Imagine you’re working with an AI tool, hoping it will be a real partner, not just a fancy calculator. That’s what the Human-AI Handshake Framework set out to test. Researchers at Chulalongkorn University looked at popular tools like GitHub Copilot, ChatGPT, and Adobe AI to see if they actually help humans and AI work together as a team. They wanted to know: where do these tools fall short when it comes to true collaboration?
Here’s what they found: every major AI tool has the same blind spot. They can adjust to you while you’re using them, but as soon as you start a new session, it’s like meeting a stranger all over again. They never remember what you taught them last time.
Take GitHub Copilot. It can suggest code and let you say yes or no, but it never remembers your choices for next time. ChatGPT can keep up with you in a single chat, but once you close the window, it forgets everything you taught it. Adobe’s AI tools let you tweak things as you go, but the next time you start a project, it’s back to square one. None of them actually learn from your feedback over time.
So every time you use these tools, you’re forced to start from scratch. Why? Because most AI is built on fixed models that can’t grow with you. The researchers say that for real teamwork, you need five things: sharing information, learning from each other, checking each other’s work, giving feedback, and helping each other get better. Right now, AI tools only do some of this, and only for a little while.
Why does this matter? If you’re running a business, you need to know if your AI tools are just giving you a quick boost, or if they’re actually learning and growing with your team. The research shows that most tools are stuck in the present—they don’t pick up on your company’s habits or needs over time. So next time you’re choosing an AI tool, ask yourself: will this thing remember us tomorrow, or is it just a one-off helper?
If you’re building AI products, here’s the catch: the real problem is those fixed training models. The researchers suggest new ways forward, like using smarter learning methods that let AI remember and adapt across sessions. And if you’re a researcher, this framework gives you a way to test why real-world human-AI teamwork still hits a wall, even if things look good in the short term.
