Training intelligent agents

When I was in school I worked as a mechanic and the head mechanic I worked with had a very particular way of training.

He would show me once then make me do it myself.  The whole operation was based on efficiency so he knew how to get me up to speed ASAP.  The faster I could come up to speed, the sooner he could get back to his work and the more successful our shop would be.

The routine went like this:

  1. I’d watch him do it
  2. He’d watch me do it and make corrections
  3. Now I can do it on my own

I eventually came to realize this wasn’t just the best way to learn, but it was also the fastest.


Fast forward to today and technology has provided some seriously powerful tools to significantly increase our productivity and efficiency.

When your accounting team is ready, they’ll have the ability to train what’s called intelligent agents to help complete their work.

You can think of an intelligent agent as a bot that has instructions and can perform tasks autonomously based on instructions and input you provide it.


Sounds interesting, but why are we using intelligent agents?

The short answer is speed and accuracy.  Given explicit instructions, the IA (intelligent agent) won’t violate those rules or ‘cheat’ plus they can complete at least 10X more work in the same period of time a human could do it.  How much more productive could you be in this scenario?

Get our awesome product content delivered daily-ish to your inbox


But if the intelligent agent is doing the work, what am I supposed to do?

Your job is now to ‘train’ the agent.  Remember the story I told you earlier, you are now the head mechanic.  It’s your job to tell the IA how to do the work, check their work when they get stuck and help them get better.  Eventually, they will be off and running to get 10X more work done in half the time.  At that point, you’ll move on to figuring out what the next big problem is that you, and perhaps your IA, need to work on solving together.


What does training the intelligent agent look like?

Given your instructions, the IA will bring back results that it may need your help classifying.  For example, if you’re performing an audit and looking for fraud, you can instruct your IA to go find anomalies and bring them back to you.  This means they will occasionally find something for which they need your input to help them learn whether or not it actually is an anomaly which might indicate fraud.  You tell the IA, yes or no, it learns for the next time around and keeps moving.


Intelligent agents aren’t replacing your team members, they’re helping them do 10X more in half the time.


PS. Read more about intelligent agents.

Find this advice helpful? I send out a short message every day to help accounting firms learn how to boost their profit without needing to hire more CPAs.

Join my growing list of subscribers below