AI pilots are fun. AI at scale is a business.

In the world of artificial intelligence, it’s easy to get caught up in the excitement of small-scale experiments. AI pilots can be thrilling, particularly when the technology they rely on is projected to generate as much as USD 4.4 trillion in annual value. A successful chatbot demo, an internal resume analysis tool—these pilots offer an opportunity to test the waters without committing (or investing) too much. But while these small-scale experiments can feel like a victory, they rarely offer a clear picture of AI’s true value to an enterprise. As the novelty wears off, business leaders are struggling to gauge AI’s effectiveness based on these isolated experiments.

This is in large part because a simple pilot or trial can’t fully unlock AI’s potential. It’s at the scale level—where AI is embedded deeply within business processes and across departments—that AI delivers on its enormous promise. Enterprise AI at scale is where the real value emerges, transforming business models, driving efficiencies and creating new growth opportunities.

There’s no question that pilots are fun, and that they represent a cautious, incremental approach to deploying AI. These projects give an organization a taste of what the technology can do—perhaps to improve a single process or optimize a small segment of a business’ operations. Often these pilots have impressive, if hyper-localized, results.

But pilots often focus on isolated use cases with limited scope. They’re designed to work in a controlled, manageable environment, like automating the customer experience for a limited segment or applying AI to a specific problem like inventory management. These trials are intended to prove a concept, not to deliver a holistic solution. They don’t provide true benchmarks for determining AI’s larger value, deployed at scale.

Scaling AI across an enterprise involves complexities that can’t be captured in a single trial run; a cross-departmental organization-wide series of solutions, built on a solid technology foundation, can generate exponential returns.

Yet a surprisingly small number of businesses have truly committed to extracting this value, particularly when the emerging landscape of generative AI. According to recent research from the consultancy McKinsey, only 11% of companies have adopted generative AI at scale.

But when it comes to the transformative potential of artificial intelligence, playing it safe can be a bigger risk than cautiously rolling out a series of siloed projects. According to a recent report from the IBM Institute for Business Value (IBV), 43% of CEOS planned to increase the pace of digital transformation with AI in 2024, concerned that without it they would fall behind. Those CEOs realized that AI is most valuable when deeply woven into their organization’s operations from end to end. 

AI at scale is where true business transformation is possible. Bringing these tools out from the silo of one-off projects and infusing them across the organization can provide a significant strategic advantage. An organization that invests in the fundamental architecture required to scale AI cannot just eliminate pain points but gain holistic data-driven insights into their operations across the board.

And scaling, in the current environment, can solve several discrete problems. Just look at the numbers collected by the IBV:

• Over one-third of CEOs already say that their workforce will require retraining and reskilling over the next three years to handle the increased use of automation and AI.

• Gen AI is considered a "critical" force in the design and development process by 85% of digital product leaders.

• Among leading CEOs who have invested in infrastructure and organized data for AI projects, 90% report that the scaffolding they've built has enabled further scaling and value delivery.

A single pilot project can’t articulate these longer-tail benefits of adopting AI at scale, or the cost of training employees and building AI tools individually.

AI thrives on data, and building a high-quality data foundation for use across an organization increases the chances of accuracy, as well as the ability to quickly build new tools. And, as AI models improve over time, long-term commitments to enterprise AI across a business are likely to yield better results. 

To see these concepts in action take, for example, communication service providers (CSPs). The telecom industry adopted AI, and particularly virtual agent technology (VAT), early. It’s a natural fit for CSPs: Many have high volumes of diverse customer interactions and a mandate to provide consistent customer experiences. Most of these businesses monitor the cost of an individual interaction closely, as individual customers can have relatively small contracts.

Applying VAT to a narrow set of customer experience use cases has improved CSP’s performance. According to a report from Forrester, a large organization implementing VAT might achieve cost savings of USD 5.50 per contained conversation. The IBM Institute for Business Value found that 84% of these businesses either met their ROI or exceeded it following adoption, and 97% reported a positive impact on customer satisfaction. By all accounts, the CSP adoption of VAT for customer experience has been a widely successful pilot. Given how profitable these initiatives have been, it’s surprising that they haven’t often been expanded or scaled.   

But Vodaphone, an outlier, did more than just apply AI to a narrowly defined function. In 2017, the company started its own assistant, TOBi, in the United Kingdom. Six years later, TOBi handles hundreds of thousands of calls per month. And TOBi also became a launchpad for a series of deeper integrations. The company created a complementary chatbot that addresses different brands and geographies in the unified voice of the brand, and then deployed it across channels like social media and SMS.

Today, Vodaphone has added chatbots tailored to suppliers and for use in retail environments. Vertical integration has made it easier not just to reduce the cost of each interaction, but to quickly add support for new channels and products. These multi-channel, scalable tools purpose-built to perform different functions across the company reveal the gap between an AI pilot and an AI-first strategic approach: The former might improve productivity, but the latter creates avenues for significant top-line growth. 

Given the potential returns of moving from AI pilots to large-scale transformations, business leaders must make a choice: Reactively create individual tools to overcome bottlenecks and under-performing departments, or proactively invest in the scaffolding to make AI a core facet of their operations.

There is more up-front effort required to train, tune, deploy and adopt AI across business functions, but in the current landscape those who don’t choose to adapt will fall behind. Already, two-thirds of CEOs we surveyed say the productivity gains from automation are so great they take on risk to remain competitive. Building the fundamental architecture for infusing AI across an organization is more than a pilot; it’s a business model for the future.