Skip to main content

The next frontier in AI: Automations creating their own automations

(Image credit: Image Credit: PHOTOCREO Michal Bednarek / Shutterstock)

Artificial intelligence (AI) has now fully entered the mainstream. Whether it’s a self-driving car or a voice-enabled music system, automation has become an exciting cultural touchpoint, and one that sometimes we can’t even tell is automated. We communicate with our bank and make payments via a virtual assistant, we allow automation to select which films we should watch or which clothes to buy, and in some businesses, we’re even working alongside digital colleagues to help us offload day-to-day admin tasks.

Although many are still getting to grips with the myriad ways that automated systems are infiltrating our lives, tech companies, like IPsoft are already looking at the next frontier of AI. The next phase in the development of this innovative technology is looking at the ability to design automations that create their own automations.

Autonomously solving business problems

So, what does this really mean in practice? With cognitive AI, which has developed problem solving capabilities modelled on the human brain, software can now automate processes that recommend new automations, aimed at making the processes quicker and more efficient.

By applying this in the workplace, businesses can look at streamlining processes, improving workers’ productivity and driving better business outcomes. A great example of this is an automotive manufacturing company that uses robotics and automation to build cars. Engineers can programme the automated machines to not only perform the tasks needed to actually build components but at the same time, to also scan and apply more efficient methods of car manufacturing to future tasks.

Analysing processes with ‘virtual engineers’

One way of looking at this is engineers programming an AI solution to create its own ‘virtual engineers’. As with virtual colleagues in back-office functions or call centres, designed to take on repetitive tasks from human workers, ‘virtual engineers’ represent the next evolution in automation.

Some companies are already doing this with IPsoft’s fully automated platform, 1Desk, composed of virtual engineers and a virtual colleague named Amelia. Both systems can watch a human worker perform a task, study that task, and recommend how best to turn this task into an automated process. In the complex world of IT operations, these automated workflows span multiple systems and can then perform intricate tasks, such as trawling through data to pull out insights that would normally take human workers away from more valuable functions, in this case, analysing the insights themselves.

Automated observational learning

For example, if an IT engineer wanted to run diagnostics on a SQL server related to an incident, they would have to analyse all logs and data files and add them to an incident log, which would be both time-consuming and may result in inconsistencies due to human error. A virtual engineer can master this process by watching how it’s carried out and recommending how it be turned into an automation. If the human engineer agrees, the process can be added to the virtual engineer’s knowledge base. Going forward, if a similar incident occurs, the virtual engineer previously-learnt autonomous skills and can do all the ‘heavy lifting’ of data to create drive results more quickly and accurately.

Virtual-human collaboration

It’s not as simple as putting all your work in the hands of a virtual engineer. Not every task can be fully automated as some require human assistance, for example, if the virtual engineer comes across an unforeseen situation that it has not encountered before. When this situation occurs and automation ceases to work, virtual engineers will ask for human assistance and then note the human response so that they know how to react in future.

One way to imagine this is by using the example of an HR team automating onboarding processes for new employees. The virtual engineer has run this exact process hundreds of times before and they handle the end-to-end onboarding programme which includes processing all of the legal documentation, insurance, pay information, benefits and even organising the setup of new equipment such as their laptop and mobile device.

The process is 95 per cent complete, however the virtual engineer reaches an impasse. The company is onboarding its first Australian employee, and the process has only ever been conducted for staff in the UK office. As the virtual engineer is not programmed to procure devices outside of the UK, it asks a human for help. AI’s intelligent learning capabilities mean that it can monitor the steps the human takes to solve the issue, absorb this information, and now ask the HR team for permission to add international device procurement to its automation skillset. In short, the automation has created its own automation.

Who makes the decisions?

The first thing people are going to wonder is “what if AI becomes so smart that it learns to override human control?”. AI is programmed to guarantee this does not happen by following established procedures that have been set out for it with precision. Responsibly created AI will never deviate from the strict parameters for which it has been programmed.

In a banking setting, AI has developed to the point where virtual engineers can approve and reject credit card payments, give mortgage advice or suggest new products based on the customer’s activity. However, they cannot autonomously reveal credit card data to consumers in tertiary conversations and customers can rest assured that the AI system cannot access a card’s payment function without the customer’s approval and following a thorough security check.

Taking back control

Responsibly-created AI, or what some are referring to as ethical AI, is human-controlled: meaning when it learns a new task, it will always ask a human whether or not it can apply this to future processes. Essentially, automations can only create automations if they have been approved to do so. If a human says no, the automation is rejected and the AI never applies what it has learned. However, in those situations where a human says yes, an AI system quickly learns how to automate automations, bringing new levels of speed, efficiency, and innovation to a company’s operations.

Most businesses are already applying some form of automation to the work they do. And as more companies adopt AI for this approach, employees and customers alike will find these autonomous actions ever more useful for supporting them in the workplace and in their daily lives as well.

Martin Linstrom, Managing Director UK&I, IPsoft
Image Credit: PHOTOCREO Michal Bednarek / Shutterstock