BPM development goes beyond low-code with AI, RPA

Low-code BPM wins favor

Use of low-code process development tools has increased rapidly over the past three years. About 67% of businesses said they use or plan to use low-code process software in 2018, up from 43% in 2016, according to Forrester Research. Low-code BPM development allows business and DevOps teams to collaborate on the creation or customization and deployment of process applications, said Rob Koplowitz, an analyst at Forrester.

Generally, there's more sizzle than steak in the hype about AI capabilities, but that's not true for AI in process development.

Once advertising agency Leo Burnett Chicago implemented a low-code product, administrative costs for routing financial transaction requests and approvals shrunk and the use of the tool grew rapidly, said Steve Hudgin, director and program manager at Burnett. "Using low-code tooling has significantly streamlined our operations and allowed us to centrally manage our financial processing with a smaller team," he said.

The advertising agency, which uses BP Logix BPM software, was able to phase out several development tools that were not designed for process management.

Once business teams can configure development on their own, they are not reliant on IT to make updates. This means they can react quickly as business needs change.

AI features come to process development

While low-code features reduce development complexity, AI can foster real-time software delivery and more personalization for a better customer experience. AI engines are a natural next addition to BPM development platforms, because machine learning enables better code output and higher-quality code to support app development, Wong said.

Rob Koplowitz

"Generally, there's more sizzle than steak in the hype about AI capabilities, but that's not true for AI in process development," Koplowitz said. AI brings machine learning capabilities that analyze many types of process forms, such as loan and job applications, and make recommendations without human intervention.

With AI, process managers can choose a function in a process for analysis, or a user can request information or status through an interface, such as chat. This eliminates the need for phone or email exchanges that often cause more friction than a natural language interface does.

AI is often used to provide analysis and action recommendations on unstructured content through a process map -- without human intervention and with greater accuracy than humans usually deliver. For example, a company that uses an AI-enhanced tool can streamline the tasks associated with bringing new hires aboard by letting machine learning take over so that employees don't need to input the same data -- such as names and addresses -- on many different forms. As companies increasingly move their process automation workloads to the cloud, AI provides a base system that customizes processes based on machine learning results from all connected internal and external systems. On the client side, a citizen developer who does not necessarily understand development processes gets wizard-like prompts. On the back end, AI can induce process optimization with root cause analysis when something is not executing within service levels.

The augmented development side of AI applies machine learning to application models and integration patterns.

Jason Wong

"IoT and other event-driven applications will increase the number of endpoints connected to process applications," Wong said. AI will help businesses apply historical, predictive and other forms of analytics to events, enabling automated decisions to be made on each event.

Marry RPA and BPM development

BPM vendors, such as Camunda Services and Pegasystems, now offer integration between BPM development tools and RPA software. The goal is to drive overall automation to improve customer services.

"RPA and BPM together provide a strong platform for digital transformation organization-wide," Wong said.

RPA is able to handle workloads that traditionally have been difficult for process vendors to address because of the complexity of the back-end integration. Traditional back-end integration processes called for complex coding in data workflows.

"RPA focuses on client-side integration, bypassing costly and complicated back-end integration," Koplowitz said. Integration of RPA with BPM breaks down barriers between disparate applications and services and opens opportunities for businesses to use any type of software in process automation.

Processes most suited for RPA have a high transaction throughput of structured digitalized data in relatively fixed processing paths or user interfaces. These do not change frequently and are rule-based activities, Wong said.

RPA tools work best when they have direct access to the data and applications. Thus, processes are more suitable for RPA if they have little or no need for remote access tools. Also, processes with unstructured data are not good fits for most RPA tools; processes that are rule-based can be.

Just because RPA relates to robotic software, many businesses mistakenly use it for the wrong reason, Wong said, thinking they can automate certain jobs out of existence. Gartner's research shows that RPA will have limited impact on top-line growth for 50% of organizations that use it for labor reduction and not improving business value.

Organizations with limited in-house BPM expertise would be wise to evaluate emerging BPM PaaS platforms, Sharma said. He sees BPM PaaS as a means to increase the participation of business users in the process cycle and enable them to rapidly compose applications.