Incorporating AI into a business strategy is something, for most companies, that has only been in discussions for the past 1-2 years. The sobering truth is that most companies don’t know what they need. Some bring in consultants who often have a narrowed approach, that typically fails to address a fully connected model. In this post we’ll review some things to look out for when building your A.I. strategy.
The Things We Know:
The one truth we can apply to most businesses is that the A.I. approach will, at some point, involve the mountains of data they’ve accumulated over the years. This data is usually inaccessible except through time consuming and inaccurate manual processes to extract the important information. Often, the documents are lost or misplaced from a lack of naming and categorization standards. If that went a mile over your head, think about how files are misnamed and how that causes you to look right over them when searching for data. A database will do the same thing. This is the most common bottleneck to effectively incorporating a machine learning strategy.
Next, and equally as important, there isn’t a “one algorithm to rule them all” for businesses. I’ve heard some outlandish things like a “Director of Analytics” saying they’ve solved a dynamic classification problem by using basic geometry. It took most of our resident mathematician’s resolve and quick “Don’t do it” glance from his CEO, to not describe how absolutely ridiculous that statement was…but if you encounter a pitch from someone saying they have resolved all of your data issues before even walking in the door, they are selling a bill of goods. Either that, or be fearful of the Terminator scenario because these guys have manifested the singularity! Selfishly, it’s great meeting these folks because they’ll inevitably fail and NthDS can show up with a superhero cape and feed our saviour complex…but this information will hopefully open some eyes to the charlatans…and we all need to do our part to remove the A.I. charlatans out of all sectors.
How things should be:
It all starts with figuring out what you have before you figure out what to do with it. For example, if you have legacy data that requires data entry, you need to assess how much you have and where its all located. Avoid taking the fragmented approach, but look at what you need from the whole of your data and work from the start toward a complete solution. The solution should be focused on finding the relevant documents, categorizing and renaming the files, and then sorting them for easy access.
Once data aggregation is achieved, the focus should shift to a digital/digitizing strategy. This is where the data needed from those sorted files can be identified and digitized into a database. your team of data scientists or external consultant should be interfacing with your data managers and IT responsible for the data. The collaboration will inform the data scientists on basic understanding of the data and its downstream role in your predictive analytics. That entire process should be centered around identifying the needed data from those sorted files so it can be extracted and digitized it into a searchable database file.
The analytics strategy needs to be carefully incorporated. Your data scientists should be focusing on what are classification vs regression problems, then understanding what paradigm to use to get to the end state. All of this needs buy in from the business as they have the experience where the devil resides…in the details.
All of this also involves the human factor of getting the tech nerds up to speed with the expectations of the business, and the business guys trusting their nerds to execute the plan, regardless of whether or not they’ve seen it done that way before.
I’d be remiss if I didn’t point out that NthDS has developed products to solve these problems. Our Crawler is able to find, sort, and rename all your relevant data. Then extract and digitize the metadata using our other product Nspect. Contact us today to find out how our A.I. solutions can be applied to your company.