Intelligent automation: The good, the bad and the ugly
Despite so much hype and a near-constant stream of chatter, there are few who actually understand what intelligent automation means
By Besjunior / Shutterstock
The real challenge facing intelligent automation is a lack of clear definition and a very noisy market, says Cortex COO Eddie Watson. “It’s really a fairly large federation of solutions that provide orchestration and automation if done properly,” he explains.
Outside the overhyped promise of some new magical AI, bots, or other fad, not much has changed when it comes to the role of automation. It is simply that the technology’s functions and capabilities are becoming more extensive - Watson adds. He should know - since starting his career in chemical engineering and process management, Watson has automated processes from critical business decision making right through to IT services.
We sat down with Watson to discuss how far intelligent automation has progressed, and what to look out for in 2018…
AIIA: Automation is not a new thing. In fact, many people believe we’re now in the third wave of automation—if we think of the first wave as being defined by machines relieving humans of basic manual labor, and the move to a service-based economy as signaling the second wave. How will this third wave of automation be defined?
Eddie Watson: A lot of organizations now aspire to implement intelligent automation without realizing that they don’t have all the foundations in place, the existing systems or the datasets they need.
LATEST WHITEPAPER: How to deploy intelligent automation for multi-vendor activation
One of the things I talk about is the gaps and the overlaps: A lot of business systems have overlapping capabilities and often we get involved with client operations where they have an aspiration to implement some form of automation when they already have tools that cover that task or function. The real challenge is that the processes are made up of a lot more intangible stuff than just the tasks.
The things you’ve got to worry about are the gaps. Often an organization has a vision to digitalize a process only to suddenly discover that key data sets are incomplete or maybe don’t exist at all. Also, often primary capabilities or systems are functionally incomplete, or don’t do what you need to do. In a manual world, these gaps are filled with small manual compensatory tasks supported by armies of spreadsheets that are typically undocumented and unregulated. These kinds of tasks are the typical target of simple RPA, or DTA tools which far from helping, just entrench and compound the problem. Dwelling in those gaps are hidden requirements that are not of the process automation world, they are analysis and decision driven actions that belong to the world of intelligent systems. A world of pain will arise which will prevent you from being successful if you don’t identify those early upfront. This way, you don’t get halfway down the line assuming a simple process based outcome and then suddenly discover a major gap, that can only be bridged with cross-domain intelligence.
When it comes to intelligent automation, you must take an intelligent approach; recognize most of the intelligence in your current workflows, processes and tasks is in the heads of the people in the process; because people are key. To transition a current operational activity into an automated state many departments and affected people should support and contribute. These subject matter experts must be the ones who define the knowledge and expertise that drives the decisions that result in the right outcome. The business must own and participate in the automation of the process, not the technologists. Ask yourself this: would you get on an airplane that had been automated by a software coder, or a pilot?
AIIA: Can you give an example of what happens when the whole team is not behind an automation project?
EW: We once took on an incredibly ambitious program for an organization that wanted to effectively optimize and make automatic decisions for the configuration of a voice network. That meant that key members of the team responsible for taking orders and writing contracts needed to be on-board to capture the process, knowledge and decisions.
Both the billing department and the configuration engineers needed to be on board, as well as those monitoring the network because that’s where the information on which the model is designed and optimized comes from.
The project eventually halted because these were complex cross-domain processes involving more than seven departments who were required to agree on how this was to be done. They all needed to align with the mapping, and capture that into the processes. Starting by capturing complex processes across too many domains with various level of expertise and knowledge in to a single process just results in failure through complexity.
Essentially, the different departments could not agree on a common set of dialed number breakouts to charge customers. The networks team used one set, the commercial team who were selling contracts were using another. They had not established the political will or the strength in the organization to drive the decisions required for complex processes like this. Apart from the fact that this is the longest and most fraught route to a solution.
You cannot successfully automate or orchestrate something directly into a single complicated process where each domain requires ownership of information and the definition. Differences and disagreements are inevitable between those various views of the world. The solution is to maintain domain ownership by starting with each domain separately; Start simple. Even use different technologies, or adopt what is already there using a universal platform to fill the ”whitespace”. Once each domain has an operational sub-process; orchestrate them together using decision driven automation to ensure acceptable outcomes. This guarantees success; each group retains strategic control over their domain. Each domain retains their own context within the larger process. The complexity of the larger process, the deeper intelligence technology required and the related data arbitration are all kept away from subject matter experts in each domain ensuring it is manageable. Automation that can be managed in production through small and incremental changes by the subject matter experts; starts successful and remains successful.
AIIA: What are some of the most common mistakes you see organizations make when they’re undertaking automation projects?
EW: A common mistake is that organizations automate in an uncontrolled, unmanaged and inaccessible way. What I mean by that is that organization end up with little islands of inconsistent specialist automation tools – typically where each one has a technologist who cuts code, or manages the configuration for the subject matter experts but each island sits in different departments with incompatible skill sets. If you start building your automation capability across a lot of different tools like this, then you can have some success in basic, simple task automation, but you will have a real problem moving to anything more meaningful. It often rapidly becomes fragile, and eventually falls into disuse, at which time a new cycle of basic automation starts; a kind of “Groundhog day” of automation.
The most important thing is that you need to recognize you don’t know everything about your operations, and that you never will, so it is futile trying to define all permutations of activities for a machine. You need to think about methods to capture the happy path and have very clearly defined rules for dealing with exceptions to that path. These should ultimately, after you have exhausted heuristics, deep learning and machine learning, escalate to a human-in-the-loop. Our smartest most automated clients accept the concept of retaining and designing in a human-in-the-loop for exception management. This means you need to think carefully through who is going to manage this automation and orchestration, how small and incremental changes to each of those simple islands and complex processes are captured, and how it will impact everyone’s role. What will the future look like, and plan for that through a carefully executed transformation plan.
“The problem with rigid automation and particularly scripts is that you solve yesterday’s problem and you want to be solving tomorrow’s problem.”
Automation must be in a transparent environment which is easy enough to change and adapt to the changing business environment.
AIIA: We’ve talked about the bad and the ugly—now let’s focus on the good. How should leaders be managing the process of automation implementation within their teams in order to ensure widespread acceptance and success?
EW: You must have an executive sponsor> This is more difficult than it appears because success is dependent not just on the automation project, but on implementing and maintaining the changes to the organization outside of the automated processes. That’s a particularly difficult one because that person is often putting their career on the line, and they need to believe in the power of automation for it to work.
The second thing is to start to develop a center of excellence (CoE) from your initial deployment. Create a small group of people who effectively know the business and know the technologies, into which you can map key advisors and roles to work together with the executive sponsor to define and maintain the business outcomes that you’re wanting to achieve.
“Don’t get stuck in the technology. Too many companies do random acts of automation”
The third piece is don’t boil the ocean; start simple and adopt an agile approach. We have seen many failures where people blindly want to adopt an agile approach but they don’t really understand that agile means that you cannot, and should not define your project deliverables at the outset. This is a waterfall approach and in automation is the shortest road to failure to deliver … our statistics repeatedly show 70 - 73% of automation projects fail because the output was achieved not the objective.
AIIA: What role does the gaps and overlaps conversation have when it comes to defining those parameters for success?
EW: Having processes defined is an area where businesses delude themselves into thinking that they know what they do, but unfortunately that sort of standard operating procedure documentation usually only captures the action pieces, not analysis, decision making and very little context. Not only are they usually out of date, but more than 75% of the knowledge doesn’t exist in documentary form. This is another reason why you start simple, and start with short-run tasks. Build a hierarchy using exception management to capture the knowledge from the subject matter experts in production.
One of the other key things that businesses make a mistake with is that they try and orchestrate or automate directly against the existing manual standard operating procedure and that is not a good place to start. Capture the process as it is operating today, preferably directly from your subject matter experts even if this is inefficient, broken, fragmented, or otherwise compromised. The technology you use should help you weave things together, fill the gaps, manually intervene and provide performance indicators for improvement. This means you will need a no code platform: most subject matter experts do not make good coders.
To summarize, what you need to do is take those standard operating procedures, define what the objective is that you want to achieve, identify what the inputs, the outputs and the decisions are that you need to make and then define how you move with those inputs, outputs and triggers to get to that objective. That flow chart will look very different to the standard operating procedure for that same objective.
AIIA: Looking ahead to the next 12 or 18 months, what trends and changes do you hope to see in the intelligent automation sphere? What do you predict will be the big focus areas for 2018?
EW: It doesn’t move as quickly as you might imagine; let me put it that way. I think that often people’s vision outstrips the reality of where they are.
If you clearly understand your objectives, adopt the right approach, and pick the right tools you will move a lot quicker, but because of the hollow promises in the market from ineffective techniques, and overpromises of poor technology, most things aren’t moving as quickly as they could.
I think what we will see is the automation imperative over the next couple of years will force organizations to adopt a higher profile and a more urgent orchestration agenda. I think we will see the emergence of real advanced AI use cases using less structured data and more human interaction because advanced AI will find a natural implementation platform in orchestration and automation.
For example; Amazon and the companies that started in the digital era have services that are so integrated and feel so neat and seamless that they set the benchmark for customer service an order of magnitude higher. If other businesses are not going to move with that, if they’ve got a lot of history and legacy, a lot of old technology, old thinking; they’ve got an insurmountable challenge without Intelligent Automation.
They need to rapidly formulate a very clear vision of their future and they need to have the will to achieve that and that’s got to come from the very top.