According to a survey conducted by McKinsey, 65% of organizations are using generative AI which is almost double the number reported by McKinsey in its previous survey 10 months ago. Organizations are in different stages of generative AI lifecycle.
Some early adopters have successfully scaled their generative AI pilots into production while others are still struggling with pilots.Some businesses are looking at third parties for help while others are creating and training their models in house.
If your business is struggling with generative AI implementation then this article is for you. In this article, you will learn about six hard truths about implementing generative AI in enterprise.
6 Hard Truths You Should Know Before Implementing Generative AI In Enterprises
Here are six hard truths you wish you knew before implementing generative AI in enterprise.
Tech Talent Hampers Generative AI Adoption
According to a Deloitte report, 33% of organizations who have a high level of artificial intelligence expertise think positively about generative AI. Despite this, they still consider generative AI as a threat to their business model. They are just adopting the technology due to the added pressure.
This clearly indicates that organizations that specialize in generative AI are preparing themselves for tough times ahead. Businesses are rethinking talent strategies and upskilling. They are investing heavily in training their existing workforce in artificial intelligence so they can make the most of the technology.
Generative AI Impact On Bottom Line Is Not Clear
Generative AI is facing some roadblocks as it moves beyond the initial excitement phase. Aamer Baig, a senior partner at McKinsey, noted that only 15% of companies surveyed by McKinsey have clear visibility into earning improvements from generative AI initiatives, emphasizing that not all use cases add value. He advised organizations to focus on projects that address real business problems, are technologically feasible and carry minimal risk.
Despite early adoption, many organizations are realizing the limitations of generative AI. For instance, CNH Industrial’s experience with Microsoft’s Copilot and developing their own large language models revealed that while there have been some productivity gains, they were less significant than expected, particularly in areas like invoice processing. Generative AI is currently seen more as an information synthesizing tool, with off-the-shelf solutions falling short in areas like financial analysis.
The proliferation of generative AI tools and technologies has become an obstacle to scaling their use effectively, as noted by Baig. David Higginson, from Phoenix Children’s Hospital, echoed this sentiment, pointing out that most generative AI advancements are still being driven by large tech companies, leaving smaller organizations in a waiting phase for mature, impactful solutions. As effective tools emerge, healthcare providers will face difficult decisions balancing risk, cost and outcomes.
Legal Questions Needs To Be Answered
The legal and regulatory challenges coupled with higher operational costs can put off many businesses from generative AI adoption. According to a study conducted by Deloitte, 28% of respondents consider compliance as a barrier to generative AI adoption while 27% consider governance as a major hurdle in generative AI adoption. 42% of businesses are confident that they have done enough to govern generative AI and mitigate the risk associated with the technology.
Jim Rowan, Principal at Deloitte Consulting shed light on the problem by saying, “This shows a lot of uncertainty in terms of how artificial intelligence will be regulated over the coming year, especially for global organizations operating in multiple regions. Looking at the bigger picture, the challenges that generative AI poses in corporate governance and risk parallel those in society governance and risk.”
More than half of the Deloitte study participants showed concerns over wider adoption of generative AI and said that it can increase economic inequality as well as centralize global economic powers.
Cost Control Is a Big Challenge
The high costs and challenges of implementing generative AI are significant concerns for organizations. According to McKinsey’s Baig, companies must proactively manage these costs, which are driven by the compute intensity and the extensive changes in workflows, business processes and key performance indicators that generative AI requires. They should Budget Vps server instead of expensive hardware.
He emphasized the need to invest in risk management, hallucination training and ongoing maintenance, similar to the investments made in digital transformation efforts. Higginson also highlighted the barriers to generative AI adoption, noting the scarcity and high cost of the hardware such as dedicated hosting server , data centers and graphical processing units, power, and data required to train models. This scarcity forces organizations to prioritize solutions that can appeal to a broad audience and generate long-term revenue.
Despite these challenges, some organizations like CNH Industrial and Briggs & Stratton have begun implementing generative AI platforms but face hurdles in achieving cost-efficiency and user adoption. CNH’s Kermisch noted that while there was initial excitement about using artificial intelligence tools like Copilot, utilization often drops off quickly as employees struggle to use the tools effectively.
Similarly, Briggs & Stratton’s Olsson pointed out the difficulty in monitoring adoption and determining the value of their investment in tools like Google Gemini. He stressed the importance of staying nimble in the rapidly evolving artificial intelligence landscape while balancing financial considerations and adoption rates.
Finding The Right Data To Train Artificial Intelligence Is Not Easy
Organizations face challenges in using high-quality data for generative AI models, as these models require vast amounts of accurate and relevant data. Instead of seeking perfect data, which is a daunting task, McKinsey’s Baig advises focusing on data that supports multiple use cases.
Deloitte’s Rowan highlights the importance of building strong data foundations for scaling artificial intelligence and CNH’s Kermisch notes that even creating large language models from text-based data is valuable and relatively quick.
The expected large-scale improvements, such as in vehicle design and cost reduction, have yet to materialize. CNH is currently fostering experimentation, training tech leaders and partnering with Microsoft to provide broader training, acknowledging that many early efforts are failing, but seeing value in rapid iteration.
Artificial Intelligence Is Here To Stay
IT leaders acknowledge that generative AI is here to stay despite the challenges it brings. Industry professionals like Phoenix Hospital’s Higginson and World Insurance’s CIO, Michael Corrigan, recognize the need for a strategic approach. Corrigan emphasizes that while generative AI is powerful and evolving rapidly, it matures slowly and requires a well-planned strategy and roadmap for effective implementation.
Without this, the technology may not positively impact business goals or enhance capabilities, leading to missed opportunities. Moreover, there is a growing concern about the rise of “shadow generative AI,” where employees use artificial intelligence tools independently, posing new information security risks.
As noted by Briggs & Stratton’s Olsson and Consumers Energy’s Dave Pawlak, organizations must implement generative AI securely, which is more complex than what the public experiences with open artificial intelligence tools. Despite these challenges, leaders like Baig believe that now is the time for companies to invest in generative AI, as it presents a unique opportunity for IT leaders to lead their organizations through this digital transformation.
Which of these hard truths do you wish you knew earlier about implementing generative AI in the enterprise? Share it with us in the comments section below.