Watch all the Transform 2020 sessions on-demand here.
RPA has played an important role in helping Mars, the candy giant, automate many of its processes and save time and money along the way. But the technology is in its early days, said John Cottongim, automation director at Mars, who is leading the company’s digital transformation efforts. At Transform 2019, he presented his wish list of new capabilities that would make it easier to scale up RPA for large enterprise deployments.
Ubiquitous AI/ML
Enterprises are still struggling with adding AI to workflows using point-solutions. “There is no framework I have seen that is overarching,” Cottongim said. The industry has yet to create standards for many important aspects of RPA, including process data and AI capabilities. As soon as industry comes up with better ways for standardizing the data, he said, the automated learning can become a more practical task.
Improved shop floor UI/UX
The RPA packages are all great at generating basic bot automations but don’t make it easy for these bots to collaborate with users. Cottongim thinks RPA tools should take a cue from lightweight workflow development like Appian or Microsoft PowerApps to enable better two-way communications with the teams using the bots.
Industry standards
Each RPA platform includes separate file formats and processes for scripting and managing bots. Cottongim believes that an open source approach would help with standardization. It would also allow bot developers to invest more effort in adding value to an interoperable bot ecosystem. This would make it easier for enterprises to weave together best-of-breed components for complex or specialized workflows.
June 5th: The AI Audit in NYC
Join us next week in NYC to engage with top executive leaders, delving into strategies for auditing AI models to ensure fairness, optimal performance, and ethical compliance across diverse organizations. Secure your attendance for this exclusive invite-only event.
Self-healing and self-learning
RPA bots can break as soon as a button in an app moves or changes color. Cottongim hopes that machine learning will make it much easier for RPA apps to adapt to UI changes and eventually even workflow changes. “We need to build in some self-healing capabilities,” Cottongim said. In the meantime, bot development tools currently make it easy to configure and change bots so that downtime is low when things do break. But the long-term goal should be close to zero downtime.
Automated process discovery tools
Cottongim sees three main approaches to process discovery today:
- Mining back-end data logs like Celonis
- RPA widgets that run on an employee’s desktop
- Tools like FortressIQ Virtual Process Analyst that use machine vision to infer what is happening on a user’s desktop
These are still in the early stages in terms of the kinds of processes they can interpret, Cottongim said. He believes a larger company may have between a hundred thousand to a million processes, and that 20% of these may ultimately be automated. But creating these automations at scale will require more practical ways to identify and generate bots in a secure and manageable way.
This is particularly important because, Cottongim says, there is a tendency to focus RPA effort on processes people complain about. In practice, these types of processes tend to be harder to automate because they require a lot of human engagement. In contrast, simple processes that no one ever thinks about end up being much better candidates for automation.
