AI and ML for the Masses
Artificial intelligence is no longer the domain of Hollywood technothrillers, nor is it available only to the Fortune 500 or VC-backed startups. In fact, use of the technology has become increasingly common at companies of all sizes.
IBM describes artificial intelligence (AI) as technology that “leverages computers and machines to mimic the problem-solving and decision-making capabilities of the human mind.” Cutting-edge? You bet. But today, even small- and mid-sized companies can leverage AI by tapping into customer, product and market data to power their analytics, reduce their time-to-market and help get a leg up on their competition.
Data makes an application of AI like machine learning (ML) possible. Companies in healthcare, transportation, law, education, and even agriculture generate high volumes of it, and they’re benefiting from AI and ML. Let’s take a closer look at how AI and ML are already transforming how companies operate, whether they see themselves as “tech companies” or not.
What Are You Really Trying to Solve?
Before a company tries to navigate an AI or ML project, there needs to be an understanding of what problem it’s trying to solve — and there needs to be a lot of data. Then the company can look to find ways AI or ML can help it complete processes faster and more reliably.
With robotic process automation, one can, for example, leverage AI to identify patterns that currently require a human eye to recognize. Additionally, ML can be used to complete repetitive, redundant tasks while learning and recognizing patterns, and making inferences, just as a human would (but faster).
For example, one company had its field employees taking photos of documents. Then, the field employees would send the photos back to headquarters for other employees to route the images to different departments. This kind of classification problem is ripe for automation. Using a tool like Google’s AutoML, developers with limited experience can build customer ML models. One of the benefits of this particular implementation was that it freed up the team to focus on more innovative work.
Or, consider football — a decidedly nontechnical product on the field but one that produces a vast amount of data on factors such as player performance, collision intensity, run-pass preference and play similarity. Another company, a sports analytics firm, needed to improve the speed at which it could provide teams with in-depth analysis. Using ML, the company saw what had previously been a four-day analysis cycle drop to under two minutes, significantly improving its ability to validate game models against one another. Such information is highly valued by the teams that rely on these analytics.
In both cases, the problem was clear. There were processes that simply had to be faster and less error-prone to satisfy a business need. Understanding what you need to improve — and whether or not AI or ML can help you get there — is critical.
Invest in a Team That Can Keep Up with Your AI and ML Needs
To leverage AI and ML, companies need to build a team with expertise. Continuing to train and hone the model is an iterative process, so even non-tech companies should consider adding a few new hires in technical roles who understand the process. Hiring a technical consulting partner may also be a good way to ramp up a team.
Even among top companies, most leaders don’t understand how AI or ML models make decisions, so it’s incumbent on IT and analytics teams, who have a greater awareness, to help leaders understand the value and the difference the models can potentially make.
Investing in AI and ML initially, and for later oversight, will pay off in a significant way. The pandemic-era focus on AI has the technology poised for continued growth. Forbes reported earlier this year that “43% of enterprises say their AI and ML initiatives matter ‘more than we thought,’ with one in four saying AI and ML should have been their top priority sooner.”
The growing commitment to the technology and associated spend will mean a greater reliance on team members and partners who can wield it effectively.
Watch Out for Common Pitfalls
When making the transition to AI or ML, the easiest mistake is underestimating the amount of data necessary to train a new model properly. Most AI or ML projects start with zero knowledge. It can take thousands, millions and even trillions, of individual data points for an AI or ML model to become reliable at classification. When a company engages an external developer or hires a data scientist, the first request will always be for more data. As companies begin to look at all the data points it’s producing at every stage of customer interaction or product usage, it may need to add instrumentation to capture the actual volume (and value) of the data produced.
Another key pitfall to avoid is inputting faulty data during the training process. Mistakes in the training set become embedded in the model and can derail the entire project. Error-checking should also be top-of-mind for companies implementing AI or ML.
In other types of software, a developer can pull up the code and try to spot an error, but with AI and ML programming, the mistakes are much tougher to locate and extricate. Correcting mistakes is never ideal, but because the software continuously optimizes and learns patterns, the effect of any mistake becomes compounded. This is not unlike a team member who was trained incorrectly going on to train new hires in the same manner.
The speed, efficiency and accuracy of output is what makes AI and ML solutions so valuable for companies, but it’s important to carefully define the problem that needs to be solved and train the models correctly. Otherwise, your downstream results will fall short of what you need. Effective use of this technology requires a commitment to training, fine-tuning and retraining models to eliminate bias and bad data. The technology is still growing, but well-established approaches and tools allow companies of any size to leverage AI and ML today — shaving hours, days and even weeks off of processes companies and their customers depend on.
About the author: Calvin Hendryx-Parker is the co-founder and CTO of Six Feet Up, a software company that helps organizations build apps faster, innovate with AI, simplify Big Data, and leverage Cloud technology. In 2019, Calvin was named an AWS Community Hero. Additionally, he is the co-founder of IndyPy, the largest Python meetup in Indiana, and the founder of IndyAWS, Indiana’s fastest growing cloud meetup.