The buzz around Gen AI has reached a fever pitch. At the recent WEF conference in Davos, it was all anyone wanted to discuss. Sorry ESG. Sorry Crypto. As my daughter would say – “you are so last semester!”
Forecasts predict that workers will lose all sorts of high-paying gigs because AI-driven bots will do what they do, only faster, better, and cheaper.
This reminds me of Robotic Process Automation (RPA) hype when it emerged on the scene 8 years ago. RPA was going to replace many people.
If RPA was so effective at automating away the simple and mundane tasks buried deep in the middle and back offices of banks, who is left to be replaced by Gen AI?
The truth is – RPA did not live up to the hype. Many financial institutions invested heavily in RPA capabilities and built and deployed bots across front, middle, and back-office processes.
But the realized value was nowhere near what was predicted.
Why?
Automating business processes to replace human labor is hard.
Humans are very adaptable.
Robots are not.
RPA was supposed to be easy – it wasn’t.
We thought we would configure and train a bot in a few weeks, put it into production, and then send the people who used to do those jobs packing.
The fact is it takes a lot longer to train and configure a bot that can do something meaningful. And when it’s done, the bot doesn’t replace five people; it replaces 1/3 of one person.
That bot that took longer to build and produced less requires care and feeding to stay productive. If the applications the bot accesses change, the bot breaks and must be repaired.
Also, the list of business processes that RPA could automate was shorter than many predicted because of the underlying complexity that required the exercise of judgment. The “happy path” for many processes only covers a small percentage of the total workload.
Lastly, organizations struggled to transform their delivery and management models to factor in the robotic workforce.
All the above issues apply to Gen AI.
Gen AI can leverage earlier generations of automation technology, like RPA, which makes it more powerful, increasing the aperture of the type and volume of work it can automate. We’ve all seen the potential when interacting with the public LLMs.
The promise is there.
The question is how.
Most organizations that I talk to are in the same place. They have many use case experiments and pilots but no real plan to extract the business value from the technology. Success with Gen AI requires changes to people, process, and technology. In other words – a transformation and successful transformations require discipline and rigor. Two words that are not associated with easy.
