In the 1990s, the promise of artificial intelligence gained attention and fueled efforts to innovate across industries. I worked for a company that designed and built a variety of laser- and camera- based measurement sensors and systems. These optical sensors paired with image processing algorithms to locate and measure the size, spacing, area, and volume of small features critical to the assembly of increasingly miniaturized circuit boards.
It was humbling to learn how hard it could be to locate a circle accurately and reliably. It was never where it was supposed to be, no matter how accurate the robotics. Despite the best efforts of our optikers, variations in materials and manufacturing processes meant the circle rarely looked like one.
How AI Has Matured While Data Intelligence Lags Behind
As a software engineer, it was my job to figure out how to compensate for the shortcomings of technologies of the time. But no off-the-shelf solution could be integrated into a reliable and robust system. Many machine vision, AI, and speech and handwriting recognition companies with audacious beginnings were actually short-lived.
A friend told me about a neural network project meant to determine whether satellite imagery contained tanks. It failed miserably. The images that contained tanks tended to be taken on days that were clear and sunny. Those without tended to be from cloudier days. They effectively created a “sunny day detector” and a neural network with seasonal affective disorder.
Flash-forward to today.
Thanks to improved data intelligence, JetBlue announced a new facial recognition system as the basis for check-in at airports. Meanwhile, more of us everyday use speech and handwriting recognition systems backed by the cloud that are remarkably reliable and capable.
But hosted solutions for commercial-grade AI and machine learning services for enterprises now advertise that full functionality is just clicks away. “Just Add Data.”
This is a false promise — or at least not the full story.
Businesses Need Data Intelligence to Avoid AI’s Pitfalls
At TechWeek Chicago 2017, a panel of experts in the data intelligence field discussed the modern realities of data science, the people, and technologies that make it possible to create truly data-driven organizations. Now that powerful analytics are accessible for even the smallest organizations, there is hope that businesses will apply these technologies to their business operations, improving their abilities to predict trends and overall efficiency.
If you follow the money, there is reason to believe this is true.
How AI Can Bring the Worst of a Business to Light
Investors evaluating potential firms to underwrite them will assess their data science capabilities and culture for evidence of “an intent to be more efficient” as a company. If that culture and value system are not evident — if they lack data intelligence — then no solution on the planet will help. At that point it’s just a poorly-run business.
The phenomenon of “data lakes” attests to the pitfalls of pursuing AI. Companies realize that cutting edge tools depend on having large amounts of structured data. Acquiring that data is actually the hardest part. Most of it is usually buried in emails and spreadsheets.
Data Lakes Negate Data Intelligence
So, many companies invest in pulling data assets together into one place — usually Hadoop. They also must source the talent required to make sense of all of that data. But when the end drives the means, the results are usually lackluster:
- The organization of the database does not support the actual business processes
- The funding for the effort is largely tied to building a data lake
- Drained resources leave the organization underfunded and unable to demonstrate data intelligence or make proper use of new information
A CEO that wants a data-enabled business can fall into the trap of hiring an employee who spends four years building a data warehouse that few can understand or access. Real working solutions depend on interactions between domain experts and technologists that evolve systems in an Agile approximation of the iterations used in Scrum.
Organizations set themselves up for failure in AI when there is an abundance of data science hammers looking for nails. This is an environment unsuited for data intelligence.
Why Your People Will Determine Your Level of Data Intelligence
But collective recognition of this trap symbolizes a turning point where we can understand when and how to best apply these tools to get the results we need.
Success means recognizing a company and culture that is ready to exploit the tools that are available, rather than attempting to mimic successful models and expect the same results. Success requires architecting an organization and a system that work together. While this is much more challenging, it is also much more rewarding in the end.
If you have the right intent and culture for continuous improvement, then powerful and affordable tools exist for your cross-functional teams to leverage the predictive insights and operational visibility that your business needs. When you understand the maturity of your business model, you can direct these teams to give you actionable data and sustain data intelligence.
Being truly data-driven requires enacting the necessary changes that convert insights into a new product or process that increases ROI.
The technology is no longer the hard part — it’s the people.
Image Source: Unsplash, Simson Petrol