Three Enterprise AI Mistakes To Avoid by Forbes – Entrepreneurs

Serebral360° found a great read by Forbes – Entrepreneurs article, “Three Enterprise AI Mistakes To Avoid.”

Add another layer to your #Business literacy. We at Serebral360° would love to know if the Forbes – Entrepreneurs article was helpful, leave a comment, like and share. Let’s dive in and discuss the information and put it to use to grow your business. #BusinessStrategy #ContentMarketing #WebDevelopment #BrandStrategy
Info@serebral360.com 762.333.1807 www.serebral360.com
Grap a copy of our NEW Business Stratgety Books #FFSS VOL1 and #FFSS VOL2

More and more business leaders are looking to artificial intelligence (AI) to further their objectives, from building new lines of business to protecting existing businesses. When done right, the benefits of a successful AI project likely result in significant returns for the business. However, new entrants would be wise to avoid these three common pitfalls, which I have observed in my work with a variety of AI companies.

Pitfall One: Misunderstanding The Capabilities Of AI

AI’s capabilities are commonly misunderstood. This is partially due to media hype and partially due to popular science fiction. In its current state, AI can tackle highly specific problems with unprecedented efficiency. Picking a person out of a crowd, identifying a supply chain inefficiency and recognizing verbal commands are all areas where AI shines.

The current state of the art has trouble when asked to be “task flexible.” What this means is that the same piece of software does not necessarily do well when asked to transfer its experiences. For example, fixing supply chains in the shipping industry and then trying the same in the aviation industry.

This can be hard to grasp. After all, a trained supply chain professional might succeed at these similar tasks. The modeling technique selected for the shipping problem might be state of the art. What gives?

Data. More specifically, the distributions and qualities of the different data. The algorithm may have learned that a specific pattern, unique to a dataset, is indicative of a certain behavior. The second problem may look similar on the surface but have strong differences when drilling down. Perhaps the distribution of target events is uniform rather than Gaussian. Maybe the data is reported too infrequently for the algorithmic method to be useful. This particular dataset could have more unpredictable “one-off” occurrences. Whatever it may be, it means more investment from your business to develop a working solution.

Pitfall Two: Underestimating The Cost Of Data

I recently received a question from a colleague about a piece of software Google had open-sourced. It went, “It’s wonderful they have released this new prediction framework. When will they release the data behind it?”

My answer was: “Never. I think it is unlikely they do.” This was met with confusion. That is when I realized it is not commonly understood that the value lies in the data and the system, not the model. One caveat is that cutting-edge modeling research can make improvements on existing results. This exists outside the scope of most projects. The vast majority apply existing frameworks to new business problems. Therefore, modeling costs are mostly time and materials.

Meanwhile, data requires significant overhead. Raw data feeds often have subscription fees. Sensor data requires maintenance and processing. Large amounts of data incur storage costs. Bad data must be thrown out. All data must be cleaned and formatted.

Additionally, different team members will access data in different ways. This must be accounted for. Data scientists may be comfortable writing SQL but uncomfortable outside of a GUI. Your management team wants human-readable reports. Your engineers may prefer RAPIDS to PySpark. Ultimately, these data-related costs accrue, likely making data the most expensive, and valuable, part of your project. Be sure to allocate your budget accordingly.

Pitfall Three: Mismanaging ROI Timeline And Expectations

The last pitfall is developing an appropriate return on investment (ROI) timeline for your project. Executives often have trouble appropriately benchmarking their AI projects because the technology is new and there are limited case studies.

Understanding how value is created in an AI system is key to a rigorous framework for assessing ROI. When starting a new AI project, I always seek to build a compounding dataset. This means I expect the product to have a good initial value. Much more importantly, I expect the product’s value to compound dramatically as it gets smarter and smarter.

Therefore, I recommend setting two ROI goals. First, justifying the project. This is a conservative line of sight return on investment. The second, an upshot ROI, projects capturing a portion of the compounding value of the project. This method sets a high floor and leaves room for radical upside.

An executive who understands what AI is good at, who accurately calculates the cost of data and manages ROI expectations, is far more likely to succeed. Therefore, these lessons can serve as valuable aids to your success.

June 11, 2019 at 08:01AM
https://www.forbes.com/sites/theyec/2019/06/11/three-enterprise-ai-mistakes-to-avoid/
Forbes – Entrepreneurs
http://www.forbes.com/entrepreneurs/
http://bit.ly/2CMy7Yu