What could AI actually see in years of real, unguarded group chat built on shared understanding, inside jokes and things left unsaid? Can it read group dynamics? Get the group banter? Tell us something new about each other?
The idea came from a video I saw online, but it felt worth testing.
So, I did a little experiment. Here’s what happened.
The set up
The premise was simple: WhatsApp chats are potentially where people show up most authentically, away from mass social media audiences and trends. Less performance, limited filter, mostly chit chat about random topics at 11pm on a Tuesday.
Surely therefore if you feed your chat into an AI that’s been trained to spot patterns (STRAT7 Assist – our internal AI chatbot in this case), it should be able to read the group dynamics and tell us something about how the group interacts – who listens, who leads, what do they actually care about, where the friction exists.
I took my group of four friends and a text history covering the last four years of holidays, job changes, relationship challenges, breakups, inside jokes, and approximately one million photographs from Copenhagen…and stuck it in AI.
What AI found
First, the archetypes. AI wasn’t afraid to call out ‘who’s who’ in our group dynamic:
- Person One (let’s call her M), considered the ‘emotional processor’. She brings the big stuff, thinks out loud and values the groups’ opinions before making major life moves. The group rallies around her wins like when she passed her driving test and keeps her going during the tough times like when she decided to quit her toxic job.
- Person Two (K) – the ‘steady anchor’. Rarely complains, mostly listens and validates so when she does share something vulnerable, the group takes note. AI noticed that K’s busy work schedule can mean she’s less responsive but is still the one who remembers the little details about all of our lives the most.
- Person Three (A) – ‘the amplifier’. High energy, quick to hype but also quick to defend everyone else when needed. Reacts fast and enthusiastically (AI notes short, sharp bursts of multiple texts in one go) and is a consistent contributor to every topic or conversation that’s floated.
- And finally…me (E) – ‘the planner’. Apparently, I take charge of the logistics, get the flights booked, sort the accommodation. AI labels me as practical, witty (ha!) and less inclined to process emotions out loud (didn’t realise psychoanalysis was part of the deal!)
None of this feels outright wrong. Some of it feels obvious. And other bits potentially a bit judgy…but seeing how we each show up named by something external was oddly clarifying even if slightly uncomfortable.
Second, the peaks. AI identified when energy in the group was highest based on content of messages (of course) but also message frequency and volume as well as number of contributors and pace of replying.
Interestingly, we didn’t get most energised about birthdays, dinners or nights out…instead, the energy blew up when we started planning, especially if it involved a flight, the beach and sun (sounds like the right group for me!?).
The moment someone suggested a weekend away, that chat blew up – Assist observed overlapping messages, typos everywhere, emojis galore. Everyone’s talking at once on top of one another (not dissimilar to when we meet up to be fair).
A combination of excitement and relief; not only for the trip itself but simply the fact that we’ve finally found a date all four of us can make.
Finally, the conflict. Or I should say, lack of. Four years of chat. AI found zero actual arguments between us.
Either we’re incredibly aligned (possible), or we’re all conflict-avoidant (also possible). When tension emerged it was always about something external, which we had no issues rallying around – a work situation, an argument with a family member, a friend being unreasonable – that’s when the group got fired up.
So, we’re either a genuinely functional unit, or we’re just very good at avoiding saying hard things directly. The AI couldn’t tell. Neither can I.
What I learnt from this experiment
Stepping outside the stream of AI chat crowding out LinkedIn for a moment, maybe we can all have a bit of fun now we’ve got AI. We’re now able to tap into parts of our worlds that might have otherwise felt too big or too complex to sift through.
I was impressed by its ability to observe who leads and who energises, take content and conversation pace into account, identify what topics spark the most excitement, and spotlight what we value – what gets revisited and celebrated versus what’s ignored.
All that being said, what started as a bit of fun quickly turned into something serious and deep for AI…
AIs takeaway: this is a fundamentally secure, emotionally mature group – they show up for each other, not in the dramatic way, but in a daily, consistent way.
My takeaway: this is either profoundly true, or AI has read way too deeply into four years of pure sarcasm it doesn’t quite ‘get’.
Either way, this has been fun. I wonder what else we could learn about human behaviours, habits, motivations and values…do any commercial applications spring to mind for you? Let me know.
Ellie Wroe Wright, Sept 26