When you do an online search for a unique gift do you find yourself frustrated because you have to dream up ever more specific words to find what you want? A cup, blue, patterned, large, etc.? Why can’t you just find the product you like? Why do the cups all look alike? You have discovered the problem of domain specificity. The object(s) you want are very specific to your needs and desires. They are domain specific. But the objects you have been shown are very generic in nature. This generalized search is very good if you are looking for some standard product, but it is less helpful when your product is unique (a hand cured, organic leather chihuahua dog collar for example) or have very specific variations (gears for bicycles for example). This is why many businesses need a domain specific search and recommendation tool.
Why Not Use the Giants?
The tech giants like Google, IBM, Amazon and Microsoft have used their vast data gathering and engineering resources to pursue a general solution to all search problems. Sometimes described as “AI for all” or “general intelligence” they develop horizontally focused solutions. Often, they are looking for problems to solve. And they can be good for broad applications, such as “what does a tea mug look like?” but they lack industry specific features and nuances that understand your business pain points. Often you will start with these powerful tools but end up requiring significant adaptation and modification, either by your own in-house team or high-priced consultants over months of work. This is fine if you are a large multi-national with budgets and unlimited time perhaps, but it is not efficient and not particularly suited for most companies. So how do you get the specific tools for your business?
What Do I Want?
AI solutions are important for businesses of any scale and if you are going to use them for commercial applications you may want to leave aside the approaches and models best left to academic and research disciplines to focus on what works and what is specific to your domain. As a company owner you need very specific applications that have the ability to add value and take advantage of well understood AI dynamics.
Below we will suggest some approaches and requirements to consider when looking for a domain specific AI solution to your business needs.
1. Understand what you are getting.
It is important to reiterate that AI has two distinct phases: training and inference. Today, most AI revenue is made from training algorithms. That is to say, using vast amounts of data to improve an AI model’s accuracy and efficiency. As important as this is, as a business you want more than a model, you want a model that works for you. You want AI inference. This is the process of using a trained AI model to make predictions — predictions of what customers may want or what products should be shown. Companies that provide AI inference are still more rare than those that train models. But moving from an AI model to a production-ready AI application is what your business needs. This is where domain specificity shines. When you can train your model to predict the demand for your specific product features, then you are harnessing the power of AI.
Most AI technology is no where near the general intelligence of a movie robot. But there are real proven technologies such as conventional neural networks and LSTM models. Ask your AI provider what models they are using and ask them to explain it clearly. This will clue you in to the depth of specificity you can expect.
3. Focus on domain specific solutions that solve real problems.
As stated above, you do not want long customization times that require large sunk costs. Domain-specific platforms reduce implementation pain and associated lost productivity because they understand what you are trying to achieve.
4. What is your success?
If you define your goals it will make life easier for the AI provider as well as your business. For example, your goal may be to get an AI-powered incremental increase in per-order sales, or to optimize TEU loadings on vessels. The more specific the better.
Data is Vital
The data set your company owns is as unique as your business. This data is what differentiates you from your competitors. Any AI platform you use should be able to handle your domain-specific variables in order to deliver real value, quickly. But how can you help this process?
1. Your business needs some data to see the full value of an AI solution.
Most products will run adequately on small data sets, but with sufficient quantities of good, relevant data your solution will shine. Not all data is needed and remember that there will be a process to clean up your data to make it usable for the intended application.
2. Your problem has to benefit from AI.
Some problems are very complicated and require human knowledge. AI works really well to adapt to and optimizes but the data has to be relevant to the problem.
3. Your problem should be of the repeatable and conclusive kind.
If you can benchmark against results without AI then you know your problem is one AI can solve. Finding and understanding patterns in large amounts of data is very useful but it takes time to understand the real value of AI.
. . .
You have a great company, use your data to make it shine.
PwC predicts that the global impact of AI by 2030 will be over US$2 trillion. Much of this will be in saved labor productivity but much of these savings will be delivered to companies that are ready to utilize the tools specific to their needs and not general solutions. Don’t settle for general solutions to your specific problems, fully utilize your data.
If you would like a consultation on how your company can build predictive AI solutions, get in touch with our team today.