The application of generative AI in enterprises is just starting to unfold, promising to revolutionize industries and transform the way businesses operate. Enterprises can use generative AI to drive innovation, automate repetitive tasks, improve decision making, personalize customer and employee experiences, and boost efficiencies. As such, businesses that effectively take advantage of generative AI should gain a significant competitive advantage.
As we know, it is the early days for generative AI adoption. As per IDC’s Future of Enterprise Resiliency Survey (March 2023), 24 percent of the respondents globally noted that they are already investing in generative AI technologies in 2023, while an additional 37 percent are doing initial exploration of potential use cases (Figure 1).
When asked about the use cases with the most promise for the technology, knowledge management applications was ranked number one followed by marketing and code-generation applications.
Figure 1: How Enterprises Are Using and Investing in Generative AI, Source: IDC, 2023
The latest class of generative AI systems has emerged from foundation models — large-scale, deep learning models trained on massive, broad, and diverse data that can be adapted or fine-tuned for a wide range of downstream tasks. The large-scale model era revolution is driven by the emergence of transformer model architecture in 2017. Transformers can take into account the sequential nature of data while also remembering distant words, which is key to generative AI’s ability to infer user intent in queries.
While generally available large language models (LLMs) such as GPT-4 and PaLM demonstrate impressive performance, simply applying these models to enterprise data poses several challenges:
- Lack of domain-specific data, language, and context.
- Lack of data security, access controls, and privacy.
- Increase in misuse and biases due to missing explanations and transparency.
- Lack of traceability, repeatability, and consistency.
- Hallucinations from combining unrelated sources to create new “facts.”
For enterprises to reap the benefits of large-scale foundation models for their business processes and integrate generative AI as part of their knowledge management strategy, they must overcome these challenges. Enterprises need to connect these models (Figure 2) with their proprietary data and thereby improve the contextual accuracy of outcomes. That data will include not just unstructured text but also tabular data, sensor data, and application data. Many industries with specific terminology and jargon can benefit from domain-aware foundation models that are fine-tuned on the relevant business lexicon. Responses must link to the underlying sources to provide confidence in their accuracy and acknowledge when there is no answer possible from the data available.
Figure 2: Turn a Foundation Model into an Enterprise Solution, Source: IDC, 2023
With proper and secure access to the corpus of data, generative AI models can, for example, answer inventory-related search queries for supply chain analysts, brand-compliant graphics for marketing managers, and identify potential compliance exceptions in new contracts. Sophisticated domain models have the greatest value potential for enterprises by not just tapping into existing enterprise data, but by calculating metrics and analyzing that data to produce useful content for enterprise users. Enterprise application providers are in a unique position to combine domain models with application-specific data to provide context-appropriate insights. A manufacturer or healthcare service provider, for example, can locate and retrieve answers that are enterprise and domain-specific, such as inventory reports or medical training curriculum, using data that is kept within their firewall.
Since 2017, the rate at which milestone AI models are built skyrocketed with the large-scale foundation model era in both scale of models and the pace of innovation. IDC projects that the worldwide generative AI applications market will grow from $1.061 billion in 2022 to $20.8 billion in 2026, representing close to 31 percent of the total AI-centric applications market. To harness business transformation with generative AI, IDC advises enterprises to:
- Create an environment of agile experimentation for the right prioritized use cases.
- Develop policies around responsible use of generative AI and inhibit nefarious scenarios.
- At a use case level, assess build vs. buy and partner with trusted technology solution suppliers and service providers.
- Find a platform with the flexibility to use different LLMs and other models to take advantage of rapid innovation and optimize cost efficiency.
- Ensure the ability to orchestrate and train other task-specific AI tools (e.g., agents, tool chains, reasoning engines, knowledge-specific modules) to deliver the most appropriate results (e.g., predictions, recommendations).
- Take advantage of pre-built applications and workflows that address specific use cases to reduce development effort and speed time to value.
- Consider flexibility of deployment options on both cloud and on-premises to cover a wider variety of use cases and data governance situations.
Guest contribution sponsored by C3 AI
To learn how C3 AI solves the challenges listed above, including data security, transparency, and hallucination, please visit the C3 Generative AI solutions page.
About the Author
Ritu Jyoti, a guest contributor, is Group Vice President, Worldwide Artificial Intelligence (Al) and Automation Research with IDC software market research and advisory practice. Ms. Jyoti is responsible for leading the development of IDC’s thought leadership for Al Research and management of the Worldwide Al and Automation Software research team. Her research focuses on the state of enterprise Al efforts and global market trends for the rapidly evolving Al and Machine Learning (ML) including Generative Al innovations and ecosystem.
