Artificial intelligence is expected to deliver somewhere between $10 to $15 trillion in value to businesses across industries over the next 10 years, based on estimates from McKinsey, PwC and many other experts in the field.
Smarter, fact-based decision-making is the catalyst that will unlock AI’s value. In functions such as the supply chain, organizations can save millions a year with AI-supported decisions that reduce working capital, optimize logistics and inventory, and improve demand forecasting.
Moreover, millions in new revenue await with AI that supports data-driven decisions to expand market reach, introduce new products and services, and drive customer satisfaction and repeat business.
Three key obstacles to realizing the value of AI
However, that value won’t be disbursed evenly across organizations. Some businesses will master AI technologies to realize billions in bottom-line value. Laggards may make game efforts, but McKinsey estimates they’ll be outperformed 10x by AI leaders. That’s because they’ll fall short in solving the three key obstacles that stand in the way of deploying AI at scale:
- Enterprise data is scattered, dirty, and poorly documented
- Data is of limited use on its own, without meaning and context
- AI alone doesn’t drive decisions, humans do (and they’re complicated)
They’re not new challenges, but nor have they been resolved despite substantial investments into data lakes and other cloud-based technologies. Many organizations have been left in a post-traumatic state, needing to find new ways to unlock the bottom-line value of AI.
Enterprise data is scattered, dirty, and poorly documented
If there is a constant in enterprise analytics, it’s that the data is always messy. It’s scattered across multiple systems, locations, owners, and functions. Even organizations that describe themselves as “data-driven”, and use analytics for a competitive edge, often don’t have a perfect handle on their data.
Even in the case that an organization employs a data-lake, the data remains scattered across multiple domains, and it’s often incomplete. For instance, the business may have moved sales data to the cloud (with some latency), but logistics information remains in an on-premise ERP, and weather data is accessible through an entirely separate API. Even when enterprise data is consolidated in a single location, it’s often “dirty” — incomplete, inconsistent, riddled with errors and a surefire way to make bad decisions.
Lastly, many enterprises lack formal data documentation — from simple column definitions to more complex metadata and data lineage structures. That makes it difficult for a data scientist to understand the data adequately and can cause significant issues with misinterpretations and errant conclusions
Traditionally, organizations have invested in ETL or integration tools to consolidate, harmonize, cleanse, and enrich data. This approach has turned out to be unwieldy, time-consuming, and prohibitively expensive to run across the entire enterprise. As a result, analysts and data scientists continue to spend the majority of their time cleaning and preparing data for analysis, one small slice of the enterprise at a time.
Data is of limited use on its own, without meaning or context
Data requires context to drive impact in an enterprise setting. By context, we mean not only knowing to identify what outcomes matter. Above and beyond picking the specific variable in the dataset itself, it is critical to understand which processes impact the result, and how the processes themselves function
Imagine a CPG company looking to improve its forecast accuracy. While it could try to optimize the forecast using advanced forecasting approaches (e.g., using XG Boost instead of a simple ARIMA model), that approach may completely miss that sales in CPG are highly dependent on controllable factors. Inputs such as promotions and marketing spend or constraints on production lead-times are likely to impact the sales forecast in a meaningful way
These factors may seem obvious to someone who has worked in CPG but are often overlooked when data is tackled in the absence of its business context. While many organizations have attempted to “brute-force” data science onto these sorts of business problems, the last few years have shown that approach is seldom successful. It’s clear that context and domain expertise is essential to fully leverage the power of data and AI.
AI alone doesn’t drive decisions, humans do (and they’re complicated)
At the end of the day, the vast majority of enterprise decisions are made by humans. This is where even the most successful analytics applications fall short — they fail to close the loop between the insights generated and the action that needs to be taken by the business.
One issue is that change management is often needed for teams and individuals to embrace analytics and adapt prevailing practices. People are complicated, biased by their experiences, and can often be an obstacle to implementing data-driven decision making. Then, even if a human accepts and agrees with the analytics output, that individual often needs to make changes in three or more different platforms to affect the desired business improvement – a significant barrier to adoption and speed.
It’s critical to reduce that friction between insight and action as business cycles accelerate, employee tenure continues to decrease, and the volume of decisions to be made increases exponentially.
Lessons from a leading pharmaceutical company
One approach that’s already proving impactful in solving those challenges is the deployment of AI-driven cognitive augmentation. A leading pharmaceutical company is unlocking over $100M a year in inventory while increasing critical service levels by leveraging technology that allows it to:
- Understand every aspect of the enterprise by using Google-like data crawlers to capture near real-time information from any number of operational systems thousands of times a day – enabling visibility at enterprise “scale and speed”. The data is consolidated, harmonized, cleansed, and contextualized – making it ready for use by humans and in AI applications.
- Recommend proactive decisions that improve the performance of the business, using a combination of business domain expertise and machine learning. Uncovering opportunities that are often hidden as a result of massive business complexity and lack of time
- Predict business outcomes by leveraging real-time data and artificial intelligence – driving confidence in the decisions the platform recommends
- Act and “close the loop” by empowering humans to review recommendations (rather than running tedious data pulls and analyses), make decisions, and execute actions directly through the platform. Avoiding the typical manual work that is conducted today across different systems.
The emergence of AI-driven Cognitive Automation
The increasing pace and complexity of global business have continued to drive these challenges over the past decade. While the emergence of cloud-based technology has given a glimmer of hope, large scale enterprises still find it challenging to deploy analytics at scale as a result of the above challenges. Even those that jumped aboard the AI wave early continue to struggle moving beyond the pilot purgatory.
While challenges to AI adoption are not easily overcome, it is clear that those capable of closing the gap will reap the rewards. That requires both a fundamental rethinking of processes and an embrace of innovative technologies purpose-built to help enterprises rapidly capture some of the trillions of dollars in value that cognitive automation is unlocking.
Aera Technology delivers the cognitive operating system that enables the Self-Driving Enterprise. Aera understands how businesses work; makes real-time recommendations; predicts outcomes; and acts autonomously.
Using proprietary data crawling, industry models, machine learning and artificial intelligence, Aera is revolutionizing how people relate to data and how organizations function. Headquartered in Mountain View, California, Aera services some of the world’s largest enterprises from its global offices located in San Francisco, Bucharest, Cluj-Napoca, Paris, Pune and Sydney. For more information about Aera, please visit www.aeratechnology.com.