June 28, 2023

Video may have killed the Radio Star, but AI won’t kill (good) analysts

By Martin Squires

Building on our month of thinking about Thinking Skills, how do analysts add value? How does the rubber hit the road in terms of applying their thinking skills to add more value than AI alone? Like all roles and sectors, that is surely the challenge we face. How can we add more value than using ChatGPT et al instead?

To help us consider that sobering challenge, I am delighted to welcome back guest blogger, Martin Squires. Previously a Director of Analytics for organisations including Homeserve & Pets at Home, Martin has now gone over to the supplier side & helps clients as a senior consultant at Merkle.

Regular readers will recall that Martin has shared with us before on topics including data storytelling, web analytics and the toolkit needed by analysts. In this post, Martin turns his attention to the threat to analysts from Generative AI. Should analysts be worried and if not why not?

Is AI coming for your job as an Analyst?

AI seems to be everywhere at the moment. Not just in trade blogs, social media and data science journals but it’s making the headlines on primetime news. They are all talking about not just the end of data science and analytical roles but of swathes of jobs across a massive range of industries. There are even some dire warnings about the end of humanity itself. Is Skynet real?

Well maybe not, but for once it doesn’t seem to be just IT sales hype. Even just a cursory play with tools such as ChatGPT and it’s clear we are on the verge of a significant step change in capabilities. We are seeing the emergence of some exciting new tools.

Why will (good) analysts not go the way of the Dodo? 

The key thing for me is that the world still needs analytical thinking to drive technology. A faster car without a great driver at best doesn’t go all that fast and at worst is a pile-up waiting to happen. Why do I think that? Well, a combination of some hard-earned experience plus a great book I read last year and would highly recommend.

What does my experience tell me?

I’ve been around the block a few times. I go back to when data scientists were called data miners and even further into the mists of time when we were simply statisticians. For 25-30 years I’ve had IT salespeople telling me I could deploy their new software and replace a bunch of expensive analysts. Promising to give business users direct access to the data with their tec. The days when neural networks first made their way into data mining tools in the late nineties, is just one example. But moving from the Buggles to Elton John, “I’m still standing” and good analysts are still in high demand. Why is that?

I think it’s down to something I first learned many years ago at M&S Bank. I was implementing a new self-serve reporting tool. All went well as we trained users, with people highly engaged and excited. At least one person asked me over a coffee break what I and my team were going to do now everyone could do this analysis stuff themselves. 

It turns out that the same colleague gave me my light bulb moment a week later. Coming to my desk she announced that she’d figured out what my team actually did. I asked her to enlighten me. Turns out she thought it was mostly just pressing buttons on an analysis tool or scribbling code. During the training she said the tool was great, just click here to drill down and click here to get an answer. But when she sat down with a real business problem what she couldn’t do was figure out where to start. Her lightbulb moment was seeing the need for people with analytical thinking skills.

You need people who bring analytical thinking skills. 

One of the things I learnt by reading the book that impressed me this year was discovering the work of Conrad Wolfram. In his “The Math(s) Fix” (2020), Wolfram developed a computer-based maths process comprising four steps:

  1. Define the question
  2. Turn it into a form that can be computed 
  3. Compute answers 
  4. Interpret results. 

The tool could do stage 3 but not the other stages. Good analysts deliver on all 4 stages, not just stage 3. In fact, stage 2 could explain why so many videos are being created already on “how to write a good ChatGPT prompt”. Step 2 is hard unless you are experienced with this sort of thinking.

What About the book I mentioned originally?

That book is “Mathematical Intelligence: What We Have That Machines Don’t” by Junaid Mubeen. It is an excellent look at why machines are wonderful tools, but why without a good analyst to partner with them they are still simply tools. 

Mubeen lays out seven elements of mathematical thinking as part of his case:

  • Estimation
  • Representation
  • Reasoning
  • Imagination
  • Questioning
  • Temperament
  • Collaboration

All have their merits and are worth a read but a few really hit home and made me think about the 5 Cs I’ve always looked for in a great analyst: curiosity, creativity, communication, collaboration and common sense. Mubeen makes some great points related to these requirements:

  • Machines that are premised purely on patterns may have predictive value, but they lack common sense and reasoning skills to explain their choices. They may say, with some degree of reliability, what will happen in the future – but not why?
  • Any algorithm that relies purely on patterns in the data, void of context, will never be capable of explaining it’s choices.
  • AI’s strength is to sharpen our creativity, running calculations en masse, offering thousands of simulated examples from which an analyst can explore and learn.
  • AI tools can only answer a question, they can’t do what great analysts can and make the connections between two or more projects and combine findings for greater insights. Nor fuel more questions. A great analyst doesn’t just deliver hard truths they also delivers wisdom and insights. Those are traits that are hard to encode.
  • AI will struggle to work out if it’s answer makes sense. Just because a pattern is there is it meaningful or can it drive action?

So why won’t AI kill off analyst roles?

There’s far more in the book and I’ve barely scratched the surface but for any analysts out there wanting to defend the value they add and why they can’t be replaced by a machine, this is a manifesto for the value of an analytical or mathematical mind and the value you bring to the table. 

So, back to the beginning and why AI won’t kill analytical roles. Based on all the above I claim that: Computers won’t replace analysts, because the two working together are superior to either on their own. 

As Junaid Mubeen said far better than me…

Computers are not a substituting force replacing analysts, rather they are a complementary force, replicating mundane and lengthy tasks previously carried out by analysts to get to the ‘what has happened?’ and subsequently leaving analysts freed up to focus their minds on the ‘why?‘”

Junaid Mubeen “Mathematical Intelligence: What We Have That Machines Don’t” (2022)

Are you reassured, analysts?

Thanks to Martin for sharing both those books and his reflections on the value which (good) analysts provide over and above AI. I hope that has encouraged you and perhaps highlighted where to focus on your development. As Tony Boobier has shared previously in his books on preparing for AI in the job market, now is the time to develop those skills which will differentiate humans.

Have a good week all and I look forward to the conversation that will ensue from sharing these thoughts.