Déjà Vu: Is Applied AI 2025, RPA 2015?

A new MIT study just dropped news that shouldn't surprise anyone in the automation trenches: 95% of generative AI implementations have zero measurable impact on company P&L. (Anyone remember RPA overpromising?)

Interestingly enough, what caught our eye in the report was that the 5% of cases categorized as successful were, drum roll... Intelligent Automation use cases, focused on back-office, tedious, mundane tasks, generally outsourced and offshored. Anyone has seen this pitch before?

It begs the question: if AI impacts our old world, what is to become of our new world?

We're learning how to navigate this shift with you, and our current thesis has two parts:

• the principles of Intelligent Automation remain unfazed by AI;

• productivity and scope increases will dramatically change because of AI.
  

Let's break down the thesis.

The Principles Remain Unfazed

While everyone else is scrambling to figure out what AI "just killed", we're here with a contrarian take: the fundamental principles of what makes enterprise automation programs successful haven't changed at all.

Treat It as a Technology: If one resists the temptation to anthropomorphize and concludes that AI isn't magic, what we get is a technology with specific capabilities, limitations, and use cases. Just like you wouldn't deploy RPA everywhere, you shouldn't deploy AI everywhere.

Make It Human +: Another important point made in the MIT piece is that the vast majority of investment (failed investment, we should say) is directed at replacing human judgment in functions like customer service and marketing, rather than augmenting judgment. This is a classic mistake that automation practitioners also made, and our take is that bringing humans along the journey remains a principle that, when broken, ensures technological improvement fails to stick around (or at the very least makes it way more costly to implement).

And lastly,

Process is (Still) King: A costly mistake we've made in the past in Intelligent Automation was overstressing technology importance in automation, often neglecting the importance of improving processes. We think process improvement stays the core pillar of successful automation. Shitty processes will not only continue to be hard to implement, but they also have potential risks when AI is applied on top of them.

So if the principles remain unchanged, what does AI actually change for Automation practitioners? Here's our take: 3 Possible Manifestations of AI in your Automation Programme: From our vantage point, we see three key angles where AI will impact Automation strategies.

AI as Another Tool: AI becomes part of your existing intelligent automation stack alongside the workhorses of Automation: RPA, low-code platforms, conversational AI, and IDP. It doesn't replace these technologies; it complements them. This impacts the way you do discovery of use cases, segmentation per technology, and means that your design teams and tech leads need to up their game to avoid siloed decision making.

AI as Capability Augmentation: Interestingly enough, all your favorite tools are embedding AI capabilities. So you'll be put in the interesting situation of having to choose between a pure play AI implementation (e.g. with OpenAI) or leveraging an embedded capability in one of your favorite tools. So you better have a point of view on what you can do with each tool.

Another interesting observation is that while pure play AI projects might require a largely new skill set (e.g. high level programming languages) that Intelligent Automation professionals don't have, AI capability augmentations within your preferred vendors often come at a much easier cost to implement. And our empirical evidence on the ground is that this is where most organizations are starting. What everyone is calling "agents" is the act of infusing intelligence in a process step that was otherwise scripted (e.g. reading unstructured data). So we have a long way to go still.

AI as an Entirely Different Delivery Framework: Lastly, and where we have spent the most time at kyma. We think AI allows for a change in how delivery is done. Instead of cutting corners on process definitions and neglecting user testing due to build consuming all the project length, we think build will be compressed with AI and Intelligent Automation teams will have further incentives to go deeper on processes and user interactions. This is the world we are creating at kyma.

Over the next weeks, we'll try to go deeper into each of these AI manifestations, so stick around for more. If you want to discuss your automation challenges, contact us here.