Adaptive Processes and the autonomous enterprise

What are adaptive processes? How do adaptive processes relate to the autonomous enterprise and how does AI enter the story? Sarah Burnett explains.

In my recently published book, “The Autonomous Enterprise” I introduced the concept of adaptive processes. These are processes that, empowered by AI, can be easily adapted to changing requirements and shifting demand patterns. When managed within governance frameworks, adaptive processes reduce the costs and overheads that traditional change programmes incur. In this article I explore the concept and discuss the shape of things to come in the enterprise.

Today, opportunities exist with process intelligence and mining solutions to capture detailed information about a process to re-engineer it. The solutions find how things are done and generate insights about processes that are usually buried deep in daily business operations. With insight comes the opportunity for process optimisation and transformation. For example, improving a process by adding a conversational interface to it, or embedding AI within data entry forms and then using a combination of AI with RPA technologies or application programming interfaces (APIs) to automate more of the process.

In my book, I provide four named case studies each of which demonstrates how business processes can be re-engineered and enhanced. The high-cost areas of the process can then be further automated using technologies such as image processing, conversational interfaces and continually improving exception handling.

There are five major ways that AI can help make processes adaptive:

1.      Reducing exceptions: Reduce exception in automated processes in two different ways:

a.      Reduce process-related exceptions

b.      Engineer out system integration-related exceptions

2.      Process discovery: Find the steps that are related to a process taken both inside and outside the expected process swim-lane

3.      Generate insights: Join the dots to provide the actionable insights that will lead to processes changing

4.      Make recommendations: Make recommendations on how a process can be adapted and improved and show the likely benefits using analysis and what-if-scenarios

5.      Automate process adaption: Depending on the current technology estate of the organisation and their capabilities, get the changes done

1 Exceptions

Exceptions, when an automation comes to a halt because it cannot deal with the case that it is handling, are common-place in intelligent automation. They interrupt operations and make it necessary for people to take some form of action before the control of the process is returned to the machine. An example of a process-related exception is when in order processing the machine cannot recognize a purchase order (PO). It would raise it as an exception to a human, who would check it and make a decision about it. The human finds that the PO came from an older system that used a different mix of numbers and letters and approves the process. The machine will learn from the human to handle those different types of POs.

The jump from static to adaptive processes can be made when with the aid of AI we can continually improve exception handling; for example, if an exception occurs more than 10 per cent of the time then it is generally accepted that it is a path through the process and not an exception and the AI can learn how to handle it.

Another common area in which exceptions occur is in the integration between systems. AI could play a big role in this scenario if we make progress with semantic intelligence and understanding of context. For example, to change a task from ‘transfer this record that represents a service request into back office system A’ to ‘a customer has reported a potential fault with these characteristics’ – a description that is more relevant to system A. The AI learns how that situation should be represented in the back office system A as well as the limitations of the import mechanism, be it through an API, RPA engine or batch record imports.

The result would be a set of processes that are continually evolving to engineer out exceptions with the AI learning from the process.

2 Process discovery

This is about discovering process variations. There may be variations in the way that a process is undertaken by different staff and teams but within its swim-lane. There are also often process steps that are taken by humans that completely bypass the swim-lane. These variations are used by  humans to get work done more quickly than the defined steps allow. Examples include approvals using email instead of ticking a box in a system, or using spreadsheets to store process-related information instead of using the appropriate shared repository that would provide access to the information to the team of co-workers. Variations and outside-process steps indicate problems, and AI can find them and highlight them.

3- Insights

AI can join the dots to identify the opportunities for adapting processes to new or changing requirements. For example, having identified variations and outside-process steps, it can analyse process times and system response times to find what leads employees to step away from the process swim-lane. Is it a slow system that eats into their productivity or have they simply found a better way of doing things?

4- Recommendations

Having joined the dots and found the likely causes of process variations AI can help by making recommendations about how a process can be improved e.g. identify a group of repetitive steps that can easily be automated and recommend the technology to use for it. Some process intelligence solutions already provide this capability as well as what if scenarios that assess the impact of change on process throughput. One use of this latter capability is where change is mandated by new regulatory requirements or policies. The impact can be analysed and the findings and recommendations presented to decision-makers.

5 Automated process adaption

Automating the process adaption very much depends on the state of data integration and other system capabilities of the organisation but I believe we are heading that way – when processes become self-adaptive. That is when AI armed with good data, improved understanding, and training can make the necessary decisions to adapt a process. Take the example of the exception that happens in one out of 10 cases that come through a system. It is clearly not an exception but a part of the process and so with good data the AI can analyse information, learn and adapt its processing to handle these. In the case of the POs generated in an old system, with the right data and access to relevant information about the old system, the AI can learn to adapt the process to handle those POs.

To conclude, adaptive processes have continuous optimisation and change management built into them powered by AI. They can be implemented quickly within governance frameworks helping enterprises improve their operations without the need to initiate costly change programmes.

By Sarah Burnett, analyst and author of “The Autonomous Enterprise”