A Strategic Approach to Precision Process Automation

The global tech labor arbitrage trade has matured since its start in the Clinton era.  We've witnessed real estate in places like Mumbai and Singapore appreciate to levels rivaling NY and London.  The rising tide lifts all boats.  As global tech labor costs have flattened somewhat, software has gotten more expensive to produce and maintain.  The market is betting that AI will largely alleviate this by supplanting much of software engineering as a practice.  However, until this plays out, being judicious about when and how to use software is more important than ever.

 

While it's constructive to build software systems as "black boxes," it's folly to view them as crystal balls.  Increasingly, software is operations.  That engineer who knows the code like the back of his hand is your foreman.  Code is not static, it's constantly evolving to meet business demands and improve efficiency.  "Software Engineering is programming integrated over time."  The motivation and history building up to your system's current form may not be captured within, but is often valuable institutional knowledge.

Japan - Even a High Quality Process Can be Automated

Progress in AI is pushing the boundaries of what computers can do but we're still far from being able to say, "Anything a human can do a computer can do."  However, for the most part we can assert that, "Anything a computer can do a human can do."

In America, we're innovative.  We're quick to dispense with outdated traditions and try new things.  Sometimes too quick.  As Churchill famously said, "The Americans can always be trusted to do the right thing...after exhausting all other options."  Japan is more conservative.  Craftsmanship is a national point of pride and is truly impressive.  There are ceramics makers that have been in operations for hundreds of years, perfecting and carrying forward the same process one generation at a time. 

I once heard a German tourist say, "Impressive but in Germany we'd make a machine to automate it."  A machine can replicate the process if enough care is taken.  However, intuitively we doubt this assertion.  The tendency we observe is towards lower quality by machine production.

Once automated, the creep toward commodification begins.  Corners are cut.  The Pareto Principle states that 80% of the effort is required for the last 20% of quality.  Automation pounds away this 80% effort to arrive at an 80% "good enough" product.  Low quality inputs are substituted to cut cost.  Unimpeded, the profit motive drives down cost at the expense of quality. 

The craftsman doesn't make such compromises. For a time, Japan's craftsman spirit was a strategic advantage.  It allowed the pursuit of quality to prevent them becoming just another commodity producing post Imperial era colony.  Instead, they were able to compete with and even displace domestic US brands.  By producing superior products, Japanese brands like Honda went from selling cheap "rice burners" to dominating the US standard and luxury markets.

However, even high quality product production can be automated and its maker's strategic position overturned.  If the demand for higher quality is apparent, competition can be counted on to do just that. 

The Iterative Mechanical Turk Approach to High Quality Process Automation

The mechanical turk approach to new product launch involves having engineering build out an MVP solution but filling in gaps with humans.  The result is ostensibly a machine, the inner workings of which are more cyborg-like than robotic.  Some elements are performed manually by humans, some are performed by the computer.  This approach is especially useful for fast deployments, high complexity processes, and/or low margin businesses.  But what happens post MPV? 

"The first rule of any technology used in a business is that automation applied to an efficient operation will magnify the efficiency. The second is that automation applied to an inefficient operation will magnify the inefficiency." 

-- Bill Gates

Having built out the operations team, allow them to gain expertise in the problem space.  Informed by their experience, take your best shot at delineating the 80/20, effort/impact split.  Have the system automate the negative space, 20% effort process necessary to produce an 80% impact.  Focus the operations team on the remainder.  Repeat the process, chipping away at the overall process in small, high impact iterations.

Allowing 80% of the human effort to remain sounds counterintuitive. However, having engineering focus on the "easy 20%" increases the likelihood that the portion being automated is mature, efficient, and ready for automation.  It means lower complexity endeavors and faster delivery cycles.  It means reduced chance of scope creep, unnecessary features, or missed deadlines.  It means that each iteration carves out 80% of the existing value-add portion of the operation, assuring high impact-to-effort ratio.  Finally, it alleviates anxiety from the Operations team that Engineering is out to replace them and thus incentivizes a cooperative team effort.

The Loach Beneath the Willow

Although not coined until later, the 90s dotcom bubble bet that "software would eat the world."  It took the outsourcing boon, cheap capital, and lots of hard work over the following 20 years to facilitate the manifestation of this dream.  Now that it's largely played out, naive software based automation and outsourcing can't be relied upon to realize the same advantages as they once did.  As the Japanese saying goes, 「柳の下にいつも泥鰌はいない」Just because you caught a fish there once, doesn't mean another will always be there.  As expressed in the anxiety of the dotcom aftermath, "Who Moved My Cheese?"  Fish, cheese, or otherwise, business strategy must adapt to the current realities of the economy.  A mix of outsourcing, operations staffing, iterative software automation, and creative thinking should be employed to stay competitive.