Don't Argue With The Chatbot
Sorry for the clickbait-y title, but it got you here, didn't it?
A few months ago, I was attending a focus group room with GenAI users in Jakarta.
Getting to learn more about them, and how they interact with GenAI in their day to day.
Now, conventional wisdom talks about finding the true magic and power of AI through conversational, multi-turn dialogue.
You ask something, the model answers, you ask a follow-up, and you banter back and forth like you’re chatting with Tony Stark’s JARVIS.
Except, that conventional wisdom seemed to have completely skipped sending to the folks in the room.
Because, whenever the AI spat out something that didn’t quite meet their requirements, the users in the room didn’t type a follow-up correction.
Neither did they start a fresh conversation.
They simply scrolled up, hit Edit, and refined their original prompt.
And that blew my mind enough to make me write this.
They understood something fundamental.
If you want better output, don’t add more noise to a confusing or flawed process.
Instead, fix the input.
AI evangelists talk about this all the time. “Good prompts = good results.”
Advertising folks talk about this all the time. “Good briefs = good creative.”
And there I was, listening to users in Jakarta, reminding me of something I’d learned but never applied in this context.
The gap between good and great lies in the Input Metrics.
I first read about Input Metrics in Andy Grove’s classic leadership book, High Output Management.
Grove explained operations through the lens of a fictional breakfast factory.
If your goal is to deliver a three-minute soft-boiled egg, buttered toast, and coffee simultaneously, you don’t manage the restaurant by standing at the table and yelling at cold toast.
Toast is the outcome metric.
By the time it’s arrived at the table cold, the failure has already happened.
Success lies in monitoring what goes into preparing a good breakfast - the input metrics.
The toaster’s heating settings.
The quality of the coffee beans.
Dialing in the egg timer just right.
If the inputs are good, the output takes care of itself.
And once you see the world through input metrics, you start seeing them everywhere.
Krishna was probably the original management guru.
In the Bhagavad Gita, he laid down the maxim:
“Karmanye vadhikaraste ma phaleshu kadachana.”
“You are entitled to your actions alone, never to the fruits of your actions.”
Basically: “Do not make the results your motive.”
You see this exact philosophy in elite sport.
Former Indian cricket captain MS Dhoni comes to mind.
Throughout his captaincy, he was almost dogmatic about one phrase: “Focus on the process, not the result.”
You don’t win a cricket match by staring anxiously at the required run rate on the giant scoreboard. You win it by managing the inputs: watching the seam out of the bowler’s hand; committing to your footwork; hitting the top of off-stump while bowling; rotating the strike to keep the scoreboard ticking.
The trophy is simply a byproduct of your focus on the inputs.
And MSD should know. He took India to more metal than any other captain we’ve had.
My trainer’s always told me the same thing.
The scale is only an outcome.
What matters are the inputs.
What I eat; how much I eat.
Working out consistently, with planned progressive overload.
Getting in my steps.
Prioritising sleep over my latest HBO Max obsession.
So obvious, once you understand the concept.
Once I’d understood the principle, I started applying it to my work. In particular, our Social Newsroom.
To my team’s surprise one afternoon in 2023, I asked them to think of us as a factory that manufactures Social Engagements. Shares, Saves, Comments and Likes.
And challenged them to visualise our factory floor and the manufacturing process.
That led us to identifying a whole bunch of input metrics that could get us closer to the goal.
Content Volume: We needed to give the algorithm enough opportunities to discover and serve what breaks through. And thus keep improving our channel’s standing. This was the first input metric we focused on getting right.
Creative Variance: Every batter needs a variety of shots at their disposal, to pierce the field and surprise the opposition. Similarly, we needed to introduce a disciplined approach to creative variations - testing hooks, formats, durations and more.
Signals: We started looking at diagnostic metrics - Saves, Shares and Comments - to help inform every successive batch of content. Focusing on these high-value engagements gave us a greater chance of breaking through and driving more Social Engagements.
Year after year, we’ve smashed our targets, typically blowing past them by Q3 and giving ourselves the whole of Q4 to experiment and set up well for the following year.
We spend most of our lives - personal and professional - arguing with results. Coming up with different excuses, and finding things to blame when something goes wrong.
The simple truth is that we need to stop fighting the outcome.
And simply scroll up to edit the input.
That is the insight.
- Samit