How to design/re-engineer and/or automate Processes & Tasks for a GenAI enabled ecosystem?

I have been thinking about this question for quite a while now and after coming across a recent HBR article on a somewhat similar topic, I decided to pen down some of my thoughts and see what you all think about it as well.

As process design, re-engineering & process automation professionals, we tend to usually design/re-engineer and/or automate processes based on whether a particular “task” within the process is/will be performed either by a human or by a machine (either entirely by an ‘unattended bot’ or with some human assistance/intervention by an ‘attended bot’). This works quite well when it comes to designing/re-engineering processes leveraging the current Business Process Management (BPM) and Robotic Process Automation (RPA) solutions/systems. However, when it comes to GenAI based systems/solutions (powered by LLMs), we need to consider something quite fundamentally different when designing, re-engineering and/or automating processes. This is because GenAI based systems/applications work by letting their users (humans) “talk” to them - just like they would to/with another human colleague. Moreover, it works with its user in an iterative fashion – i.e., it continuously improves as it “learns” more about the user’s specific needs and behaviours (and vice versa). Hence, the concept that a particular task (within a process) will be performed either by a human OR by a machine does not align well with the fundamental concept/aspect of GenAI.

So, to effectively integrate and leverage GenAI as part of process design, re-engineering, or automation, we need to design for “Human-AI Dialogue” – wherein machine/AI and humans can share responsibilities dynamically as part of a task within a process. Each can take the lead at different points throughout the task - based on the context of the task and/or their level of competence/expertise. The idea is to treat GenAI more like a coworker and switch tasks between humans and the machine/AI - based on who is better at doing which task (instead of rigidly pre-assigning tasks to or for either?humans?or?machines/bots). It goes without saying that we should allow the more repetitive and pattern-based tasks as well as tasks that involve a high degree of work with Words, Images, Numbers, and Sounds to be performed by the machines (or GenAI in this case) while tasks that require more EQ, very complex decision-making as well as nuanced judgements should continue to be performed by humans. ?This will allow AI and humans to perform a task jointly - learn from each other, and continuously optimize/improve the process in near real-time.

Hence when designing/re-engineering processes for GenAI powered/enabled ecosystems or environments, we should first define how GenAI should engage with human operators and vice versa, “learn” from each other as well as establish clear guidelines regarding how & when AI should seek human input and vice versa.

This is certainly a very exciting and new paradigm (for me at least) and I would love to hear your thoughts regarding this.