Keeping someone “in the loop” means keeping them informed or involved. The term human-in-the-loop (HITL) means involving humans in processes that leverage innovations like AI, robotics and machine-learning algorithms. HITL ensures that a process isn’t run automatically without human supervision. On the surface, this might seem like distrust of automation and AI, but it’s a layer of precaution used to avoid mistakes or misses that AI and robots can easily make. More importantly, HITL will enhance these processes.


Almost every industry has seen a rise in the use of automation and AI. It’s not difficult to figure out why given the business impact shown by these technologies. AI and automation have proved that they can do a lot of work that humans do, especially large amounts of rule-based, repetitive work. What could take humans hours to perform, these technologies do in minutes, for a fraction of the cost. Does this mean the future of humans as analytics and technology professionals is bleak?


We believe that the foreseeable future is far from bleak and that AI, robotics and machine learning will benefit both humans and businesses. That benefit could be perpetual if we can manage to re-skill the workforce at the pace of automation’s growth. The skills and capabilities of humans and automation technology, contrary to popular belief, are not at odds. Research has found that intelligent automation is likely to complement the skills of humans, who are more suited to perform tasks that involve their creativity, curiosity, imagination, ability to socialize and communicate effectively. Robots and AI are suited to perform more labor-intensive work such as sifting through large data sets. Ceding rote tasks to automation will free humans up to solve problems that require more creativity. We’ll see the evolution of newer roles such as human exception handler, bot trainer, bot supervisor (responsible for the smooth functioning of automated machines in a factory) or bot auditor (responsible for filtering out biases in AI by providing unbiased data for machines to learn from.) As long as humans learn to adapt to an ever-changing workforce, we’ll continue to reap the benefits.

Such cleanly split responsibilities might lead one to ask why HITL is necessary, but this division of labor is an over-simplification of the human-machine relationship.  Humans and machines have long had and will continue to have an interdependent working relationship. Consider the case of intake processing in the pharmacovigilance function of a pharmaceutical company. Today, it’s mostly managed by a large team of people who take direct calls, read faxes from organizations like CIOMS and skim through copious literature to identify adverse events. With intelligent process automation (IPA), some companies are already re-imagining operations with an HITL approach. These companies use technologies such as natural language processing to transcribe phone call audio and identify key entities such as people, places, brands or events. They use computer vision to scan through the documents and extract key data fields and RPA to orchestrate the process end to end and upload into safety case management tools like Oracle Argus. Humans oversee the whole process, manage all exceptions and quality check anything that goes into regulatory systems.


As mentioned earlier, HITL is concerned with fusing the most relevant attributes of both humans and machines to drive success for businesses. It’s primarily focused on four activities: exception handling, compliance management, supervision and quality checks. HITL ensures that machines constantly learn under the supervision of humans. For instance, a chatbot is effectively programmed to answer specific straightforward questions. But when it encounters a question (an exception) it cannot recognize or analyze, it transfers that exception to its human supervisors. This would be an example of exception handling. This layer of supervision ensures that the issue doesn’t go unaddressed. Data from the interaction can then be fed into the machine-learning algorithm to enable the chatbot to learn and respond better to similar situations in the future.


HITL can help us achieve the best of both worlds in human-machine interaction. On the one hand, keeping humans present to validate and train machines can help allay fears of an AI takeover that renders human skills and capabilities irrelevant. On the other hand, freeing humans up to focus on the most interesting challenges and creative solutions offers new and exciting possibilities for the future of human employment.