By Justin Silverman
Before we try to explain how automation and artificial intelligence can integrate productively, it might help to define their differences. Many people no doubt confuse the two, and that isn’t helped by the way the media often conflates the two.
First off, “automation” involves the application of technologies for carrying out processes with minimal human intervention. Robotics and software are forms of automation, but they don’t necessarily include AI.
Artificial intelligence is the simulation of human intelligence by machines. Some see “artificial intelligence” as a monolith, but it’s really a catch-all term for several different capabilities.
Artificial Narrow Intelligence (ANI), for instance, is highly specialized, like a chess program that can beat a human being but will never be able to operate a light switch. There are good examples of ANI now available for commercial usage in natural language processing and machine learning. Artificial General Intelligence (AGI) is “strong” AI, like IBM’s Blue Brain project that simulated—but still in a limited way—human problem-solving and learning processes. Artificial Superintelligence (ASI) is the Ultron or HAL 9000-level stuff of movie nightmares, which doesn’t yet exist, and no one yet knows if it will.
The benefits of teaming the two
When automation and artificial intelligence come together in present-day usage, there are serious benefits to be had. So let’s examine some of the ways they complement each other.
1. Like we said: AI is a form of automation, but some types of automation are entirely devoid of AI. Workflow automation, for instance, can fill in documents and make recommendations without AI. But when AI is added to a workflow solution, a human contributor or gatekeeper can be subtracted from the equation, so that AI-empowered step or steps can be completed in zero time.
2. Non-AI software can automate tasks where it’s highly certain what a human would do or should have done, like forwarding a document for a required review. It does this via conditional logic: the structured data captured in a certain field will dictate what the software does next. By adding AI, though, automation can address more complex situations where unstructured data—which makes up 80-90 percent of the data in most organizations—is involved. An AI risk management platform, for instance, will be able to analyze unstructured data to recognize risks and then recommend a mitigation action. Non-AI software would not have been able to do this, or would have needed an enormous number of fields to be filled out by human users.
3. AI can be self-training, thanks to machine learning. By analyzing unstructured data and through repetition of processes, it can hone its ability and efficiency, which in turn optimizes the automated processes it’s powering.
4. Natural language processing (NLP) is another facet of AI that can benefit automated systems. An example would be sentiment analysis of NPS responses, where an AI tool would read those responses and identify where there are potential issues, or extract insights you can leverage for your purposes. A platform like Qandai, for instance, automates the process of reviewing sales calls in order to call out insights from those calls.
5. Moreover, a trained AI tool can monitor data in real time to detect risks or undesirable trends far faster than a dashboard that relies on user engagement. By alerting users to potential problems or even triggering response automations, these risks can be mitigated far more proactively.
6. AI and automation are most effective when used to target problems or take on tasks that are high-volume and low-to-medium complexity, which are highly time-consuming and tedious, but where there’s high risk if a human isn’t careful; unfortunately, “tedious” usually entails “higher chance of human error.” High-volume tasks also means there’s more historical data available to train the AI. One example where this targeted approach to AI-embedded automation offers significant value for an organization is in software for reviewing complex legal contracts, where risky clauses or language can be red-flagged for attention.
7. AI and software-based automation work best when they’re developed iteratively, meaning software architects and designers should work with historical data and subject matter experts within an organization to develop initial versions for rollout, possibly only to a limited area of your operations. Once you’ve put these pilot versions into use, you’re able to make iterative updates to AI models or software to drive steady improvements and broader implementation.
8. Properly implemented, automation and AI can drive quick time-to-value and ROI, and ladder up to even more improvement. For instance, you might deploy a no-code workflow automation solution to quickly create, streamline and accelerate business processes, then use targeted AI to remove even more human intervention in the processes you’ve automated.
To sum up: Automation software and artificial intelligence are highly complementary, each bringing different strengths to bear on business challenges. By taking advantage of their respective strengths, dramatic improvements in cycle time, efficiency, and quality are within reach.
About the author
Justin Silverman is responsible for leading the Mitratech product management team to drive new innovation, strong platform integration, and streamlined user experiences that bring differentiated value to Mitratech’s customers. He brings over 15 years of product management and strategy experience, including many years in legal technology.