Scaling Human Ingenuity

August 22, 2018
Marin Orlosky-Randow
Managing Partner

Long-range planning, marketing solutions, and QA strategies. Grammar nerd. Classically-trained ballet dancer.

Individuals Don't Scale, But Their Ideas Do

There are many articles simply listing numerous workflow tools that use AI (or claim to do so) to help you make smarter decisions. This is not that kind of article. Instead, I'd like to focus on how to train AI to automate parts of the creative process--the parts of your business that are driven by truly innovative individuals. Because while these individuals may be the key force distinguishing your business from the competition, relying on them for daily tasks will never scale.

Artificial Intelligence, Powered by Human Intelligence

"Contrary to corporate marketing, machine learning isn’t black magic. Instead, it's a class of algorithms that allow computers to perform pattern matching extremely efficiently." -Ashish Datta

Let's start with some basics of what AI is good at, and what it's not. Artificial intelligence, as it currently stands, excels at pattern recognition and large-scale data analysis. It can provide insight into how individual data points cluster together to indicate behavioral trends, or learn how to perform predictable, repetitive tasks of varying complexity. But it cannot generate new ideas, nor reliably draw conclusions that would indicate the best way to improve your strategies. 

For businesses handling a variety of complex processes, or preparing for rapid growth, the most scalable approach is to automate as much of the time-consuming, repetitive parts of processes as possible, in order to free up more time for creative humans to innovate strategically. So the question becomes, how can we translate smart humans’ workflows into automated processes?

Isolate the Creative Parts, Automate the Rest

More can be automated than you think. Instead of trying to automate an entire process or workflow, break it down into individual steps. Isolate the truly creative parts of the process, and then offload the rest to automation. Sometimes this requires articulating in detail the tasks that you usually do by instinct or subjective feeling: the flow of a particular process, your rationale or criteria for decision-making, or the value structures by which you determine one data point to be better-quality than another.

For example, mass content curation typically is higher-quality when performed by humans. But machine learning presents the possibility to train software to follow humans' criteria for curation, with greater speed and accuracy than programmed directives. Instead of analyzing data pools, ML can "watch" human curators select, organize, and rank content. In effect, ML can learn to predict and implement humans' curation criteria, without needing to specifically tell the machine what that criteria entails. With enough observation time, the machine should be able to handle the bulk of the curation process, with a smaller number of human curators only needed to tweak the machine-generated curation, or to re-train the algorithm if the company's curation strategy changes.

Similarly, SaaS tools provide an increasingly large variety of data points to selectively analyze within your decision-making processes, such as re-targeting a marketing strategy or adjusting customer retention tactics. Some humans are particularly skilled at recognizing behavioral trends from seemingly disparate data pools of individual customer actions, and then developing innovative strategies to better respond to these trends. But ML excels at clustering data from individual points into a group that indicates a pattern. For example, automating data collection and analysis, particularly of unstructured data, frees up smart humans to spend more time considering how to act upon the findings of the automated process. So why not automate the trend-identification part of the process, so that humans can spend more time evaluating behavior and responding strategically? 

Maximize Time for Strategic Thinking

Training an algorithm on how to properly make decisions according to your approach can be a time-consuming process, so it's important to truly select tasks that will dramatically free up your time once they're handled by a machine. Depending on the uniqueness and complexity of the task you're trying to automate, it may also be possible to save time and effort by using a pre-existing ML model as part of your company's suite of cloud-based services provided by someone like Amazon or Google. Again, if there are concerns about the quality of automated results, try to figure out which parts of the process are suited for human intervention, both to adjust the machine-generated work and to use it for higher-level creative tasks. 

Technology doesn’t inherently replace people; it simply frees up their time to think strategically while the machine handles execution—rapidly, reliably, around the clock. Additionally, the process of articulating your "secret sauce" and teaching it to a machine can help conceptualize your unique approach and skill set, providing a clearer sense of your own true creative value. It can externalize, and therefore commodify, your own personal version of common sense or intuition.

You don't have to be a software architect or developer to do this, though it's certainly easier if you work with one. But quite often, innovations in scaling emerge organically from deeply informed, creative problem-solving, regardless of the problem-solver's technical capacity.

“Before you become too entranced with gorgeous gadgets and mesmerizing video displays, let me remind you that information is not knowledge, knowledge is not wisdom, and wisdom is not foresight. Each grows out of the other, and we need them all.” -Arthur C. Clarke

Key Points to Remember

The possibilities of AI can be overwhelming. Here's a quick list to keep it simple:

  • Identify the repetitive tasks in your workflows, and assess how it might be possible to automate them.
  • Examine how you train other people to do a task. Consider how this training process could be adapted to train an algorithm to handle part or all of the work.
  • Review current data collection methods to see if there are potential key metrics you're overlooking because the data is tricky for humans to collect or analyze.
  • Assess your own subjective decision-making processes. Document your personal criteria, and determine if these criteria can be taught to a machine, through an iterative process if necessary.
  • View machine learning as enabling and augmenting human capacity, not replacing it.