Catching the AI wave in the enterprise

ai in enterprise

AI (artificial intelligence) is creating much excitement as an emerging technology with applications that could be genuinely life-changing. However, there are challenges to applying it, so it’s important businesses are aware of these implications and how to address them.

AI is emerging as a must-have technology even though most business leaders find it difficult to truly understand what it really is or how to implement it. Businesses looking to profit from AI should think less about the technologies they buy today than the foundation they’ve already built up till now. It requires strong fundamentals, which includes a strong data science foundation with advanced analytics capabilities.

Businesses that don’t have those basics in place need to move now to catch up, or risk missing the AI wave. However, it’s not true that competitors have figured out how to turn it into a silver bullet, so the most important thing is to build those data fundamentals and then be agile enough to change.

Rather than focus on technology, businesses need to invest in an architecture that delivers flexibility and ensure their decisions are all data-driven. This will position them to leverage it more successfully. Three key ways have been identified as to how enterprises should view AI:

1. Consider domains where AI has been proven
Most breakthroughs in this technology are in areas that humans are good at. This is logical because deep learning networks are inspired by the human brain, and there is extensive publicly-available research, code, and pre-trained models in these areas.

Applying AI outside of those systems is likely to be more problematic. Use cases such as fraud detection or preventative maintenance models, often touted as ideal for the medium, show more incremental progress because there is less ready-made material available.

This means businesses should focus their resources on areas where quick wins are possible, such as solving problems that involve vision, language, or robotics control, where there is an enormous body of research and experience available. For other domains, enterprises should expect progress to be slower.

2. AI isn’t magic
AI isn’t a magical or blanket solution for organisations. It’s better viewed as an extension of traditional analytics and machine learning. The best areas to apply it are those where there is available research and where the business has already been applying machine learning. It requires massive data sets for success, and the data must be clean and integrated, making it fairly complex to apply across the board.

3. Multiple experiments will be necessary
Organisations should start by applying AI to a large number of problems to see which ones produce the best results. This prevents time-wasting on issues where it isn’t the right approach and delivers quick wins.

Jumping in the deep end with AI means many companies will soon find themselves in over their heads. Deploying, monitoring, versioning, and tracking the performance of models is complicated and, if companies don’t have experience, the process won’t be seamless. It’s better to smooth out the transition by taking a deliberate approach where AI is applied gradually to a number of use cases.

Alec Gardner, Director – Global Services and Strategy, Think Big Analytics