How much has your business really changed since implementing AI? I’m not asking about lift for discrete metrics such as customer loyalty, campaign revenue, or even process optimization. Was the DNA of your enterprise altered by tweaking the edges? Of course not, because the elasticity quotient was not met.

The elasticity quotient links the KPIs of the business with economic and market behaviors, letting enterprises expand, contract, and adapt at will for seen and unforeseen events.

Forrester’s analysis of more than 100 case studies provided by software vendors and service providers shows a disconnect. AI leads to measurable improvement in process outcomes, but less than 10% of firms demonstrated AI implementations affecting overall business revenue, profitability, and shareholder value. In one case, the board of directors of a transportation company asked why they should be investing more in AI after three years with no significant overall return.

Machine Learning Isn’t Learning

The underlying issue for AI is both in the algorithm (machine learning) and business model. Business models are still linear, driven toward efficiency, and hardened. Machine-learning models are deployed for discrete outcomes of business processes, increasing algorithm use. That is a positive outcome. However, big market changes break machine learning, accelerating dives in process outcomes.

  • Buyers stop buying or shift completely online during a pandemic. COVID-19 showed that algorithms are trained on typical buying behaviors and not ready to handle purchases outside the norm. While a global health crisis shows the extreme in thwarting machine-learning intelligence, it is a window into vulnerabilities where industry and startup disruption can emerge. It also exemplifies how the effectiveness of algorithms for identifying cross-sell and upsell opportunities is capped.
  • Promotional blocking confounds marketing influence. Jessica Liu, senior analyst and expert in social marketing, confirmed my suspicions that I broke Twitter’s ad algorithm. Promotions in my personal Twitter feed were based on my work account — too much technology, data, and analytics. So I blocked the promoters. Twitter algorithms kicked into high gear, with more and increasingly diverse technology, data, and analytics ads in my feed after I blocked the promoters. The relevance of ads in my feed has decreased, negatively impacting my customer experience.

The elasticity quotient overcomes these hurdles by acting as the economic indicator for the C-suite. Here’s what you need to know.

Use AI For What It Is Good At — Stop Seeing It As Analytics

AI is an ecosystem of models. That is the key to unlocking its potential. Enterprises must augment today’s automation models with holistic, strategic models to achieve AI for resilience and competitive advantage. The model is a simulator for near-, mid-, and long-term strategy, planning, and execution. Thus, AI needs training on multiple objectives. This is where AI shines: the analysis of relationships and formulating links. It is also good at breaking and retraining on new links. To overcome machine-learning rigidness, AI needs training and execution of a broad set of enterprise metrics and KPIs, augmented by data outside automation.

This approach allows for top-down, bottom-up, and cross-channel knowledge sharing. That is where elasticity provides the proactive data and dials that the C-suite, line-of-business leaders, and operational teams use to balance and rebalance the ecosystem. Additionally, AI overcomes even hardened automated processes as intelligence is expanded and connected.

Automation Investments Are Not Lost — They Are A Foundational Step To AI

This is not a call to scrap intelligent automation. The elasticity quotient is the next step to move from machine learning toward AI. Linking machine-learning models inside and across channels ensures that operations can work asynchronously while leaders have synchronous oversight and influence. This can be complex. To overcome it, strong attention to best practices within and across the AI chain in model ops, data ops, and development ops is mandatory.