Many years ago, I watched an elderly couple trying to get on the metro escalator at Connaught Place in Delhi. Their son was with them, showing them how easy it was. Step on. Hold the rail. That’s it.
They were still very hesitant. They almost took the stairs.
Escalators were still fairly new in India then. For many people, the metro was their first ride on the magical moving stairs. And it was interesting to watch first-timers. They found it hard to get on, and were usually just as awkward getting off.
It wasn’t so much fear of the unknown. It was fear of the unknown in all its engineering glory, right in front of their eyes. Or should I say, right under their feet. You can see the metal steps folding into each other at the ends. You can see the machine working. One had to match the pace of each step emerging and keep the balance as you boarded it.
And having someone experienced beside them didn’t reduce that anxiety much.
Now compare that with an elevator
First-time elevator users usually have a pretty neat experience. With someone experienced beside them, it’s a non-event. Even alone, they figure out the buttons and get moving. Maybe they hold their breath when the car moves a little too fast. But they’re never as anxious as their cousin on the escalator.
Why?
Because in an elevator, the magic is hidden. You don’t see the cables, pulleys, counterweights and brakes. For all you know, you’re busy looking at the floor number on the display.
Back in 2014, this observation gave me two questions for product managers :
- Can you design the product so that its complexity is hidden from the user?
- Can a new user’s onboarding be helped by an experienced one?
I still think those are very relevant questions.
But what does the answer look like – seems to be an evolving concept for me.
In 2018, I had argued that all friction is not bad – some users need a few extra steps to feel safe.
Indian payments became an elevator
Think about what happens when you pay with UPI today. Scan. Enter amount. Enter PIN. Done.
Behind that simple action of tapping, there are two banks, a payment app, a payment network, multiple switches at banks, fraud checks and a settlement cycle. All working at massive scale, without minimum lag – behind the scenes. Nobody needs to know what a VPA resolves to, or which bank is the acquirer. The complexity is entirely hidden.
And the second question also saw an interesting answer in the realm of UPI adoption. How many people made their first UPI payment because a shopkeeper, a nephew or a colleague showed them how? The experienced user did the onboarding. No product tour needed in most instances.
Closer home, my dad finally started using UPI after seeing me pay with my smart-phone across big and small merchants, witnessing how I could fetch and pay electricity bills so easily while there was an ad-break on TV. He finally learnt it all – P2P, P2M (Scan to Pay), bill payments – and along the way chose the same TPAP (UPI app) I did – GPay. Never tried PhonePe or PayTM or his own bank app (!!).
AI in finance is an escalator right now
Now look at where AI in financial services is today.
Chatbots that tell you they’re AI and might be wrong. Disclaimers pop-up before every answer. “This is not financial advice.” Consent screens. Explanations of why a decision was made. A button to talk to a human.
Feels like we are showing the customer the moving steps. In all its complicated glory. And, just like the couple at Connaught Place, many of them hesitate.
And (most) will continue to hesitate for banking, payments etc. Unless things change drastically!
Some of this is unavoidable. Financial services is regulated for good reason. A customer has a right to know when a decision about their money was made by a machine, and to have someone accountable when it goes wrong. You can’t just hide everything and call it good design.
So is the answer to keep building escalators?
What elevators actually show you
Here’s what I missed in 2014.
Elevators don’t hide everything. They hide most of the complexity and moving parts. But they show you the controls and even the safeguards. There’s an alarm button. An emergency phone. Often an inspection certificate, with a date on it.
It doesn’t tell us how the elevator works. They are there to help us feel comfortable, that it will work. And if it doesnt, we are still in control.
I think that’s the design principle AI in finance needs. Hide the mechanism. Show the controls and safeguards.
- The customer doesn’t need to see the model, the prompt or the confidence score.
- They do need to see who’s accountable, how to reach a human, what happens if the answer is wrong, and what the system will not do.
It also connects with something I wrote about recently in The Human-in-the-Loop Paradox. In that post I argue that If humans (by very nature) make for poor continuous reviewers of reliable automation, then the human’s real job isn’t to watch every step. It’s to be the alarm button: the one who steps in on exceptions.
A human-on-exception is to AI what the emergency phone is to an elevator.
And the experienced hand still matters
The second question around onboarding matters even more for the AI era. Because for every magical video on what AI can do, there is a corresponding scary one, telling how it can unleash havoc in our lives.
Perhaps the first person to use an AI feature shouldn’t be the customer. It could be the relationship manager, the branch officer, or the call-centre agent, using it alongside the customer. They become the son at the escalator: someone the customer already trusts, showing them it’s safe. Once enough people have ridden it with someone beside them, more of them will ride it alone.
This is also something, we have seen before – self-checkin kiosks at airports (assisted model), ATMs inside branches to help walk-in branch customers see how to use their debit cards for cash withdrawals.
So, what are you building?
If you’re building anything with AI in financial services, here are four questions to ask:
- What machinery are you showing that the customer doesn’t need to see?
- What safeguards are you hiding that they do need to see?
- Who is the experienced hand in your onboarding, and are they using the product first?
- Is your product an escalator or an elevator today? And is that by choice, or by accident?
I don’t think there’s one right answer for every product. Some journeys should stay escalators for a while, where seeing the machine is part of building trust.
But I do think the best AI products in finance will feel like elevators: you press a button and get where you wanted to go, with a very clear alarm button if anything feels off.
What do you think? Are we building escalators or elevators? If you see it differently, I’d love to hear from you.
The original Linkedin Post is here .