Artificial Intelligence (AI) is a combination of technologies that allow smart machines to extend human capabilities. It uses massive processing power, intelligent algorithms, and huge datasets to mimic human cognitive abilities such as sensing the immediate environment, comprehending what’s happening, acting on this information and learning through experience. Autonomous vehicles are the most well-known example of AI using these human-like abilities to accomplish a complicated task that requires constant assessment and recalibration in real time.
Core functional capabilities
From an agency or operational perspective, the core functional capabilities enabled by AI technologies include:
- Enhanced Interaction — Enabling more natural and seamless collaboration between individuals and systems. Examples include chatbots, virtual assistants and virtual advisors.
- Intelligent Automation — Delegating common steps and decisions for system execution. This would include solutions like robotic process automation (RPA) that automates routine processes requiring limited human judgment.
- Enhanced Judgment — Making more contextual determinations that go beyond static business rules. Next Best Action guides caseworkers through complex decision-making processes, and machine learning employs self-learning to extract insight from large datasets.
Applying automation and augmentation
One way to think about the range of available AI solutions is to evaluate them in terms of both automating and augmenting. Automating involves taking over underlying tasks routinely done by humans, such as simple data entry, to allow them to focus on higher value work; augmenting bolsters workers skills, knowledge, and experience, helping them become smarter and more productive or effective.
Mapping tasks to AI solutions
AI solutions can be considered based on two criteria: 1. the predictability and repeatability of the task and, 2. the complexity of the data that is either structured or unstructured. In the matrix below, on one end are routine, predictable and rule-based processes —typing information from a paper form to the computer, or checking that dates use the correct format and are not missing digits – that allow little deviation. At the other end, processes may be more ad hoc and unpredictable and require the expert judgment of an experienced benefits case manager or intelligence analyst. Likewise, the information driving the process can vary dramatically, ranging from fairly simple, highly structured data to very dynamic, unstructured text or imagery.
Activities that are closer to the bottom left are more likely to involve automation, while those closer to the upper-right part of the matrix are more likely to use AI to augment human capabilities. It is also important to note that many jobs will have aspects or characteristics that span the different models. For example, some elements of the work of a lawyer or doctor might fall under Efficiency and Effectiveness, while other aspects might be more in line with the Expert and Innovation models. For more detail on the models, see Turning Artificial Intelligence into Business Value Today.
Agencies can use the work complexity/data complexity framework to map tasks to four primary types of activity models: efficiency, expert, effectiveness and innovation. By using the matrix, agencies can see which tasks cluster around either automation or augmentation and can begin to map the appropriate types of AI solutions for exploration and testing.
Related: Read more about Process Reimagined