Initiated the Situational AI inquiry after two decades building software and, most recently, AI coworkers. Responsible for the hypothesis and the research program.
About
Prasad Pillai initiated Situational AI and is responsible for its hypothesis and research program. He is the founder and CEO of Mindspace AI, the organizational and engineering home of the work, and the founder of BotsWork, where situation-centric ideas are being tested in working AI coworker systems.
Why I Started This Inquiry
The inquiry began with a practical frustration. Modern AI had become remarkably good at reasoning once it was told what problem to solve. Given the right prompt and the right context, it could write, plan, analyze and explain. And yet the systems we were building still failed in ordinary situations, not because they reasoned badly but because nobody had told them what was going on.
Real work does not arrive as prompts. A customer goes quiet. A payment fails for a long-standing account. A project drifts. Nobody asks anything; something changes. The harder problem was recognizing what problem exists, for whom it matters, and what the moment calls for. That question, once noticed, was hard to put down.
From Building AI Coworkers to Situation Recognition
Building AI coworkers at BotsWork made the gap concrete. Giving an agent more context helped, up to a point. What it could not do was tell an ordinary card decline from a valuable customer being blocked by a temporary outage, or a cooling prospect from a resolved support ticket, when the underlying data looked the same. Smart was not the same as competent.
The pattern kept repeating: the missing capability was not more knowledge or better reasoning but recognition, an explicit account of the situation the system was in, from whose perspective, and what that implied. That led from context engineering to situation recognition, and from there to the question of whether situations should be first-class objects in the architecture rather than something inferred and forgotten inside each prompt.
The Situational AI Hypothesis
Situation deserves to become a first-class computational primitive of intelligent systems: something a system detects, represents, compares, tracks, remembers and revises as evidence changes. Situation is to intelligent behavior what the token is to language modeling, an analogy about computational role, not a claim that the work is done. The manifesto sets the hypothesis out in full; the research program is the attempt to test it.
Current Research Questions
- Can a system recognize which situation is emerging from changing evidence, and notice when its interpretation is wrong?
- Can experience be remembered as situations, responses and outcomes rather than as text?
- Can structurally analogous situations be found across domains that share no vocabulary?
- Does an explicit situation layer actually improve timing, decision quality, coordination and explanation?
These are the open questions of the program, not settled results.
Professional Background
Over two decades of building technology to solve real-world problems, including roles at Microsoft and co-founding Elsys Intelligent Devices. Founder and CEO of Mindspace AI, which conducts applied AI research, proofs of concept and engineering around emerging AI architectures; founder of BotsWork. Speaker, including a TEDx talk, with coverage in The Better India, Manorama News, Crazy Engineers, Mathrubhumi News and Raksha Net.
Events
- Situational AI Open Research & Technology Forum
- Introduction to Situational AI for the JSOM MBA Class
- JURS Students Meetup and Recognition
- JURS Student Presentations on Situational AI — Day 2
- JURS Student Presentations on Situational AI — Day 1
- Introductory Address to the JURS Students
Essays
- Situation Is Not Context
- Why Situational AI Isn’t Another Agent Framework
- Introducing Situational AI
- All essays →