Optimize or automate

Optimize or Automate? How RPA is redefining Rapid Cycle Improvement

Do you optimize first or dive right into automation? 

To most the answer is obvious. It isn’t. Recently I chaired a Process Excellence conference for the OPEX Exchange in San Diego where leading companies from Financial Services, Pharma, Healthcare and other industries as well as BPM providers and top consulting firms participated in a workshop on that very question. We had all heard the expression, “automating a bad process just produces bad outcomes faster,” so a quick poll of the group produced a clear consensus: optimize first, then automate.

For decades, applying technology to improve the way business processes worked was no small undertaking, whether done through internal development or external purchase of an application or platform. From ideation to implementation took from months to years depending on the scope and costs could run into the millions of dollars. Designing without the customer in mind or failing to simplify the process or not addressing underlying quality issues would inevitably result in time lines slipping to the right, cost overruns, and user dissatisfaction. If you were looking for quick wins, for “rapid cycle improvement,” the preferred path was a Lean or Six Sigma project. Automation came later after the process was standardized and optimized.

LSS process outline

These were the experiences we brought to that discussion supporting the “process first” approach.  But that was then, this is now:

·      Robotic Process Automation projects, start to finish, can be done in a few weeks

·      Actual design and coding of the bot can be done in a few hours

·      Cost to implement can be less than buying a new PC

Let that sink in – an entire RPA implementation can be completed in less than the time it used to take us to set up a steering committee for automation programs of the past.

How can that be? Two reasons:

  • RPA target processes that are clearly defined, repeatable and rules based. The bots execute transactions based on these simple business rules.
  • The robots – or “bots” – are implemented in the presentation layer, the IT layer where information is captured from and presented to the user. This means that RPA implementations do not require changes to core processing and platforms.

This targeting enables fast capture of business requirements, rapid translation into bot design, ease in implementation and lower risk – all translating into lower cost.

automate process

Here are few examples of processes well suited for RPA:

  • Marketing to prospects via email or social media
  • Creating a client profile or customer account
  • Registering a patient
  • Onboarding an employee
  • Processing a customer application, order or claim
  • Updating customer account information
  • Order updates and shipping notifications
  • Response to customer inquiries or complaints
  • Client, customer or patient billing
  • Reconciling accounts
  • Exception processing

By focusing on business areas which require significant manual work and potential sources of delay, implementations like these save hundreds of hours of cycle time and hundreds of thousands of dollars, eliminating labor cost to execute manually.

optimize or automate

Further, RPA’s short project cycle times and small team resource requirement allows for parallel deployment and multiple waves enlarging the scope of what can be automated within a specified block of time. RPA deployment is more like Agile than waterfall, moving quickly from one opportunity to the next.

Now back to our discussion at the conference.

As we mulled over this new information, sharing recent stories of how much digital transformation could get done how quickly, of how much impact could come from only automating a step vs. an entire process, which came first optimize or automate became less clear. A new paradigm began to emerge:

  • Automate first when scope is focused on a process step or two and elimination of manual work not only reduces cost but also maintains or improves quality output.
  • When scope is broader, more end to end, optimization first is preferred, ensuring redesign takes full advantage of opportunities to simplify flow and enhance value.

To be clear, process design and improvement tools must play a key role at the front end of automation projects if they are to succeed – documenting the way work is done, noting variation, identifying key inputs and requirements for outputs, applying lean principles in the design. Some version of these steps can be found in all RPA methodologies. But the focus here is automation, not optimization.

So back to our question – which comes first, optimize or automate?

In the age of Digital Transformation, the answer is – it depends.

globe north

The original text of this material can be found in a pulse article on LinkedIn.

Gregory North
    Senior executive with extensive global experience in operations, strategy and business transformation―including Lean redesign of end-to-end business process, re-engineering of business process and systems architecture, and optimization of outsourced and shared service models. Broad industry background includes manufacturing, electronics, defense, financial services, healthcare, consulting, and business process outsourcing.Exceptional combination of senior executive consulting and communication skills combined with strong business acumen and enterprise process insight tuned for driving transformational change at all levels of the organization. Recognized subject matter expert on process excellence, business process outsourcing and digital transformation.

    2 Comments

    Dharani Kumar

    Dharani kumar Nov 9, 2017 at 7:11 AM

    Very important & useful

    Jun 11, 2017 at 10:36 PM

    The framing of ‘optimize OR automate first’ does not recognize the existence of automated AI which: repeatedly learns from data generated by past operating decisions, and modulates operating decisions to improve and optimize (overall) performance of existing processes, up to their capabilities. The function of giving directions to a regulatory control to implement is called Supervisory Control in industrial production. This supervisory control AI is, in simple terms, “learning from experience how improve performance”.

    After optimization is achieved, sensitivity analysis of the process barriers to higher performance — e.g., constraints or inaccurate measurements — gives directions to process re-engineering and/or re-definition of objectives for a “macro cycle” of another type of optimization.

    So, the answer I suggest is “optimize and automate, and repeat”.

    More specifically: (A) first optimize operations manually, where managers and personnel can guide the evolution of the definition of objectives, what needs to be measured, and what decisions need to be made (yes, in that order) — which sometimes bring to light process design improvements, some easy to implement; (B) automate and refine optimization — e.g., as conditions change; and (C) periodically do sensitivity analysis for process design improvements.

    Leave a Reply