Skip to main content Skip to footer

October 16, 2023

Jumpstarting the gen AI journey: 5 common mistakes to avoid

To create a platform for future success, business leaders should apply lessons from past technologies.

It hasn’t even been a year since the introduction of ChatGPT, but we’ve arrived at an interesting and important junction in the generative AI journey. On one hand, we’re far enough along to know that gen AI will play a sizable role in the future (Bank of America estimates that by 2030, AI's overall economic impact could skyrocket to well over $15 trillion dollars); on the other, it’s still early enough for companies that have not yet begun to effectively catch up.

At the same time, it’s important to remember there are no shortcuts to success. Just as some organizations tried to expedite cloud migration plans at the onset of the pandemic, only to find themselves in a digital quagmire a few years later, companies that fail to craft a thoughtful generative AI plan may find that today’s investments and projects fail to build the foundation for long-term, sustainable growth.

With that in mind, here are five common mistakes companies may be making with their gen AI strategy—and ways to avoid those mistakes.

Mistake #1: Not getting started

Uncertainty in the regulatory landscape, coupled with the rapidly shifting nature of technology, are prompting some companies to take a wait-and-see approach to generative AI. That’s a mistake. Gen AI will change the way businesses of all kinds operate, even if we’re not yet sure exactly how.

In fact, it already has. In retail, for example, the technology is leaving its mark on the customer experience in the form of conversational commerce applications, hyper-personalized campaigns, and next-generation customer service tools. Indeed, in our September 2023 survey of senior business and technology decision makers at large businesses in the US and UK, 61% of executives expect generative AI to result in a complete business transformation.

The value of investments made in generative AI technology today is likely to compound over time. Every initiative a company launches now represents a chance to establish the partnerships, build the capabilities, develop the skills, identify the best practices, and create frameworks that will make it faster and easier to launch and scale programs in the future. This is also the time to test and learn, gaining insights not just from successful programs but also initiatives that fail, so that teams better understand the limits of the technology within the context of their organization.

Launching a small use case or proof of concept for gen AI isn’t nearly as expensive or complicated as it is for other digital technologies, such as traditional AI and machine learning or RPA. This is especially true if companies already have some of the underlying infrastructure set up, such as a relationship with a hyperscaler which would allow them to use out-of-the-box gen AI and large language model (LLM) capabilities. 

Bottom line: Gen AI is here to stay, and businesses must not only get started, but set themselves up to play the long game.

Mistake #2: Trying to go it alone

Gen AI is one of the most talked about topics of the past year, but very few people have any real expertise in the field. The growing demand for skills across industries coupled with a relatively small talent pool is making it difficult for businesses to build an internal team through strategic hires, so most will need to rely on a partnership model to help them develop capabilities and build maturity. In our survey, respondents named lack of skills as the top barrier to quickly adopting generative AI.

When selecting a generative AI partner, organizations should consider, first and foremost, the prospective partner’s technology capabilities and its investments in teams and tools. Beyond that, there are two elements that should be high on the list of considerations:

  • Industry experience. There is no substitute for sector experience. Companies will gain important advantages by working with a partner that understands the landscape and its challenges and can apply gen AI to the issues most relevant to the sector.

  • Enterprise IT experience. Niche gen AI players may know their own domain, but how well do they understand the complexities of the corporate IT function? The size and structure alone of a large client organization is likely to overwhelm a gen AI startup. Further, enterprises are grappling with other technology initiatives, such as cloud migration and optimization, that must progress concurrently with the gen AI roadmap. It will be immensely helpful to work with a partner that understands these complexities.

Mistake #3: Starting with advanced use cases

One of the factors holding companies back from experimenting with gen AI is the perceived risk of doing so. As was the case with the shift to cloud, some companies are cautious about using any new technology that has the potential to compromise their intellectual property or customer data. In our survey, security risks, privacy risks and reputational harm were the top three concerns of execs when considering the impact of generative AI on their organization.

But there are ways to start the gen AI journey that that are relatively low risk—namely by focusing on internal use cases. One of the most prominent low-risk, high-reward use cases for gen AI is around developer productivity. Using this technology to automate routine tasks (identifying and fixing vulnerabilities, refactoring code to improve quality, automating environment setup and configuration, and writing code documentation) can generate a substantial return without introducing much risk. Recent research puts efficiency gains for developers at 20% to 50%.

On the other hand, more advanced use cases, especially customer-facing ones, tend to bring greater risk. For example, using gen AI-enabled chatbots within the customer service function may offer incremental efficiency gains, but creating those tools is complex and requires companies to think through any and every interaction the customer may initiate.

This is not to say businesses shouldn’t use gen AI in the customer service function at all. But it’s wise to begin with an internal application, such as a tool to assist agents, as opposed to customer-facing bots. Generative AI makes an excellent knowledge management tool, helping agents quickly comb through past cases to find the optimal solution and next best action for common issues, like addressing product defects or managing returns.

By focusing on internal use cases, companies can begin to draw value from the technology while creating a launch pad for their broader gen AI strategy. They can build momentum and excitement among teams, develop the necessary infrastructure, and create frameworks that will enable them to advance as the technology matures.

Mistake #4: Customizing out-of-the-box tools

Organizations have long relied on customization to level-up out-of-the-box software capabilities, developing specialized use cases and adapting platforms to meet their needs.

But many enterprise IT leaders make a compelling case against system customization. When you customize, you add complexity. You make it difficult to scale. You make it more challenging to upgrade. In other words, the immediate gains of adapting often come at the cost of future agility. This is as true of gen AI systems as it is of ERP.

In these early stages, most gen AI tools on the market function in a very similar way and offer comparable capabilities. While these models may evolve over time and develop important differentiators along the way, for now they are largely interchangeable. That means companies need not spend significant resources on evaluation.

Moreover, these existing LLMs offer incredible out-of-the-box capabilities that can provide the foundation for many initial use cases. Given that the field is rapidly evolving, it’s best to focus on the fundamental capabilities of the tools and models without customization. This will put organizations in a better position to adapt over time.

That said, there are features that can be built around an LLM to make it more effective without resorting to customization. For example, IT teams can use prompt engineering as a communication technique to help the interaction stay on track. By offering specific conversational prompts, the agent (in this case a bot) guides the customer down a designated path to improve their chances of solving their issue.

Companies should also consider how they can use integration to make existing LLMs more enterprise oriented. At present, gen AI tools are available largely through a web interface, which is not a seamless part of the existing process flow. Through integration, businesses can increase flexibility within the workflow and draw more value from the technology.

Finally, companies should recognize that simply using gen AI is a way of building the maturity of the technology. The sooner an organization begins and the more it uses gen AI, the more precise and targeted the models will be. While customization may appear to be a value accelerator, it may in fact hinder the natural evolution of the models and negatively impact the organization’s ability to leverage the latest capabilities over the longer term.

Mistake #5: Getting stuck in a pilot cycle

Many organizations are in the process of launching pilots to prove the value of gen AI. And that’s absolutely the right thing to do, especially since the technology is still evolving.

At the same time, they need to remember that current investments and activity should help the company build maturity over time. Pilots are not just about proving the value of the technology or, in some cases, identifying applications that do not work. They also serve to build momentum and excitement for the program internally, establishing best practices, creating the necessary frameworks, and developing the required skills that will take the company to the next level.

While the focus today may be on experimentation, organizations will eventually need to develop a plan that extends beyond the pilot or proof-of-concept phase. In many ways, this process should mirror the introduction of any other new technology within the organization; businesses should create a path to production that allows them to develop capabilities and roll out the technology in a thoughtful, methodical way.

While it may be premature to create a comprehensive strategy or dedicated team today given how quickly the technology is evolving, this is something that will need to be done eventually. As such, companies need to be mindful as investments are made, teams are formed, and projects are launched.

What’s next: taking the right steps to get ahead

We’ve spent the better part of the last year living through a generative AI hype cycle. But as we approach the anniversary of the launch of ChatGPT, we may need to ask ourselves, at what point do we accept that this technology isn’t hype, so much as an undeniable part of our day-to-day lives?

Gen AI is here to stay. What that means for enterprises is that they need to play the long game, investing in generative AI and enabling technologies and developing capabilities so that they are at the forefront as the landscape continues to mature. The only way to do that is by acting now and launching their generative AI journey, while there is still time to get ahead.

To learn more, visit the Generative AI section of our website or contact us.

Scott TumSuden

Vice President & Global Managing Partner

Author Image of Scott TumSuden

Scott oversees strategy and growth for Cognizant's Retail division, leading relations with a top Fortune 30 client. A former Fortune 10 tech executive, he drives successful digital transformations in the industry.

Latest posts

Related posts

Subscribe for more and stay relevant

The Modern Business newsletter delivers monthly insights to help your business adapt, evolve, and respond—as if on intuition