Why is automation today so bad? The future: Teaching Automation.

Ahmed Datoo
Mesmer
Published in
3 min readJul 15, 2020
A pile of books on a table with a red apple on top. Three blocks stacked on top of each other with letters A, B, C

I can safely say that automation today stinks.

It’s broken. And it’s been broken for the last 25 years. Lots of tools out in the market. Some that you pay for and some free. But all equally bad.

Let’s face it. Automation today is slow. It breaks. It’s hard to use.

Why? IMHO we’re stuck with the wrong approach. The right approach will seem so clear to you.

The wrong way: “Telling” Automation

Almost all products today use what I call “telling” automation. You have to tell a bot exactly what it needs to do. You’re basically giving it instructions to follow.

Everything is fine and dandy as long as nothing changes. Changes cause headaches. You’ll find yourself always updating your instructions for changes. Huge pain.

Here’s an example. You’re automating an important customer journey using telling automation. One step in that journey is logging into your app.

Here’s what v1 of that app looks like.

Rectangular button that says sign in.
Rectangular Button that Says Sign In

Based on this you tell the bot to look for a rectangular button that says “sign in.”

Everything works great until v2.

Here’s what it now looks like.

Link that says Log In
Link that Says Log In

The bot follows the instructions you’ve given it. It gets to this step and it’s looking for a rectangular button that says sign in. Only now the page now has a link and it says log in. So the bot can’t find what it’s looking for. And boom your automation breaks.

So off you go to update the instructions. You tell the bot to now look for a button or a link. And text that says sign in or log in. And this works well until v3 when another change is made. And so the cycle of updates continues.

The new way: “Teaching” Automation

There’s a better way to do this. A different way.

What if you showed the bot thousands of examples of login pages. And it learns based on what it sees. It learns the shape of the object (e.g., button, link, icon, etc.). It learns the location of the object, the text used, the code used.

With this knowledge, the bot can find objects without instructions. You change the text from sign in to log in? No problem. Bot sees it. You change the UI from a button to a link? No problem. Bot sees it.

Teaching automation is the holy grail. It provides a way to finally automate your automation. Tesla self driving car? Teaching automation. Google Assistant/ Siri? Teaching automation. Alexa? Teaching automation.

RPA solutions today unfortunately use telling automation. That said, analysts have coined a new term for next gen tools: Intelligent Automation, Intelligent Process Automation, Hyper Automation. All of these tools incorporate teaching automation. It’s the future!

You will see the market segment into companies solving broad use cases across every department within an enterprise. And new companies focused on specific use cases.

Here’s the market as we see it.

At Mesmer we’re focused on solving a specific use case: customer experience problems. Across development, staging, and production. Imagine tracking your most critical customer journeys throughout all stages of your CI/CD pipeline. And using intelligent automation to know of problems before customers call in with complaints. And using this same automation to fix problems before anyone notices.

You’ll finally see speed. You’ll see resiliency.

Welcome to the future of automation!

--

--