Reimagining Robotic Process Automation

At Clear Ventures, we looked into the reasons behind the widespread dissatisfaction with today’s RPA industry. After several discussions with large enterprise companies in our own portfolio, it became apparent that RPA’s poor success rate can be boiled down to three primary factors:

• High upfront cost: Contrary to Humans, who learn over time by doing and adapting, RPA ossifies current business processes in software. This mandates a great deal of upfront work to ensure that the business and associated exception cases are clearly captured and programmed. Upfront costs are significantly higher both for bots and, commonly, business consultants who first must optimize the current business process prior to automation.

• High maintenance cost: In theory, a diligent job upfront can reduce ongoing maintenance cost, but the reality turns out to be quite different. Many business processes are simply too complex to effectively document all of the exception cases upfront. And because business requirements change, the processes do as well. This leads to high maintenance costs that often turn out to be the hidden “Achilles’ heel” of RPA.

• Finger-pointing (the blame game): When the automated process fails to work as expected, a blame game often ensues. This RPA project finger-pointing typically is a three-way exercise between the business user, the RPA tool vendor, and the RPA tool programmer. Undermining RPA project failure accountability aggravates the frustration and further increases costs.

RPA at a Crossroads RPA is at a crossroads, and we need a different approach as nobody except perhaps the RPA software vendor is being served well in the current environment. As we looked beneath the proverbial “tip of the iceberg”, we came across deeper reasons for the poor success rate of current RPA implementations including:

1. Shiny and unrealistic demos from RPA vendors: RPA vendors such as Automation Anywhere and UI Path have enticed organizations with glitzy demonstrations of sophisticated “robots” taking over previously human-led processes. These demos understandably lead to unrealistic expectations and eventual disillusionment as the actual automated processes are often too rigid to handle real-life exception cases.
2. Inability of the bots to learn and adapt: When exceptions and edge cases arise in a process (as they almost always do), bots are unable to ask questions, learn and adapt themselves. The business user has no way to interact directly with the bot and instead must work with an RPA programmer to make any changes. The additional costs and delays erode the value of RPA.
3. Environment instability: In addition to evolving business processes, a changing environment can also negatively impact RPA. As an example, bots dependent upon a CRM or ERP tool may suffer unintended consequences when a change is made in either. As with the initial automation of the business process, a team of programming experts is frequently required just for ongoing upkeep.
   

   Reimagining the future of RPA

   It became obvious to us at Clear Ventures that the RPA market is primed for disruption. We evaluated several startups and eventually came across Kognitos which takes a radically different approach from RPA. Kognitos’ solution is centered on two pillars:

   • Natural Language programming: Kognitos enables bot and human interaction based upon natural language, eliminating or dramatically reducing the need for experienced programmers. Using natural language slashes both maintenance cost and time to value for the line of business.

   • Run-time learning: Current RPA approaches simply crash or freeze when faced with exceptions and edge use cases. When a Kognitos bot encounters unexpected situations, it interacts directly with the business user. Similar to the way in which humans resolve unexpected hurdles, the bot asks questions, learns, and then applies the learnings to modify and continue the process.

   We have partnered with Kognitos since the days when the concept was just a glimmer in the eyes of Binny Gill, the gifted product author and architect of the vision. We are thrilled that Binny and his team have come out of stealth and have launched their first product. We look forward to continuing this exciting journey with the Kognitos team as they move into the next phase of the journey to reinvent the RPA industry.

   Please visit Kognitos.com for more reading.

How to Reimagine RPA with AI-Native Automation

1. Calculate the total cost of your current RPA program. Pull 12 months of RPA maintenance, incident response, and development data. Calculate total annual cost: licenses, developer time, operations team time, and business productivity loss from bot failures. This is the baseline for the AI-native comparison.
2. Identify the RPA bots with the highest maintenance burden. Sort your RPA bot portfolio by maintenance hours per bot per quarter. The top 20% of bots by maintenance cost are the migration priorities. High maintenance indicates fragile selector dependencies or high exception rates that AI-native handles better.
3. Map the capabilities required for AI-native automation on your top migration targets. For each high-maintenance bot, map the capabilities the replacement automation needs: document understanding, unstructured input handling, multi-system orchestration, or natural language configuration. This capability map guides platform selection.
4. Run a parallel pilot before decommissioning any RPA bot. Deploy AI-native automation alongside the legacy RPA bot for at least 30 days. Process the same transactions through both. Only decommission the RPA bot after parallel operation confirms AI-native produces equivalent or better output.
5. Plan for incremental migration rather than a big-bang replacement. RPA portfolio migration over 18 to 24 months is more sustainable than a big-bang replacement. Migrate by maintenance burden priority, measure cost reduction at each migration milestone, and use the cost savings to fund subsequent migrations.