A poll of major businesses from the Capgemini Research Institute has found that while using artificial intelligence (AI) for automating repetitive business tasks offers significant savings, the tasks being improved are relatively simple. A significant number of the business leaders surveyed also say they prefer proprietary AI implementations over open source alternatives.
According to the survey of 1,607 executives from organisations with at least $1bn in global revenue, business leaders are seeing a 40% reduction of customer operations costs thanks to AI and generative AI (GenAI). The executives polled also saw a reduction of 26% in people operations costs, and a 24% reduction of finance and accounting costs. Respondents also achieved a 21% reduction of supply chain and procurement costs.
As an illustration of agentic AI potential in the industry, Capgemini highlighted Yum Brands, the parent company of Taco Bell, which operates 60,000 restaurants worldwide. The company introduced an AI-powered restaurant manager to track crew attendance and plan shift patterns, as well as suggest adjusted opening hours to align with market conditions.
While such examples show the potential of AI and GenAI to improve business efficiency, the research found that a significant portion of the gains reported by respondents tend to be associated with automating straightforward, repetitive tasks. According to Capgemini, this suggests the use of AI and GenAI among the executives polled represents early stage efficiencies rather than long-term transformational impact.
Savings need to be balanced against the cost of running the AI system. Capgemini Research Institute noted that the price of querying a trained model is falling dramatically. For instance, OpenAI’s GPT 3.5 experienced a decrease from $20 per million tokens to $0.07 per million tokens, while GPT-4 had a reduction from $15 to $0.12 in a year.
Techniques such as model pruning, quantisation and distillation can be used to reduce the size and complexity of AI models. As Capgemini Research Institute points out, these optimised models require fewer computational resources, which lowers inference costs. Alongside more efficiency algorithms, Capgemini Research Institute said that efficient hardware utilisation, batch processing of inference requests, dynamic scaling to adjust the number of computing resources based on current demand, and energy-efficient algorithms can significantly reduce the power consumption of AI models.
However, although open-source models such as DeepSeek have been shown to achieve an 11x reduction in compute costs without compromising performance and can address the advanced hardware bottleneck many organisations face, the poll showed that business executives are less enthusiastic about open source AI compared with proprietary AI models.
Despite the increasing performance and cost advantages of open-source AI models, Capgemini reported that a significant majority of executives continue to favour proprietary AI implementations. Three-quarters of the executives surveyed prefer proprietary models, with 43% opting for those developed by hyperscalers and another third choosing models from niche providers.
Capgemini found that the preference for proprietary models and AI systems is particularly strong among organisations that have scaled up their investments in AI and GenAI, which, the report’s authors suggest, indicates a clear trend towards trusted, enterprise-grade AI products that offer robust support, security and integration capabilities.
The findings, published in Capgemini Research Institute’s AI in action report, identifies a number of trade-offs that curb enterprise adoption of open-source models due to the trade-offs IT and business leaders need to make. These include the need for greater technical expertise, potential exposure to security vulnerabilities and reliance on community-driven support that can impact update cycles and documentation quality.
Oliver Pfeil, CEO of business services at Capgemini, said: “GenAI and agentic AI can truly transform business services – enabling the shift from traditional cost-focused models towards an AI-enabled, value- and insight-driven business. Those that adopt an integrated approach with data and AI at its core will be set to achieve a truly connected, frictionless enterprise.”
However, he noted that the research suggests organisations face numerous barriers scaling up the deployment of AI agents. “Adopting a pragmatic approach, fostering trust in AI and creating a strong data foundation will go a long way in transforming business services into a strategic powerhouse to fuel any enterprise,” he added.