Lately, I’ve been connecting some dots from things I’ve watched or heard in my formative years with what’s going on today. Maybe it’s my way of making things easier for my brain to process, or maybe I’m doing the opposite and over thinking things. Regardless, I’ve been talking about this clip from Friends with coworkers, fellow CX professionals, and basically anyone who let me practice telling this story. This is the story of how Rachel Green made a traditional English trifle and how it is a stark warning to the blind adoption of AI.

For those unaware of the series and characters…

• Rachel, while determined and confident in her ability to deliver a delicious desert at arguably the most celebrated meal of the year (American Thanksgiving), doesn’t normally cook.
• Monica (a professional chef), Phoebe (a vegetarian), Joey (a perpetually hungry actor), Ross (a paleontologist with his own history), Chandler (no one can say what Chandler does), and Jack & Judy Geller (Ross & Monica’s parents) are our guests for this year’s gathering.

Let’s break down the clip

• Rachel made a traditional English trifle using a cookbook
  • Rachel follows the cookbook to the letter. She constructs the trifle meticulously, starting with ladyfingers, jam, custard (which she made from scratch!), raspberries, more lady fingers, beef sauteed with peas and onions, more custard, bananas, and topped with whipped cream.

Wait, did you catch that? How many desserts do you know that have beef, peas, and onions in them? Here’s what Rachel had to say about the layer of beef…

“The beef was weird to me too, mince meat pie is an English dessert, these people just put very strange things in their food.“

Rachel believed that since the English have a mince meat pie that “beef custard” was a common ingredient in an English Trifle. For the unaware, mince in English cooking is s mixture of dried fruits, spices, and butter, often with spirits like brandy or rum.

While Rachel goes to get a bottle of rum from Joey & Chandler’s apartment next door, Ross discovers the pages of the cookbook were stuck together. Rachel unknowingly made half a traditional English trifle, and half a Shepard’s Pie. Joey’s retort sets the stage, “only one layer of jam?!“

The time has come, and everyone is seated at the dinner table with a plate of Rachel’s trifle in front of them.

• Phoebe asked if it was a traditional English trifle, and if it has beef or eggplant. Since she doesn’t eat beef, Phoebe gets a pass from having to consume the trifle.
• Monica eats it and feigns delight and even says, “it’s soo good that I feel selfish for being the only one eating it”.
• Chandler is also trying his best not to express his disgust as he begrudingly eats the trifle and then says he wants to enjoy it on the balcony.
• Jack and Judy want to leave the table so they can call their friend from Monica’s bedroom to tell them how good the trifle is.
• Monica wants to go to the bathroom so she can look at herself while eating the trifle.

At this point Rachel is getting the feeling that something is amiss and reaches for Ross’ plate. Instead of telling Rachel the truth he quickly consumes the dish so there’s nothing left, only to tell Joey after Rachel leaves…“it tastes like feet”.

Joey can’t get enough and gobbles it all up. And this is what he says…

“I like it, what’s not to like? Custard, good. Jam, good. Meat, GOOOOOD”

Let’s consider some of the character motivations…

• Rachel is motivated to prove that she can make a great desert that everyone enjoys.
• Joey and Ross have a double date and want to get out of dinner as soon as possible. They can’t wait for another dessert to be made and deal with just having Rachel’s.
• Phoebe Monica, Chandler, Jack, and Judy are motivated to not hurt Rachel’s feelings, in addition to not eating this awful trifle.

Now let’s get to why I think this is a lesson on AI adoption.

Rachel represents the creators that rely too much on AI and not enough on consumer needs.

Her determination and pride interfered with delivering a remarkable & delicious Thanksgiving dessert. She used the cookbook with stuck together pages and incorrect assumptions about English cooking while constructing the trifle.

Ross, Monica, Phoebe, Jack, and Judy all represent different consumers of AI products. Each have their own motivations, but underpinning each is avoiding hurting Rachel by telling her that the dessert is terrible.

I left Joey out of that last list on purpose because his behavior is the one I’m most concerned about.

Joey may represent the majority of future consumers of AI generated output.

In addition to being time bound and hungry, he genuinely enjoys the less than appetizing desert. “What’s not to like?” he says, while everyone else wants to put this trifle behind them as soon as possible.

So, where do we go from here?

As creators, I believe in operational transparency and allowing consumers the chance to be part of the product development lifecycle. Imagine if Ross told Rachel that the pages were stuck together sooner?

These aren’t just focus groups and interviews (which can be biased based on financial compensation or represent more Joey’s than the rest of the table), but also incorporating listening posts at the right time and place, in addition to community/beta users to scale insights and get ahead of issues that would be catastrophic at GA.

With all that rich data, creators must develop a system to understand what is being shared at scale. They also shouldn’t expect to get the same kind of insight from their sources too. For example, you shouldn’t expect to get the majority of your feature requests from support interactions.

Lastly, closing the loop with customers that do share feedback, whether it be through an in-product notification or auto-reply email, with confirmation of receipt, how to get help in the future, and what to expect (e.g., we’ll reach out if we need additional clarification), increases the odds that customers continue to use those valuable channels.

But James, wouldn’t that slow down development? Not if you bring the right partners in first.

Mapping out known points of friction, where customers feel the least confident, and also getting ahead of issues that developers won’t see (could be pride, or are too close to the problem, or don’t understand customer needs as well as support reps do, etc.) are all ways that you can increase the slope of an adoption curve as opposed to trying to claw back customers who have already moved on.

Lastly, having the right measurement in place to determine which users are saying what and why, to not just understand current customers, but target for the future ones, is necessary to ensure ongoing growth and focusing on the things (e.g., marketing dollars to target the right customer, revenue generating cohorts, multi-product adopters, etc.)

As consumers, we must use those opportunities to provide honest and useful feedback. Personally, I’ve spent the better part of 20 years presenting data and receiving feedback about how well my message is being received (similar to when I’ve tested getting my point across using Rachel’s trifle).

While I love hearing praise (who doesn’t?), it’s often most helpful to receive direction on what I can work on to be even more effective & efficient. Over the last 7 years my profession has allowed me to read & synthesize solicited and unsolicited feedback from customers, in addition to millions of AI and human assisted support interactions. We expect to receive a normal distribution of “you suck” and “you’re great” at the bottom and top deviations, and while these can be useful to measure the ratio of praise to criticism, the most actionable feedback highlights what isn’t working. And this is where the alchemy of using feedback and turning it into trust happens.

Understanding at scale what consumers feel and expect at onboarding, first time use, when things go sideways, and deciding whether to grow or go is the first step. Using that information to make decisions on what to build is next. Finally closing that loop with customers to let them know that they talked and you listened, can result in a loyal customer base that act as your greatest lever for future growth.

Building trust with customers is not a one time event. Creating a flywheel of actionable feedback is necessary to satisfy current and future expectations.