For nearly three years now, Generative AI (GenAI) has captured the imagination of enterprises worldwide, promising to transform customer experiences, boost productivity and unlock new revenue streams.
However, today, many large organizations find themselves grappling with the reality behind the hype. Market research and advisory firms have placed GenAI firmly in the Trough of Disillusionment phase, as firms come to grips with its true potential and limitations.
Investment continues across the industry, yet many companies are frustrated by slow, measurable returns. In this critical phase, senior business and technology leaders are asking: how can we manage the rollout and scale up of GenAI to deliver real business value and avoid becoming part of the 30% of GenAI projects that Gartner predicts will be abandoned by 2025?
What challenges do enterprises face in scaling GenAI?
Large enterprises racing to adopt GenAI are encountering a host of practical challenges, including poor data quality, inadequate risk controls, escalating costs and unclear business value, which threaten to derail projects before they reach production.
A major hurdle is the mismatch between investment and immediate returns. Another key challenge is organizational readiness. Many enterprises lack the data foundation and AI literacy to support GenAI at scale.
Low-maturity organizations struggle to identify the right use cases and carry unrealistic expectations, while more mature firms face talent gaps and need to instil GenAI literacy across teams. Ensuring data quality is also a persistent challenge, as GenAI systems like any other AI model, is only as good as the data it’s trained on.
Poor data leads to unreliable outputs. Governance and risk controls are often playing catch-up, with early adopters facing issues like model hallucinations, bias and meeting emerging regulatory compliance, such as the forward-looking, legally binding, EU AI Act.
All these challenges highlight that GenAI adoption is not purely a technology challenge, but also a people and process challenge. Siloed innovation efforts falter without cross-functional buy-in and projects pursued in isolation of business needs risk delivering no clear business outcomes.
How can organizations avoid GenAI project failure and drive value?
To move GenAI initiatives from pilot to production, enterprises must take a strategic, value-focused approach from the outset. First, establishing a clear business case and success metrics is essential.
Rather than deploying AI for its own sake, companies should start by identifying high-impact use cases where GenAI can solve a real problem or unlock measurable improvement, such as reducing customer service wait times or automating costly manual processes.
At the same time, organizations must rigorously analyze the total costs and potential business value of the initiative up front to make informed investment decisions.
Another best practice is fostering strong cross-functional collaboration from day one. Successful GenAI programs break down silos between IT, data science, business units and risk management.
This cross-functional approach ensures that technical teams understand business context and value drivers, while business stakeholders stay informed of AI capabilities and limitations. Promoting collaboration across teams empowers people at all levels to make informed decisions and drive innovation together.
One approach is to set up an “AI Council” or similar governance body with representatives from multiple departments, who can champion the initiative, align it with enterprise strategy and monitor ethical and compliance considerations.
Equally vital is managing the cultural and change aspects. GenAI often augments or redefines jobs and processes, so organizations need to prepare their workforce. This means upskilling and change management to help employees trust and effectively use AI tools.
Some early adopters have found it useful to start with pilot projects that involve end users and iterate based on feedback; demonstrating small wins helps build momentum and buy-in. In the current climate of high expectations, setting realistic milestones and celebrating incremental progress can prevent disillusionment.
While the hype may have promised immediate value, in practice GenAI success comes from a sequence of well-executed, value-focused steps.
What framework can guide a successful GenAI rollout at scale?
Rolling out GenAI in a large enterprise calls for structure. Companies need an operating model that can take AI from ideation to industrialized impact by enabling multidisciplinary teams to stay agile, without compromising on safety or accountability.
Many enterprises also use product-aligned operating models to connect AI work to business outcomes.
An effective way of guiding AI deployments is to apply a three-step framework from pilot to production.
The first phase, Discovery and Baselining, focuses on understanding the enterprise’s readiness and opportunities. This involves assessing the current data landscape, technology stack and AI maturity, while also identifying priority use cases through workshops with business leaders.
The goal is to define the problem, align on success criteria and build a shared understanding across stakeholders.
The second phase, Tooling and Design, covers the heavy lifting of building the solution. Here, organizations select the right tools and models and architect the solution with scalability, security and governance in mind.
It includes setting up the cloud or on-prem infrastructure and integrating the GenAI model with business workflows. Design also extends to user experience. For example, how a GenAI-powered assistant integrates into an employee’s daily tools.
The final phase, ROI and Scaling, is about proving value and then scaling up what works. In this phase, the GenAI solution is deployed in a real-world environment, often starting with a limited scope or user group, and closely measured against KPIs established in the discovery phase.
If the outcomes meet or exceed targets, the organization can confidently expand the use of the AI and start to institutionalize it as a capability. The focus in this phase is on adoption and enterprise change management as well.
Responsible AI must be embedded across all three phases of scaling GenAI. In discovery, define intended use and guardrails up front, assess data provenance and quality and set measurable responsibility metrics alongside business KPIs.
In design, engineer the system to those standards, including incorporating policy enforcement and access controls and applying bias and safety testing. In scaling and adoption, incorporate human-in-the-loop oversight for high-risk steps, continuous monitoring and incident response, audit trails and regular model re-evaluation.
Where are enterprises seeing success with GenAI?
Following the right approach, GenAI can deliver impressive results. For example, in the banking sector, an Australian Bank applied GenAI to its software testing process, which traditionally is a time-intensive, manual effort.
By leveraging GenAI, the bank was able to significantly accelerate its testing lifecycle and improve software quality, fostering a more collaborative and adaptive testing culture. In practice, this meant faster releases of new features to customers and higher confidence in those releases.
Another example comes from the pharmaceutical industry, where a North American pharma company used GenAI to reinvent its compliance and audit processes. The company’s existing rule-based document auditing system was costly and not user-friendly, so they worked with a partner to integrate a GenAI solution.
The result was an AI-powered assistant that could review regulatory documents and identify potential quality gaps with over 95% accuracy, while reducing manual document development efforts by 65% and increasing readability scores by 50%.
A marathon, not a sprint
The journey to rolling out GenAI across big businesses is a marathon, not a sprint. Many organizations are currently entering the trough of disillusionment, where initial experiments haven’t yet yielded the promised ROI. But this phase is survivable, as businesses rethink their AI strategies from hype to reality.
By addressing data quality head-on, investing in organizational readiness and fostering collaboration across IT and business domains, companies can avoid common failure points. Crucially, by turning to product-aligned operating models and setting realistic expectations, organizations can unlock significant value.
But achieving that at scale requires thoughtful Responsible AI, governance, iteration and a continuous focus on business outcomes. Enterprises that treat AI deployment as a holistic transformation, aligning technology with people, process and purpose, are the ones already turning initial AI investments into sustained ROI.
We’ve featured the best business intelligence platform.
This article was produced as part of TechRadarPro’s Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro
