I was recently talking to a friend who serves as CTO at a mid-sized company and was struck by his sudden shift in perspective on AI. Despite initial skepticism, he now believes artificial intelligence (AI) will revolutionize his industry. However, his main challenge has been convincing the rest of his executive team to adopt an AI roadmap. This scenario is not isolated.

In the last year, we have seen one contracted cycle of hypes about AI, which has led many leaders to question whether an investment in AI can really deliver proportional returns. These concerns are not without merit. VC firm Sequoia Capital recently evaluated The AI ​​industry spent $50 billion on Nvidia chips to train AI models last year, yet brought in just $3 billion in revenue.

Despite the disparity in investment, Sequoia went on to hypothesize AI is likely the “single greatest value creation opportunity” humanity has ever known, comparing its impact on business to that of the cloud transition. Unlike the cloud, however, which replaced software, AI has the potential to replace services, which the VC firm estimated has a total addressable market in the trillions. It's the reason tech giants like Microsoft and Amazon keep going double AI investment.

Connected: What is Artificial Intelligence (AI)? Here are the benefits, uses and more

With so many competing narratives about the future of AI, it's no wonder companies are confused about the best approach to integrate it into their organizations. The problem is that most leaders are still seeing AI in its limited capacity as a software or tool, rather than its ability to operate in a human-like capacity. Here are three common mistakes I see companies make when it comes to implementing an AI roadmap.

Underestimating and limiting the potential of AI

AI is widely seen as a tool or software, but because it can create and reason, has the ability to interact in a human-like capacity. Like a new employee who improves at his job with experience, AI has the ability to learn from his interactions and refine his methods to improve his output and get more work done overtime.

For this reason, leaders who think of using AI as “smart people” rather than software are better positioned to exploit its full potential. Think of a company's organizational chart. If you were to write down the skills and tasks associated with each employee, then you can begin to imagine where AI could be trained to augment or automate these tasks.

AI already outperforms humans in areas such as image classification, visual reasoning and even understanding English, according to Stanford University. Recently published artificial intelligence index report. As of 2023, the report showed that AI has surpassed human-level performance on some standard tasks, helping workers become more productive and produce better quality work. Another study from the University of Arkansas showed AI people passed on standardized tests of creative potential.

However, unlike humans, AI scales effortlessly as business demands grow, handling workloads without the physical and mental limits of humans. Adopting AI in this way means rethinking our team structures and workflows. It involves training teams to work alongside AI to improve their roles and drive innovation.

This shift in perspective is essential because it allows leaders, who may not be accustomed to using the technology themselves, to naturally understand how to best use AI throughout their organization.

2. Trying to mimic another company's AI use case

The more you start thinking about AI as smart people, the more you realize how individual each organization's approach to building an AI roadmap needs to be. I like to think of implementing AI as introducing new team members who need to adapt to the specific dynamics of your company.

Take human resources for example – a company might have 10 people there; only three more, even if they are the same size. This difference is not just about company size or revenue. It's about how these companies have evolved.

Every business has its own unique structure, culture and needs. To realize the full potential of generative AI, PwC reportedbusinesses should take advantage of its capacity to adapt to a company's specific needs and avoid the use case trap.

Of course, there are general use cases for AI, especially when it comes to enhancing customer service or sales. But when looking at a deeper integration of AI into a company's operations, the approach needs to be custom-built, not copied and pasted from external case studies.

Connected: I've tested the AI ​​tools so you don't have to. Here's what worked—and what didn't.

3. Buying off-the-shelf products — not tailoring AI solutions to your needs

There are some great AI products like ChatGPT, Dalle, and translation tools that solve specific problems within a company. The challenge with investing in a limited AI solution is that many leaders fail to see how AI can improve operations at a systemic level.

The true power of AI lies in its ability to fundamentally transform your operations, not just perform isolated tasks. PwC AI Predictions Report 2024 says many companies will find attractive ROI from generative AI. However, few will succeed in achieving transformative value from it – the biggest obstacle is the inability of leaders to think beyond established solutions and reimagine how they work with AI.

When building an AI roadmap, leaders must first make a thorough assessment of their company's processes. This means identifying areas of redundancy, identifying outsourced tasks that can be automated, and determining where the company invests heavily in human capital. By understanding these dynamics, leaders can tailor AI solutions to their company's needs and transform the way they operate.

The more I talk to company leaders about integrating AI into their businesses, the more apparent it becomes that we leaders need to change our perspective. When we see AI not just as a technological enhancement, but as the introduction of smart people, we are better able to integrate it into our internal operations, increasing performance and human intelligence along the way.