing useful information, offering feedback, or prompting a reflection. Often, the right call is to do nothing.
That sounds simple; it isn’t. You’re trying to mimic a human‑like interaction that is contextual, timely, and relevant; that respects preferences and style; that can read nuance in feeling and intent; and that has situational awareness of what might be happening for that person.
The sheer volume of variables and nuance for each individual makes it impossible to codify precise behaviour. So you have to provide all the relevant context and trust the agent’s reasoning to make the right decision in the moment.
There are plenty of hard problems here — memory, context engineering, timing, evaluation, safety and more.
I haven’t yet seen other attempts at autonomous agents with genuine situational awareness and agency — not even from the frontier AI labs or humanoid robots. If you’re building in this space — or have great references — I’d love to learn from you.
The analogy that comes to mind when thinking about codified behaviours vs first principles guidance for an agent’s reasoning is the difference between Waymo and Tesla: the former layers lidar/camera/radar with codified rules; the latter relies on vision and neural nets to make decisions. What we’re attempting feels closer to the latter.
Would appreciate any thoughts, experiences, or good references to explore.
#Agents #AI #Wellbeing