As we are long past the stage of AI hype, it’s becoming apparent that the technology’s biggest issues revolve around gaining profits rather than figuring out how to make it useful. With the growing number of AI experts and machine learning services, AI is capable of providing immense value for many organizations. However, when it comes to deploying AI, companies often fail to even cover their initial investments. This seems a bit contradictory, isn’t it?
A recent IBM research reveals that only 21% of companies are able to integrate AI into their operations. This is where the root cause of the problem lies: it’s impossible to achieve economic returns on the technology that hasn’t been put into production. Moreover, even those AI projects that get deployed often don’t bring the expected value.
Let’s discuss the hurdles companies face on the way to AI profit-making and how they can be overcome.
Given that AI is always data-heavy, it’s paramount that the adopting organization’s culture is data-driven. Unsurprisingly, a lack of data culture is one of the most recurring problems that companies have to face on the way to realizing the full potential of AI.
If the company’s leaders and key employees have poor data expertise, AI initiatives will most likely fail. Even expertly built AI systems won’t realize their full potential if the staff doesn’t apply data-driven approaches to decision-making. A lack of change management is another widespread mistake in AI implementation.
More often than not, AI calls for significant changes in organizational structure and strategy as well as employees’ mindsets and skills. Therefore, consider change management as a core part of the AI implementation roadmap and ensure that your company’s leaders have the necessary knowledge and drive to foster the AI-centric culture.
While goals are basic success prerequisites for any project, when it comes to AI implementation, many companies still fail to clearly determine them. It’s essential to have clear expectations about the outcomes of an AI initiative. More often than not, end users don’t participate actively in AI projects, so when the technical team builds flawless AI systems, they provide little business value. This is why it’s critical to involve all the stakeholders from the beginning of the project.
Also, AI projects often bring value that cannot be measured. For example, enhanced employee satisfaction or better customer experience is much harder to keep track of than cost or time savings. Or, let’s say you build an AI system to decrease the time it takes for the IT department to categorize tickets. First, given that the system will have to make sense of free-form text using NLP, it won’t be 100% accurate, especially in the beginning. So your team will need to determine the permissible error rate and account for that in the ROI calculation.
Here is another example — let’s say there is a critical issue which needs immediate attention of IT staff and an AI system mistakenly identifies this ticket as low-priority. This significantly complicates ROI calculation as it’s hard to measure the negative outcomes of such a case.
This is why it’s critical to start with projects where ROI expectations can be properly calculated. For example, many manufacturing companies succeed in achieving economic returns on AI initiatives applied for quality control, as their ROI is comparatively easy to measure.
While it’s tempting to build large-scale AI systems, aiming for low-hanging fruit is often a much more effective strategy, especially in the beginning. It might be a good idea to start with robotic process automation (RPA), which tends to be more affordable than AI and provides relatively fast ROI. RPA implementation is non-invasive, meaning that it doesn’t disrupt the flow of legacy systems like many AI solutions would do.
AI projects that turn out to be quick wins can also help to justify more ambitious AI investments and ensure stakeholder buy-in in the future.
While it may sound trivial, companies that are more mature and experienced have a better shot at reaping the benefits from AI. Such companies tend to have established data governance practices, elaborate training programs, performance tracking systems, and clear project goals. These are critical differences between companies that succeed in AI implementation and those that don’t.
Given the volatility of project success rates, AI calls for a solid foundation in key management areas more than any other technology. The degree to which companies can track, measure, and organize processes often correlates to their probability of profiting from AI.