If AI were a real employee

Oct 26, 2023

See the sensationalist headlines: 

“AI will take over many human jobs and make thousands redundant!” 

“AI is as creative as humans and much faster and more accurate” 

Anyone who has dipped their toes into the world of AI from a lay-person’s perspective, can’t help but be in awe of the power and response time of bringing together information in a lifelike way. For some this is liberating to enable them to churn out reports, document and social media in a fraction of the time it used to take. 

But is it really taking over someone’s job? 

If it did, it should be subject to the same levels of scrutiny that any other employee would be…

Having that “difficult” conversation

Line manager: “Look, Michael-AI, I asked you to write me a summary of our company, based on what could be found on our website, and I gave you our website address. You came back with a description of a clothing and fashion retailer with multiple stores and a reputation for being low-cost. We are an engineering company who work with small manufacturers. How could you get that so wrong?” 

Michael-AI:” I apologise, I looked at the wrong website. Here is the correct one about a company that specialises in nanotechnology and microelectronics and based in Worcester.” 

Line manager: “That’s slightly better, as you have the location reasonably close, but we’re still nothing to do with nanotechnology and microelectronics. Look at the web address again.” 

Michael-AI: “My apologies, I was looking at some old information which may no longer be relevant. Here is the corrected information. You are a training company offering a range of online and in-person training courses with online booking and options of a way to pay…” 

Line manager: “Stop! Training is a small part of it, and we certainly don’t offer standardised packages. Where are you getting this information from, because when we updated our website this year, we made sure to emphasise the Leader Liberation Plan and the manufacturing support and process improvement, and you’ve discovered nothing of this from your own search. Where do you get your information from?” 

Michael-AI: “I search my databases but if you have failed to update those databases with your new information then it may not be available to me. Would you like me to search again?” 

Line manager: “No, don’t bother. It’s quicker for me to do it myself.” 

Michael-AI: “My pleasure to help you. Is there anything else you would like me to do?” 

Line manager: “Yes, could you please provide me with our company disciplinary procedure and write a note in your own HR file with a verbal warning for having failed for the 15th time this month to check the work you produce, use the information that is available to everyone else and not even attempt to improve your performance following our previous 14 little “chats”.” 

Michael-AI: “Of course.”

Is this for real?

Is this an illustration of how flawed AI is, or of how our expectations of what is possible from using it exceeds the practical? 

When used with consideration and precision, AI can be a powerful tool to help productivity, but at the same time with human inaccuracy and an assumption that the tool is more intuitive than it is, AI can be a big time-burner, and a source of frustration and mistrust. 

For anyone hoping to use AI to replace interactive humans, then remember: 

  • The only new information is what you add to its instructions
  • It has to be given the context
  • It’s output (almost always) should be checked
  • Fortunately, it can’t read facial expressions
  • It has a long memory and tells its friends everything

How do we get the best out of this?

As with any emerging technology, an understanding of what it can do and what potential exists, is rapidly evolving. There’s no disputing that AI has become significantly more accessible, but as we do with any industrial process that we look to improve for businesses, we have to consider where and how it might give us some advantage. 

Process definition, or process mapping, or any other description, helps to break down individual activities and will show which tasks can be packaged to AI, along with specific direction about the sources of information, and which need to be done by humans. 

We do this for many of our clients – to a level that many of them don’t expect – and it does mean that where a company really wants to add AI into the mix, we can help them find out precisely how and when it’s appropriate, and the scenario illustrated above, wouldn’t need to happen.

 

Yet!

Oh, and in case you were interested, MATL specialises in manufacturing process improvement through engineering and our Leader Liberation Plan to release business leaders from the day to day grind of operational matters. We don’t sell fashion clothing or specialise in nanotechnology – honest! 

Our website at https://matl.co.uk tells all. Give us a shout!

Images courtesy Andrea Piacquadio