Did Visual Basic halt the rise of our robot overlords?

image of type writer with red and blue background | RPA

Key Takeaways

Robotic Process Automation (RPA) has its roots in the 1980s with early tools like Lotus 1-2-3 macros, but the introduction of Microsoft Visual Basic for Applications (VBA) in the 1990s limited its accessibility and slowed the democratization of automation.

Democratized automation presents challenges, as the ability to modify workflows can lead to unintended consequences, emphasizing the importance of change control in business process automation.

The rise of AI, including tools like ChatGPT, raises concerns about the unregulated implementation of automation technologies, highlighting the need for thoughtful governance and intent analysis to avoid mishaps in business operations.

Robotic process automation (RPA) isn’t new.

In the 80s, my dad ran a small engineering company. They bought their first IBM PC because it was the thing to do – and nobody could really fathom it out.

I’d cut my teeth on ZX81s and Spectrums in my bedroom, so aged 14 I was invited in during my school holidays to help “do stuff” with their new PC. This invariably involved Lotus 1-2-3 and Borland Quattro spreadsheets, financial forecasts and production planning schedules.

Explore related questions

I didn’t know it at the time, but I’d accidentally been picked up in a new wave of technology and was learning the depths of one of the world’s core software applications with real-life applications in my teens. And I was getting paid for it; I’d spend my money on ZX Spectrum games and magazines, further reinforcing the cadence of my geek cycle.

Lotus 1-2-3 had a thing called macros. A primitive language that involved mimicking the actions of a user with commands like [RIGHT], [END] and [PASTE] to automate menial tasks.

The clever thing about Macros was that you wrote them in a series of cells vertically, naming a range and then telling Lotus 1-2-3 to execute the macro in the range. It ran the “code” vertically and you could watch the automation unravel before your eyes.

This was pre-undo days – and the risk was that your Macro went wayward and deleted a load of data that you actually wanted (we’ll come back to that later).

I figured that I could make the macro comments equations with conditional results – if this cell is X, then change step 10 from [LEFT] to [RIGHT] for example. Suddenly there was a whole world of essentially dynamic code and logic.

In a half-term holiday, I’d automated the company’s cash flow spreadsheet giving daily reports on cash flow by extracting their Sage ledgers and graphing inbound and outbound transactions.

In other words, 1980s RPA.

But, in 1993 – the year I graduated with my manufacturing degree – Microsoft replaced this with Visual Basic in Excel (more specifically VBA – Visual Basic for Applications) and curbed the RPA adoption curve, making it available only to people who knew how to declare variables or write For-Next Loops.

Microsoft accidentally undemocratized RPA. And, in doing so, probably knocked real-word automation back two decades. 1980s workforces in conglomerates and SMEs were full of bright young things without IT degrees who could parse a comma-separated values file and churn out some automated reports for their board members, who all had their automation wings clipped when VBA arrived on the scene.

AI Intent Prompter – an expert in giving AI leading commands to get the “right answer”.

Had this not happened, like Hush Puppies in Malcolm Gladwell’s Tipping Point, in-cell macros would have become super mainstream like @functions and pivot tables.

Instead, we had a two-decade lag before challenger democratized RPA tools arrived on the market, and application vendors started embedding decent native workflow solutions within their cloud platforms.

 

Should you automate everything?

Democratized automation isn’t without its challenges. Any ability to change the inner workings of a solution – whether code or configuration – needs a modicum of change control. You can’t simply have everyone creating workflow rules to do stuff because it seems like a cool idea.

With my SAP hat on, I’ve always been amazed that some changes to the workings of your business processes require formal change control (aka Transports in SAP) whereas others can be done on a whim – the difference being that SAP simply chose to implement certain functionality in different ways, which resulted in different change governance.

One of Resulting’s partners who sell RPA solutions was recently telling me how their own RPA solutions sometimes wreak havoc in their internal systems, deactivating customers and sending out inappropriate emails because of some rule somebody set up. We got into a whole discussion around doing things “because you can” vs. doing things “because you should”.

Yes, you can automate practically anything. But should you?

Which kind of brings me on to AI and new tech like ChatGPT. As cool as it seems on TikTok, probably warranted more than a modicum of change control in the business world.

As humans, and as animals I guess, we’re think/do programmed. We think about stuff, make decisions and do things. I know that’s a depressing way to think about our existence and millennia of evolution, but it is what we do in abstracted form.

If RPA is aimed at replacing some of the “do” part of our existence, then AI will replace some of the “think” part. I’m not going to get into the morals of AI – my daughter is studying social anthropology at Edinburgh, and she’s much better equipped to address this than me – but I am going to put some generalizations out there to act as thinking prompts.

AI like ChatGPT requires prompts. It requires that we provide intent. I recently even saw a talk on a (near-) future job as an AI Intent Prompter – somebody who is an expert in giving AI leading commands to get “the right answer”. If we acknowledge that intent is an input, we have to accept that intent requires both analysis and hypothesis. We can’t provide orphan intentions like “Give me 10 things I can do today” and expect meaningful output from AI.

Analysis and hypothesis require data. And most business data is average – that’s how averages work, right? On average, things are average.

Weak-minded CIOs will feel compelled to “have a go” at Chat GPT and automation because they can, rather than because they should.

RPA also relies on data. Worryingly, RPA’s actions sometimes change data. Or, worse, unleash it on the outside world in the form of emails, reports and so on.

This AI/RPA cycle is the most worrying thing for me in an undemocratized world. Putting this power in the hands of ‘everyone’ without due consideration and control will end badly for many businesses.

Weak-minded CIOs will also feel compelled by peers, benchmarking or Gartner to “have a go” at Chat GPT and automation because they can, rather than because they should.

They’ll do this without thinking properly about the control required. And, because of Microsoft’s VBA Macro switch out in 1993, the world has had 30 years without hardcore automation. CIOs aren’t ready for it.

If you’re one of these CIOs, or you know somebody who is, seek help before unleashing our future overlord masters on us.