When are Autonomous Systems the best solution for your data-driven business challenge?

When are Autonomous Systems the best solution for your data-driven business challenge?

There are multiple paths to solving business challenges using data. As the saying goes “for a hammer every problem looks like a nail”. This is also true for applying advanced data-driven technology solutions to business process optimization or transformation. If you only consider one approach, you’re selling yourself short.

Fortunately, Neal Analytics has developed a broad range of AI, machine learning, data science, and optimization algorithms expertise over the last 10 years. Therefore, we can look at each new data-driven business challenge with the full suite of tools and technologies available and pick the right one(s) to solve it. In this post, we will share with you a simple yet powerful approach to help you find the right technology solution for your unique challenge.

To select the best possible approach, you can leverage this simple decision matrix. It will help you parse, at a high-level, which technology is the most appropriate for your need.

Narrow vs. Generic use case

First, you need to consider if your business use case is narrow or generic. This dimension plays a big role in the level of customization and the type of data you’ll need in your solution.

To gain clarity on your use case, ask yourself:

  • Is this a very specialized one-of-a-kind application or is this a broad one?
  • Are we trying to optimize for something unique and well defined, or is it more a generic problem with only inputs, outputs, and parameters that are unique to my process?

Well-defined vs. Open-ended challenge

The next point to analyze is whether you expect that your subject matter experts and engineers should be able to design the solution because the problem is well known, or whether it is an open-ended issue that requires a more research-oriented approach?

Another way to look at it is to think about the challenge being open-ended without an initial or proven hypothesis or, on the contrary, being well defined beforehand.

Decision matrix: Which data science technologies are best for your scenario?

 

Autonomous Systems and data science technologies decision matrix

 

Using this fairly simple two-dimensional taxonomy, you can select what is the most appropriate approach to your problem:

Broad or generic use cases and open-ended parameters

Open-ended and generic challenges will require the development of custom AI solutions built by data scientists using standard data science platforms such as Azure Machine Learning, Amazon SageMaker RL, Google DeepMind, or linear programming. Although these core platforms are existing ones, devising the right solution will require advanced competencies in the ML/AI/DS fields. These tools will be useful when lots of raw data is available but little is known about the underlying correlations among them as well as the underlying causalities for these correlations.

Narrow or specific use cases and open-ended parameters

More defined, narrow, but still open-ended challenges will require the development from scratch of custom AI or data science solutions. They could involve significant research before a workable solution is found. These are the hardest ones to solve and often require companies to leverage internal research divisions, academic partnerships, or external experts such as Neal Analytics.

Narrow or specific use cases and well-defined parameters

For challenges where both the application is narrow and the context is well defined, solutions will often already exist with hardware and software products dedicated to these specific use cases. Typically, these solutions will be based on standard control systems (PID controllers, PLCs, etc.) in manufacturing, or their hardware and software equivalent in other industries.

Broad or generic use cases and well-defined parameters

Autonomous Systems shine here. These solutions are the most appropriate when applied to systems with well-defined parameters (inputs, outputs, process steps, and control modes, etc.). By leveraging real-world heuristics devised by existing subject matter experts (operators, process engineers, etc.) and advanced platforms such as Microsoft Project Bonsai, customers can develop robust solutions that can optimize outputs even in changing environments and with competing goals.

What comes next?

Obviously, this matrix is just the first step in defining the most appropriate technical solution to your data-driven business transformation or optimization challenge. However, knowing where to start is crucial to help you define the best technology strategy. From timeline to internal and external stakeholders to the most appropriate deployment phases, knowing which technological strategy is the best can make the difference between project success or failure.

Want to discuss strategies and solutions for your data-driven business challenge? Contact us to connect with one of Neal’s experts. 

Additional resources:

This blog was originally published 1/15/2021 and has since been updated.