What is the most important element that will determine the success or failure of an AI/ML project?
Most people, including technical professionals in the field, would think it’s the datasets: Quality, quantity, and a data engineering pipeline to produce them. This is because machine learning algorithms perform only as good as the data used to train them.
However, business leaders are quickly realizing that the most important element of AI adoption is actually defining the business problem(s) correctly.
Therefore, domain knowledge of the specific business/industry is the most important thing to have before embarking on the AI adoption journey.
Here is the machine learning process in practice (courtesy of AWS ML):
In a previous blog post titled, “Lessons From Millionaire AI Entrepreneur” we learned that solving business problems is always far more important than the AI itself.
This may seem obvious, but it’s staggering how many data scientists and machine learning engineers focus on the AI algorithms themselves with little regard to true business value. This is part of the reason why last year 85% of AI/ML projects failed.
It’s pointless to ask questions about data, data infrastructure, and machine learning models before selecting one or more key business problems that can be solved using AI/ML.
Selecting the business problems correctly and framing them as machine learning problems are by far the two most important steps in the machine learning process (see image above).
Executing these two steps correctly requires domain knowledge of the business, specifically all the things that contribute to a given business problem/solution. Lacking enough domain knowledge can be dangerous when adopting AI.
Let me give you an example:
There is a Fortune 500 company that has digital marketers as part of its customer base. They provide these customers with marketing photos and other materials to help their conversion rates (for example, helping them increase the number of people that click on their Facebook ads).
This company decided to use artificial intelligence, specifically computer vision, to automatically provide users on their website with “high-converting” photos at the top of search results. Supposedly, these are photos that have been shown to convert customers at higher rates than others (based on visual characteristics).
Suppose a digital marketer intends to use these photos within Facebook ads. The conversion rate of any Facebook ad depends on 3 variables: Image, audience, and copy (ad headline and body text). The performance of a Facebook ad depends on the congruence of all 3 variables.
For a Facebook ad, if you use an image that “looks good” (such as random stock photos as in the case of this Fortune 500 company), but has nothing to do with your target audience, or it has nothing to do with your ad copy, then your Facebook ad will convert poorly (or not convert at all).
How can the computer vision algorithm possibly know this during an image search?
How could the people who labeled the images know this?
How could the machine learning engineers know this?
The only people who would have picked up these nuances are the ones who understand their customers deeply, including their digital marketing domain. This is where machine learning product managers and other subject matter experts must come into play.
This lack of domain knowledge leading to improper AI/ML solutions is extremely common.
We see this even within structured datasets where one or more columns are missing from the data. This missing data may never have been collected because engineers never realized its importance, and it may be critical to solve a given business problem with machine learning.
Make sure you invest time thoroughly understanding the business domain and the problems you are trying to solve with AI/ML. This can take several weeks, but it can easily save you months of time and other resources.
Always develop a clear AI adoption strategy for a given use case before going all in with technical implementations. AI adoption is only valuable because of its potential to solve business problems and add value to the organization.
If you need help to accelerate your company’s machine learning efforts, or if you need help getting started with enterprise AI adoption, send me a LinkedIn message or email me at email@example.com and I will be happy to help you.
Subscribe to this blog to get the latest tactics and strategies to thrive in this new era of AI and machine learning.
Subscribe to my YouTube channel for business AI video tutorials and technical hands-on tutorials.
Client case studies and testimonials: https://gradientgroup.ai/enterprise-case-studies/
Follow me on LinkedIn for more content: linkedin.com/in/CarlosLaraAI