Generative AI, namely ChatGPT, are on the front of everyone’s mind.
Thought leaders in the automation space are eagerly trying to apply the technology to use cases and companies like PWC and Deloitte have established a business line and invested over $1 Billion to start. Like others, I think it’s too early to predict the impact it will ultimately have on automation solutions and society, in general. Just like other technologies before it, factors such as security, viable and scalable use cases and the breadth of data that can be fed to the model will impact its success and adoption. That being said, I do think there are some pretty clear areas where Generative AI will gain traction and adoption.
Low Code Application Development
This will be a game changer for many of the Low Code vendors out there. As they have built “activity based” code generation that makes it easier to assemble a workflow. However, it has an inherent ceiling by which users are beholden to the constraints of the activity that was compiled. While there is the ability to provide unique parameters, that is the extent of how much one can customize or control the outcome. Many Low Code apps are built to be “easy to use” but intrinsically don’t have the same level of functionality as a fully-fledged development environment. Generative AI takes care of this issue immediately. The use of Python and other open-source coding languages can make coding or getting the code needed to create a robust workflow an easier task. The feedback model of ChatGPT will only improve the code as issues are identified. I anticipate, over time, that there will be a new breed of Low Code Applications built on Python or other languages in which code will be created by instruction voice or text and the feedback loop provided by ChatGPT will improve the code with minimal developer intervention. The current heavy UI Activity type application will have a hard time building this type of capability into their current architecture in a timely fashion. Another factor to keep in mind is that much of the technical talent entering the workforce has some background in Python, so this will be a natural fit with the use of Generative AI.
Instant Feedback for Real-Time Decisioning
I have long held the belief that the power of AI is to reduce human decisioning to 5%. Meaning AI, in whatever form, can provide a correct, non-emotional answer 95% of the time and require a human to make the more nuanced decision-making for the remaining portion of the decision. With generative AI this possibility seems to be getting closer. The challenge is with data and most notably security. How do you govern what is asked and how much information is fed to generative AI?
This becomes an issue with PHI, PII and any other personally identifiable information. In healthcare, a generative AI agent could provide a clinical diagnosis, but how accurate can it be without a vast amount of historical data? Those problems will be solved over time and businesses will become more sophisticated in their utilization of and in their strategy-building for the use Generative AI. There is still quite a bit of “plumbing” that needs to go into this. For example, a proper feedback model needs to be installed before an operational model for businesses can use this technology in a scalable and reliable way.
RPA will not go away, it will flourish.
I always laugh when I hear other people who claim to be experts indicating that RPA is dead or not a technology that can stand on its own. I think part of that comes from not understanding what RPA really is. UiPath is an automation platform that uses RPA, as do other vendors in this space. With the platform you can literally do anything (document understanding, utilization of stand-alone API’s, use of sophisticated AI models to help with decision-making, and even the utilization of .net custom code). To me, it only makes sense that the orchestration model that UiPath and others provide will only increase the use of these AI platforms. Generative AI is part of a solution but isn’t the whole solution. It needs to be orchestrated with other complimentary software platforms to do its job. RPA interacts with applications from the desktop. Generative AI will increase the speed of the time to market for typical RPA development and provide a feedback loop for Intelligent Document Processing, as an example. We are already seeing vendors implement this in their current tech stack.
These are a few of the areas “I think” tech can and will evolve but as it becomes more robust and widely used it will become clearer which use cases are more effective and useful.
Written by: Peter Camp. CTO & Founder
