Winning the AI Transformation - In five Baby Steps

7 min

Artificial intelligence (AI) is one of the most critical mega-trends in business in recent years. It streamlines transactions for higher productivity and automation of routine tasks such as production lines, banking, insurance companies, and so on.

Many organizations understand that they must embrace AI disruption within their primary processes — and that it is sometimes crucial to the survival of a business.

It’s either you succeed in it, or you may lose your competitive edge.

Undoubtedly there is a hype about AI; hence, every CEO / CIO / CDO wants to drive AI transformation, but many organizations fail to do it right.

A few months ago, I had a conversation with an insurance company C-level executive. We had a meeting after the board of the company he works for made a strategic decision: They are investing in technology, and they would like to dramatically improve their efficiency. He told me they currently have a few million clients and are putting an emphasis on quickly growing their digital marketing efforts. But behind the scenes, they still function like a traditional company, with some major labor-intensive bottlenecks in the claims and underwriting processes.

We discussed the right way to adopt AI in the company. My recommendation, which is summarized in this post, was to do it in baby steps. Implementing AI is a complex process, AI problems are inherently different from company to company, and thus most “off-the-shelf” products are not suitable. On the other hand, building an internal data science team and developing solutions from scratch are processes that have an extremely high failure rate.

A recent Accenture survey shows that:

84% of C-suite executives believe they must leverage artificial intelligence (AI) to achieve their growth objectives, yet 76% report they struggle with how to scale.

So, how can you successfully scale AI Transformation? In this article, I would like to suggest a 5-steps guideline for AI transformation in your organization. How to do it right, avoid failures, and leverage the potential of this technology:

5-steps guideline for AI transformation

I will describe our personal experience here at BeyondMinds, demystifying some real-life use-cases.


The transformation of Artificial Intelligence must be incremental for organizations. I would suggest five steps:

Step #1 — understand the value AI technologies can bring to your organization

There are no shortcuts when it comes to understanding the value of AI in transforming your organization. First, assess how the benefits of this technology will meet your organization’s strategy. One way of doing this is by choosing the most frequent processes and analyzing ways to reduce costs. Another way is to look for a competitive advantage and grow your revenue through automation.

In both cases, analyzing a path to a return on your investment is mandatory because there has to be some investment in implementing AI technologies that will eventually create a more profitable company. Existing AI use cases from similar industries is an excellent starting point.

At some point in this phase you should assess your data assets. What data does your organization collect from different channels? What data (structured and unstructured) is processed and stored within your information system? What is the analog (not digitized) data that is involved? We strongly recommend that data management and data architectures be systematically handled before driving the organization’s AI transformation program.

The outcome of this step is an extensive list of AI capabilities that can help your organization to better fulfill its strategy.

Take, for example, manufacturing companies. A brief look at these companies’ strategies will unveil the importance of enlarging the capacity of production lines whilst increasing product quality. AI capabilities can automate the manufacturing processes and create better results in many KPIs.

Source: Alex Vilar on Unsplash

Step #2 — prioritize quick wins first

After planning the long-term roadmap and considering the company strategy, it is time to choose the right solutions, at the get go.

Low costs and high benefits are preferred (aka quick wins). But, when it comes to AI, a promised return on your investment is not enough. Mitigating risks should also be taken into consideration.

The primary obstacle when getting started is the skill set needed to accomplish AI development. According to VentureBeat, 87% of AI projects never make it to the production stage.

The best way to start is to purchase a customized solution from an experienced provider that will lead your organization along the path to success; you need a quick win under your belt that will generate the right trust and engagement for the transformation ahead.

The outcome of this step is 2–3 quick wins implemented with high ROI and low risk.

For example, suppose your company aims towards having a low recall number within your manufacturing process. Every time your customer returns the product, it costs  you— a lot. You can assess precisely how much: manual labor, loss of credibility and customer retention. It probably also hurts your brand (but this one is harder to assess). AI-based defect detection can create immediate value for your company. By integrating an already proven model that is tailored for your product lines and defects, the value generated by implementing AI is clear and becomes a “quick win” to communicate to management.

Source: Photo by Franck V. on Unsplash

Step #3 — a divide-and-conquer approach for each use case

At this point, you have a proven ROI solution on your hands. Yet be mindful that AI can be a double-edged sword. Specifically, the deployment phase should be treated with great care.

You can’t just fast forward to the end of deployment and full automation. Plan each process as a multi-layered part of the equation and solve each part separately. This way, you “divide and conquer” the mission. 

Here’s my recommendation of how to divide-and-conquer AI deployment:

After completing the Proof of Concept, the AI model outcomes should be checked on real-life data. This is a tricky step since it takes time to trust the AI model. The best way to do so is to test the model’s outcome in parallel to the non-automated process. After approving the model, it is time to integrate it inherently to the process and automate the process as possible.

The outcome of this step is an incremental plan of deployment and integration for the new AI use case (the quick win) to your organization’s process.

In a defect detection use case, you start with one defect in a specific product. Then you separately detect different defects in the same product; only then can you run it all together. The testing of the defect detection model outcome should be performed in parallel to human inspection, and after a short while, the model can replace the human in the loop.

After accomplishing this task, you can approach a different production line (again starting from the beginning of the axes for this new “layer”).

And finally, using AI-based models to detect defects can eventually help you find anomalies within product lines, without being defined explicitly by the machine. This would be the “holy grail” of this multi-layered solution.

Source: BeyondMinds™ Defect Detection on steel

Step #4 — share your success

Congratulations! The first AI success for an organization is something to celebrate. And by that, I mean to communicate the successful use case on any possible occasion and medium: forums, conventions, company’s website, board meetings, etc.

AI storytelling has become a new emerging art. In the beginning there are barriers to deal with while using AI. Explain how the AI technology works “under the hood” in the successful use case, show the results and their explainability, but more importantly why it creates clear value for the organization!

While communicating what has been done, it is important to emphasize the endless potential AI solutions can bring ahead.

The outcome of this step directly feeds the next step of creating a long-term roadmap together with the enthusiastic support of management.

Step #5 — long-term plans

Now is the right time to create a long-term roadmap (3–5 years) that includes resources: a budget, data lake infrastructures, human resources, upskilling staff, etc. It’s recommended to establish an AI transformation task force that controls the progress of the roadmap.

Source: Summary of AI Transformation steps, by Nurit Cohen Inger

The outcome of this step is a roadmap for AI transformation  controlled by C-level management.

Now, all that is left to do is lead your plan to success!

Takeaways

AI transformation roadmaps are easier said than done.

Source: Jaromír Kavan on Unsplash

It’s one thing to plan your path to deliver value through AI, and it’s another thing to succeed in bringing it to fruition. Taking baby steps with your roadmap as well as choosing the right use cases to focus on first, and customizing them for your exact needs — are key to successfully implementing your AI journey. You will work your way up through the chain of success, and the endless potential of AI will be revealed.

Need to solve core business problems with customized AI solutions?
See how you can solve individual use cases or achieve a company-wide AI transformation using one platform and gain a competitive advantage.
Request a Demo