Making AI Work For Your Company…Really

For the past year, I’ve been writing a new book called Implementing AI Systems: Transform Your Business in 6 Steps (by the way, the CEO and co-founder of Aisera, Muddu-Sudhakar, wrote the forward).  It’s been a lot of work but also a great learning experience.  I had a chance to talk to many companies about the AI journey–from startups to global powerhouses.

AI, autoML

Why write this book?  Well, for the most part, I often get questions about how to get started with AI.  This is even from those who have solid tech backgrounds.  AI remains a mystery for many people.

Another reason I wrote the book is that companies often fail with their own AI initiatives.  Just about every IT survey shows this.

Keep in mind that AI is not like traditional enterprise software, such as a CRM or ERP platform.  The technology is about probabilities, estimates and predictions, which can be murky and confusing.  AI also requires considerable effort with wrangling data, finding the right algorithms (which are often complicated), and putting in place a system for tracking and monitoring.

OK then, so what are some recommendations for boosting the odds for success?  First of all, there needs to be education across the organization.  This means learning about the core concepts and categories like Conversational AI, deep learning, machine learning, NLP (Natural Language Processing), and so on.  Then there should be an understanding of what AI can realistically accomplish.  Technology is far from a silver bullet.

Next, an organization must be clear-eyed about the build-versus-buy decision. The good news is that there are quite a few solid off-the-shelf systems that can make the process easier. Consider that your existing systems, like, already have powerful features. So are you using them?

You should.

But of course, there are some initiatives that need more customized approaches.  Having one-size-fits-all solutions will not differentiate a company from its competition.

Then what to do?  A good option is to look at AutoML platforms.  They have low-code/no-code functions that help to streamline the process for creating effective AI models.  An AutoML platform will also have the ability to track the performance of the model and provide alerts if something goes wrong.  Oh, and then there are built-in guardrails for compliance, which are absolutely essential since working with data can be fraught with risk.

Yet this not to imply that you can achieve complete automation (hey, as Elon Musk once tweeted, “Humans are underrated.”)  It’s critical to bring in the legal department into the mix.  The issues of privacy and ethics are extremely complicated and vary from country to country.  In Europe, for example, you must deal with GDPR.  As for the U.S., there is a patchwork of different privacy laws emerging.  But if there is no governance strategy in place from the start, then the AI efforts could easily be in jeopardy.

Something else to consider about AutoML platforms:  If you do not focus on solving a definable problem with AI, then there will inevitably be a failure.  To this end, it’s usually best to start in the back office – say with invoices or other repetitive processes.  These are often where you will find the low-hanging fruit for AI initiatives.  It’s easier to measure the ROI.  And yes, the early wins will generate momentum that will lead to more buy-in.

Note that AutoML platforms are still in the early stages of development.  They are generally best for BI-type applications, not customer-facing applications.  This is often known as operational AI, which is the most common.

If you want to take your AI projects to the next level – such as what you see with applications like Uber — there will be a need to hire a team of data scientists and to purchase state-of-the-art tools.  An organization will also need to have access to large amounts of quality real-time data.

Despite this, the experience with AutoML will certainly build AI muscle and help gain more buy-in.  It will also allow for a smoother transition to creating more ambitious projects.

The Bottom Line With AI

Even though there is much hype, AI is certainly real and constantly getting better, say with the theories, the new software systems, and hardware.  Companies like Instacart, Wish, and Lemonade are transforming massive industries because they are leveraging the power of this technology.  We are also in the nascent stages of this trend.

This means all companies need to be serious about this.  It is not something that can be put off.  But again, there needs to be a disciplined journey:  education, off-the-shelf solutions, AutoML and custom projects.  And along the way, an organization will increase its competitiveness at a rapid pace – and there will then be true digital transformation that will last for the long haul.