Generative AI (GenAI) has cemented itself as a topic of discussion in many boardrooms, and leaders are keenly aware of the technology’s transformative potential. In fact, IBM’s Global AI ROI survey found that 85% of IT decision makers report making progress in executing their 2024 AI so it’s no surprise that organisations are doubling down on innovation. Despite their best efforts, however, many businesses are still struggling to turn their GenAI ambitions into meaningful results for customers, employees and shareholders.
Investigating the root causes
Keen to outpace competitors, many companies are currently in the third or second year of GenAI proof of concepts (PoCs), and some have even deployed the technology in one or more small corners of their business. Yet many are struggling to scale their GenAI solutions across the enterprise, and some even have trouble going beyond the experimentation stage. With companies investing significant resources into GenAI development, ‘Why can’t GenAI scale?’ is now a billion-dollar question.
In many cases, the issues do not stem from the technology itself, but are inherent in the businesses trying to adopt it – every organization is unique, so there is no one-size-fits-all remedy. That said, through the client projects we have worked on, we have identified common pitfalls that companies encounter in their GenAI adoption journeys.
Inflexible operating models
Most organisations are federated into a series of business units or departments, each with their own responsibilities and mandates for change, often specialised to meet the needs of specific markets, customers or regions. So, when organisations try to drive company-wide GenAI initiatives centrally, they inevitably start challenging the traditional federated operating model, creating friction between departments.
Similarly, if distinct business units start launching their own GenAI transformation projects without broader buy-in from colleagues and leaders across the business, they frequently face pushback. For instance, we’ve seen cases of business developing their own GenAI-powered assistant, but running into legal issues in a certain region because hallucinations combined with a lack of guardrails caused the solution to break local regulations. Had the IT department, legal department and board executives been involved in this initiative from the start, the issue would likely have been avoided.
Whether innovation comes from top-down initiatives or bottom-up experimentation, traditional federated business models need to be ready to flex to support successful adoption and effective governance. Strong partnerships, especially between IT and other business units, are essential for enabling this. Most of all, successful GenAI initiatives will have full executive-level support from business units and give leaders the mandate to drive change and set ambitious targets.
Lacking good quality data
Data fuels GenAI, so it comes as no surprise that low-quality data can create a poor solution that doesn’t function as intended. Most commercially available GenAI foundation models have been exposed to almost all of the information publicly available on the internet—from cat videos to product manuals. However, much of this data will be irrelevant to most business use cases,.
The key is for businesses to apply the power of foundation models to clean, quality-checked data that sits behind the enterprise firewall. When working with a major electronics company to build a GenAI assistant to help consumers fix common issues, we found that the majority of the generalist GenAI solutions did a very good job – after all, they have consumed vast amounts of data from electronics forums. However, some of the answers they gave were incorrect, and when it comes to fixing electronics, one small mistake can lead to significant health and safety concerns for consumers.
To protect consumers and derisk the GenAI journey for the business, we put in extensive guardrails and governance measures. While these measures initially meant that the assistant could handle a narrower range of queries, it could do so with much greater accuracy, creating a near-zero risk of providing false instructions. Starting with clean, validated and accurate data from reputable sources is essential to ensure that GenAI tools can serve their intended purpose.
Ineffective reuse of AI developments
When organisations develop an AI solution for one business workflow, whether from a top-down or bottom up approach, reusing the solution for other use cases can cause challenges across industries. Generally, this arises when different business units develop their own unique GenAI solutions without integrating data sources or aligning outcomes at a company-wide level.
As a consequence, different leaders start independently reinventing the wheel in the specific business function they oversee, which ultimately drains resources and delays return on AI investments. If leaders instead took a holistic view of company-wide GenAI adoption, they could build the concepts of reuse and integration into the heart of their objectives.
One of the best ways to enable reuse and, in turn, maximise return on investment, is to embrace standardisation. We recently worked with a multinational goods manufacturer selling to regional markets around the world to help them standardise and optimise their marketing activities. Previously, campaign ideas were developed by local marketing teams in over 40 locations. These teams each worked with local agencies to execute their strategies, leading to cost inefficiencies when developing and executing marketing campaigns.
Using GenAI solutions, we established a platform that enables marketing teams at head office to generate standardised marketing campaign ideas, which can then be automatically tailored to meet the needs of local audiences with AI tools covering everything from translation to content creation. Using this solution, the consumer goods manufacturer has reduced the average cost of a campaign and can run hundreds of campaigns per month, instead of one or two. Standardisation sets the framework for the reuse of assets and ideas on a global scale—helping organisations extract as much value from GenAI as possible.
Adopting the scalability mindset
Those leaders and companies that are willing to adapt the way they work and think will become the most successful adopters of GenAI. It’s about embracing flexible, creative thinking across the entire organisation so GenAI capabilities can be released out of the sandbox and into production to drive value enterprise-wide.
John Sullivan
John Sullivan is EMEA Managing Partner at IBM iX. A passionate advocate of continuous reinvention, John has spent his career helping clients deliver a step-change in business results by leveraging new and emerging technologies.
