AI adoption has skyrocketed over the last 18 months. As a result of the pandemic, businesses have been pushed into digital transformations, including the transition from traditional data and documentation methods to AI workflow automation.
Sergio Suarez, Jr., Founder and CEO of TackleAI, recently guest-starred on the Digital Transformation Podcast hosted by Priscilla McKinney to discuss critical information regarding adopting meaningful AI solutions in the business sphere. Below you will find an informative and intriguing overview of their podcast episode.
Often the term “AI” is used as a catchall, and similarly, big data was a catchall for a bunch of information in a database. Today, there is more of an academic approach to artificial intelligence, where people want AI to completely learn on its own. Theoretical AI doesn't work for companies, it just isn’t practical.
It's really cool to use a tool and teach a bunch of information with millions and millions of data points hoping to get good answers, but the truth is, it still needs to be guided. The reason AI adoption has often failed is that people are not doing enough to make AI work for real-life business applications. Technology vendors currently build something and then try to pass the AI on to the customer for implementation. But the customer doesn’t want to try to train a platform for the next six months, spending millions of dollars. They looked to the technology vendor to provide something to immediately improve the efficiency of their work. They want to use it on day one - not start a long and arduous project of training AI. That can be scary and is a huge reason why it never gets to implementation.
Computer vision is a form of AI under deep learning models. So, if you have an iPhone, all of the pictures that you have on your phone are being scanned. Now, you can say things like, “Hey, find a picture of me in Hawaii where I'm eating food.” The AI is then going through and computer vision locating photos that align with this statement.
Now, we are to the point that AI can make connections between older photos — it’s picking up on age in photos and the algorithms, figuring out how we would have looked when we're younger or what we're gonna look like when we're older. That's an example of AI learning completely on its own.
"Analytics". If you noticed, you haven't heard the term "analytics" a lot lately. You used to hear it a whole bunch. This is because they're trying to refer to "analytics" as AI. But there's a big difference between data mining for insights and actually employing artificial intelligence. Automating big-data mining leads to people saying “I have all of this information, and this is what I'm gonna do because of the information.” We used to call it actionable intelligence, but that's a big difference. That still does not mean they are employing true AI to get this insight. It's unfortunate that people are now mixing those phrases. Nobody says "analytics" anymore cause AI is the new best thing. We need to really ask into this and require companies to explain how they are employing true AI.
Often you will hear terms like AI-enabled or AI-assisted. This basically means someone is going to make a tool that will help humans be able to do things faster, but in reality, the AI is not actually doing it.
An example could be an expense report. There's AI out there that will look at all the tables and deliver where it believes the important information is. So, instead of having to look at 2000 pages of documents provided, the AI will narrow it down to 14 pages where the information you're looking for could live. That's cool, but I think that's cool for five years ago. The hard part is all the manual work that still has to get done by a human. They will still have to read all 14 pages of the expense report the AI has pulled.
Now, a platform like TackleAI will look at that expense report and deliver specifics like, “How much was spent in this area?” Or “What did we do for X, Y, Z?” Our AI will go through and find those data points, pull them out and put them into your analytics platform so that you can go and find the insights that you're looking for.
Let’s take it a step further and talk about how receipts make it into your expense report.
Today, you can take a picture of your receipt to extract the information it'll end up somewhere, but the truth of the matter is, that usually like 5,000 people in another country where these things are being sent are just manually extracting the information. Some companies don't care that it isn’t AI, as long as it gets done in the next three, four, or 24 hours. But, for big companies that have thousands of those receipts that need to get done quickly - that’s just not an option for them.
So, when you want to digitally transform your company and you want to use AI—which can be seated on your server, at your place of business—you're outsourcing your data and privacy issues come into play. How comfortable are companies with sending their data to foreign countries? How comfortable are customers with the company outsourcing data overseas?
A prime example here is the way things are done in healthcare. Almost all of the healthcare industry is sending their data to be processed overseas. So, your private health information does not stay at your doctor's office. It makes its journey halfway across the world, through various servers, to be manually handled and available next time you're in the system. It would be much better if all of that was done here in the United States with the data secure in the company network. TackleAI is faster, more accurate, and 70% less expensive than human labor overseas. This AI solution will save businesses time and money, and ultimately provide them with peace of mind by eliminating privacy and security issues as well.
Robotic process automation is antiquated workflow automation. What happens in RPA is something called zoning. The AI will be introduced to the information on the document and, in the future, it will recognize that it has seen this document before. There's a slight intelligence in knowing what your receipts are going to look like, so it can go through and extract the information. For example, it is mostly true that a credit card receipt has the name and address of the merchant at the top. So, the AI learns that "zone" and can quickly return that information. Ultimately, the small amount of AI in use to process the documents will improve with repetition. But what happens when it is presented with a differently structured document? All of the zone learning is not useless and in fact, detrimental. Should one new document overwrite what the AI has learned about that zone in the past? In this way, RPA is far from the best artificial intelligence-structured data solution out there.
The end goal is the ability to process a piece of paper or a digital document the AI has never seen before. That is the whole concept of how TackleAI started.
Big data is a big problem. Instead of trying to make limited AI move faster, the whole premise of TackleAI is to make artificial intelligence smarter and more flexible - able to handle real-life business tasks. By adopting TackleAI, a business can fully access the critical information contained in unstructured and semi-structured documents, paper or digital. We don't create software that requires companies to change their processes in order to implement. We see what the business issue is and apply AI to solve it - letting companies continue to focus on the work they are doing. In that way, we create meaningful AI for enterprise business solutions.
We’re the next generation of AI, and leveraging our workflow automation will give your company a competitive edge.
Sound too good to be true? Seeing is believing.
Want to listen to the full podcast? Check it out now.