Marketing systems Recorded on a real call

Why does a system only get good once people actually use it?

From a client call with an insurance broker.

By Manan Mehta · 16 November 2026 · 5 min read

Unscripted. Clipped from a real call. Only my side is shown.40 seconds
The short answer

A system is only successful when it is used, because only in its usage do we understand its flaws. I can only build so much. I know the mistakes in my own work, but I don't know how mistakes happen in your business. You tell me what broke, I patch it, and the system gets better.

A system only gets good once people use it, because usage is the only place its flaws show up. I can build a lot before launch, and I know the mistakes in my own work. But I don't know how mistakes happen in your business. You tell me what broke, I patch it, and the system fits your work a little better each time.

Why can't a system be built right the first time?

On a call with an insurance broker, I was walking them through a system my team is building for them. It covers the quote form, the follow-up emails and a tracker. Questions kept coming up about how it would handle this case or that case. My answer was simple. A lot of this is still being built, and there are going to be problems that come in the middle.

That is not a warning. It is how systems work.

The key to a system being successful is it being used. Only in its usage will we understand its flaws.

I can only build so much before that point. Here is how I explained it. If you tell me to build a system for making automatic marketing proposals from a questionnaire, I can build it to perfection. I have been doing that for the past seven years, so I know the mistakes that can happen.

But I don't know how the mistakes happen in your business. That knowledge does not sit with me. It sits with the people doing the work, and it only comes out when they use the system.

What happens when someone finds a flaw?

We find the flaws when you start telling me. Something like: this client submitted the document and attached it, but I didn't get a notification.

Then I go into the system and tell it: okay, you made this mistake. It creates a patch, it deploys it, and that mistake won't happen again.

So the loop is short. You use it, you hit a flaw, you tell me, it gets fixed. Each round, the system fits your business better than it did before.

How should you report a problem?

This part matters more than people think. The best way is to email me with a few screenshots. Keep it plain: I was on this page. This is the error that happened. This is the screenshot of it. Please can you check this?

That's it. I can then give it to Claude and process it. Even five requests a day is fine.

What I need is for each one to be catalogued properly. If I misread the problem and change the wrong thing, it just creates a loop of new problems. So write it as if you were explaining it to a child. This is the mistake, here is the screenshot, this is what I came across, please check. If you do that, I will understand, and the machine will understand.

If your team wants to collect notes together first and send them in one go, that works too. But if combining everything is hard, just email me directly.

Why test it before it goes live?

I spent a lot of time building this system. Most likely it is working. But I still need to wiggle through it myself and make sure.

I asked the people on the client's side to do the same. Take your time with it and see wherever you spot loopholes. It's early, so even if something breaks, we can solve it in real time.

I also suggested a live walk-through call. We go through the system together, practise with it, and find what's wrong so we can work on it. That is usage too, just with all of us watching at once.

When it does go live, I'd rather start clean. Pick a date, and from then on every new lead goes onto the new system. The leads you already have, you keep handling the old way until that circle is over. That way the system is used properly from day one, rather than being squeezed into a pipeline that is already moving.

Why not perfect every design choice first?

This is also why I don't spend long on small design choices before launch. On the same call, the question was whether clients should attach their documents to an email or upload them on a platform.

I was honest about the limits. In the system as built, AI can't go into Gmail, download an attachment and process it. If the documents come in by email, checking them has to happen manually. But the system can already tell you that something has arrived.

And if people are not going to click a link and upload, there is no point building that place. Then we get confused: somebody emailed it, somebody uploaded it. Honestly, I don't feel it's a big difference between attaching a file and uploading it. Start with what works now, and we can build the rest in the future.

What I ask of anyone getting a system built is this: start using it, and tell me what you see, clearly. The flaws you find are what make it good.

“The key to a system being successful is it being used, because only in its usage will we understand its flaws.”

Said on the call, 27 August 2026
Take this into your AI

Open this article in your AI assistant and ask how it applies to your business.

Read more of this on Google

Add mananmehta.com as a preferred source and Google will show my posts higher in Top Stories and AI Mode for you.

Add as a preferred source on Google
Manan Mehta
Written by Manan Mehta

I run MarketinCrew, a marketing agency of around sixty people across Mumbai, Lyon, Paris and Toronto, and AI Crew, where we set up AI and marketing systems for businesses. I work with founders as a Virtual CMO, teach Search Engine Marketing and Social Ads at YNOV Campus in Lyon, and host The Marketin Podcast. Before all of that I co-founded American Joint, a restaurant in Mumbai that grew to three outlets.

If you want a system that gets better the more your team uses it, start with an AI audit from my team at AI Crew. We look at how your business actually runs, then build and fix from there.

Get your audit withaicrew.com