Why Hyperautomation Needs A Bottom-Up Approach

Updated: Dec 3, 2020


How is it that we are inundated every day with advertisements telling us about the benefits of artificial intelligence, with companies spending more than $50 billion a year on AI solutions, yet nobody you or I know is likely using AI to solve their daily business problems?


Why is it that 87% of data science projects never make it into production, and the few that do only deliver analytic dashboards that use data to draw attention to business challenges, but don’t use data to proactively help us solve those challenges through automation?


Where are the companies building AI solutions for everyday people? After all, it is the employees who have to deal with the problems that can be solved through automation, not the enterprise itself.


In this post, we will share why we think there is a better way to operationalize AI-driven automation called “hyperautomation”, and why this promise must be delivered from the bottom-up in order to be successful.


One reason AI strategies have failed to deliver is that AI is only one piece of the intelligent automation puzzle. In order to deliver on the promise, there needs to be a way to:


  • Analyze our business processes

  • Identify and accommodate different data sources

  • Identify the minimal data needed to drive action

  • Apply algorithms to the data

  • Automate activity based on the analysis of the data


This drove us down the path of building a platform that combines AI (specifically machine learning), with robotic process automation (RPA), and process mining in order to deliver a complete end-to-end solution. This approach may sound overly complicated on the surface, yet it is proving to provide real benefits and was recently validated when Gartner gave it the name “hyperautomation” and listed it as their top strategic technology trend for 2020.


However, for hyperautomation solutions to work companies must also abandon the top-down approach to AI solutions that have failed them in the past, and instead embrace a bottom-up approach. Here’s why.


Why Top-Down Approaches to Hyperautomation Fail


One reason AI-driven solutions fail to deliver results is they have traditionally been too difficult and expensive to purchase, customize, and deploy. In the traditional top-down approach, the CTO or CIO would develop an AI strategy and work with large outside vendors over a period of several years to bring in new tools that could be used to solve problems at the enterprise level. This makes some sense in that AI -- and ML, in particular -- are fueled by large data sets that are typically managed from the top down in an attempt to minimize data silos. However, these giant corporate AI initiatives too often turn into a technological arms race, a continuous building of capabilities with insufficient clarity around the problems to be solved.


I recently spoke with the chief strategy officer for a Fortune 500 company who has sunken three years and more than $5M dollars into building an arsenal of AI tools and software. He boasted that he must have spoken with every vendor out there and tried or purchased most of the available offerings. He was also excited to have recently hired an expensive top-level machine learning engineer to utilize all these new toys. But, when I asked him what problems he was solving today, he looked visibly frustrated and seemed at a loss for words. Without a hint of irony, he explained that that was what he hired the machine learning engineer for - to use all the tools to find important problems to solve. I was embarrassed for him, but unfortunately, his approach is fairly typical. He is now the owner of an expensive solution, in search of a problem, and has shown few results.


The most infuriating part of this top-down approach is that while it often fails to target or solve any real problem, almost everyone else in the organization is sitting on a mountain of real business problems that are just screaming out for hyperautomation. But, the average person doesn’t have a million-dollar budget, so their problems go unsolved. Meanwhile, billions of dollars are spent each year on corporate AI strategies (soon-to-be corporate hyperautomation strategies) that will never reach them. At best, the average person’s budget affords them BI tools that empower them to report on problems using data, but not to act on them. In other words, problems are going unsolved, frustration is mounting, and massive amounts of time and money are being wasted. To be blunt, it’s pretty messed up.


Why Take A Bottom-Up Approach to Hyperautomation?


How do I know that most people have unmet business challenges that could be solved by hyperautomation? It’s because, over the last five years, I have spoken with thousands of people about their most frustrating business problems and why they’ve gone unsolved. Many are people like “Bill,” a senior testing engineer, who came to us after he had scoured the market and failed to find what he was looking for.


Bill is responsible for hundreds of thousands of consumer-grade security devices manufactured and sold by one of the five largest tech companies in the world. Each part of the device is manufactured in a separate factory, and each factory produces a data log containing millions of rows of data on the parts being produced. If a part failure goes unnoticed, the cost of replacing the faulty devices after the customer has received them severely impacts profit margins and customer goodwill. More importantly, a catastrophic large-scale failure could cause irreparable damage to the company’s reputation.


Bill also knows that since he is ultimately responsible for quality control, he could get fired if too many devices fail in the field. So, he decided to put a tremendous amount of effort into due diligence when looking for a solution to automate the detection of part failures. His requirements included:


  • Process mining to identify all the critical stages in the manufacturing process

  • Monitoring of millions of rows of data from the part-manufacturing logs

  • Predictive analytics to identify when a part had failed or was about to fail

  • Prescriptive analytics to automatically send an alert to the right people in his organization with the appropriate measures to take to fix the issue

  • A consumer-grade application that was easy to use and deploy


Bill told us he first tried to find an off-the-shelf solution, to no avail. He found no vendors that met his needs “out of the box”. When he asked about customization options, the sales people would try to change the subject because they knew that the 10 to 20x cost associated with customizing their solution would prove to be cost prohibitive and would ultimately kill their sale. The problem was that the offerings designed and developed by these product-based companies solved a very specific set of tasks and did not offer the flexibility to accommodate many customer’s one-off requirements. Their entire business model is based on achieving economies of scale by selling the same few products over and over again to everyone, regardless of their requirements.


Frustrated, Bill then set up calls with a few custom AI consulting shops to see what it would cost to build something from scratch. The quotes all came in at $1M or more. Bill, like most, does not have a $1M budget to throw at every problem. So he did what all of us do: he left this potentially job-threatening problem unsolved for months because he had no other choice. That is, until he found RelayiQ.


RelayiQ - Democratizing Hyperautomation


Now, Bill is working with RelayiQ to identify and customize an off-the-shelf ML algorithm and turn it into an app that he and his colleagues can easily use and update. The RelayiQ platform makes it easy to develop no-code apps and then deploy them into your own custom app marketplace. Think of it as being like a hyperautomation app store for your enterprise. The RelayiQ platform features:


  • Connectors for many of the most popular data sources

  • A library of algorithms designed to work with a variety of data types and use cases

  • A sophisticated notification engine to automate the delivery of prescribed solutions to everyone in your enterprise who can and should act on them

  • Self-service marketplace to socialize all the AI applications in your enterprise


We prefer to say that we sell business outcomes, not products, and as a result our pricing directly correlates to outcomes to better align with customer goals. Bill now has a solution that monitors all of the data logs across his entire manufacturing process for potential part failures. In the event that a failure is detected a ticket is automatically sent to a service tech to test the product manually. Likewise, root cause analysis is used to automatically generate a Jira ticket for the plant manager to go and investigate the source of the problem further. Moving forward we are also looking at using natural language processing with the call center data to identify batches of devices that are failing in the field. If enough customers call with complaints about devices manufactured in the same time period the app will send an alert to Bill asking if he would like to automatically ship new devices to all the customers potentially impacted, before the products even fail. The business outcome? Better quality devices, happier customers, fewer returns, and job security.


Bill is thrilled that our solution is an order of magnitude less expensive than the lowest quote he received from the other vendors - and it is being delivered in less than half the time. He already has three more apps lined up that he wants to deploy into his RelayiQ marketplace, and we are providing training so that he can build those apps on his own in the future, saving him additional time and money. He has introduced us to his business analyst team and as a result, they are eager to interview people throughout the organization to find similarly important and low-lying issues that could benefit from custom hyperautomation apps. They are excited to move from simply making dashboards to empowering people to deploy actual solutions that can really move the needle and deliver measurable business impact. The RelayiQ solution is proving to be transformational for their business unit’s operations and discussions are already taking place with other business units to replicate the success.


How To Get Started With Hyperautomation Today


We are solving dozens of problems completely different from Bill’s (not everyone is a testing engineer). From predictive spending for finance to data unification for marketing CRM to automated sales reporting and quota alerts, the sky’s the limit with the RelayiQ platform.


When building out the solution, we operated on the assumption from day one that everyone’s problems would be different and each solution would need to be customized. However, we also knew that there would likely be significant overlap in data sources, core requirements for algorithms, and the need for an enterprise-grade application that almost every organization could use and manage on their own. RelayiQ has built an affordable and approachable hyperautomation solution developed with business consumers in mind. This bottom-up marketplace approach to business automation is proving to be highly effective, and once adopted it quickly gets applied across numerous use cases and spreads throughout the organizations it is deployed in


If you want to move from using your data to simply report on problems to actually using data to proactively solve your problems, with intelligent prescriptions and automation, contact us and we’d be happy to provide a demo.


306 views0 comments

Recent Posts

See All

Astreya announces the acquisition of RelayIQ

Astreya announces the acquisition of RelayIQ Company reveals strategy to expand its technology-enabled, managed IT services SAN FRANCISCO / PR Newswire - Astreya, a leading global IT solutions provide