The more banking moves into AI assistants, the fewer places are left to engage people. Progress bars, streaks and status tiers live on screens; a chat window answers the question and ends the exchange.
What is left to work with is what the assistant says back.
The problem is that those assistants are built around a single response style. The person on the other side – Achiever, Explorer, Socializer or Killer – needs a different answer to the same question to remain engaged.
To adapt the response you need to know which archetype you are dealing with, and almost nobody has defined which behaviours give that away.
This kit is that definition. It is a 28-page working document, written to be used by product, design and data teams at the same time, so they can build engaging, gamified communication with users inside an AI chat.
Inside you’ll find:
- 10 behavioural signals for detection: unprompted goal checks, screens opened but never used, fronting a group bill, etc.
- The events you need, named with fields and flags + the derivation logic for each signal.
- How to combine signals into a profile without averaging all back into 1 neutral voice.
- The Player Archetype × AI Companion Behaviour Matrix: tone, initiative, reward framing and failure mode for each of the 4 types, on 1 reference page.
- Privacy, compliance and the fairness line for behavioural adaptation.
- An event checklist to take into your next planning session.
Who it’s for
Built to be read by 3 people at the same time:
- Product Manager who owns which archetypes are worth detecting at all
- Designer or Conversation Designer who owns what the assistant then says
- Data Engineer who owns whether these events exist yet
Our Kit is structured so that all 3 can read the same page and take different actions from it.
Where it comes from
The signal interpretations are ours. They come from our hands-on UX breakdown of Revolut AIR across 100+ scenarios, and from patterns we see repeatedly across usability testing — more than 2,700 sessions in 20+ countries.
What the kit deliberately doesn’t give you is thresholds. Those depend on your product’s baselines, and a number copied out of someone else’s document will send you in the wrong direction. The signal interpretations are ours; the thresholds have to be yours.
The thinking behind the kit can be found in Player Archetypes in the Age of AI: Why 1 AI Companion Doesn’t Fit All Players, and the archetype model itself comes from our Gamification Guide for Financial Products.
