In this era of digital abundance, Generative AI promises to be a transformative force, capable of enhancing service quality and user experience across industries and value chains. Early use cases have revealed opportunities to automate processes, improve customer support, and empower workers to be more efficient and creative. However, the most valuable applications are likely to require creativity from the businesses themselves.
Currently, many businesses are consuming AI passively through off-the-shelf applications, chatbots, HR portals, and transcription services. For maximizing the value of Generative AI, organizations must transition from being passive AI consumers to proactive AI value creators, adapting foundation models to fit the unique needs of their business.
The Transformative Power of Foundation Models
Foundation models are AI models trained on very broad, vast datasets, enabling them to be easily applied to a wide range of use cases. Generative AI makes it easier to train and adapt foundation models for diverse downstream tasks.
This versatility is tremendously valuable. A single model can underpin chatbots, HR self-service, marketing email generation, and legal document summarization, enabling innovation across a full spectrum of applications. The key is to see the model not as the final form of AI but a foundation on which to build. However, this can only occur if businesses are actively engaged in AI innovation and take a platform approach to AI consumption.
There are three main approaches to AI consumption:
- Embedded AI
This is pre-packaged AI that comes with off-the-shelf software, like buying a smartphone with a built-in camera app. The software vendor creates the AI, and consumers use it. It can be handy — like having a tool that helps write emails with just the right tone or edit images automatically. But here's the catch: This handy tool is also available to other businesses, including competitors. It sets a new standard and baseline, but it doesn't make any one business stand out. - API Calls
As users build custom applications, those applications can make calls to another company's AI service through APIs, kind of like asking a friend for advice. Users can differentiate themselves by cleverly using these services, but there are drawbacks to be mindful of. First, everyone has access to the same services. Second, it's a bit like connecting to a black box; users can't see what happens on the other end. Long-term value is also a concern as companies may be giving away more than they realize. - Platform
With the platform approach, users become the master chef of their own AI kitchen. Businesses get all the ingredients to create their own AI solutions, including foundation models, business data, as well as the tools to tweak them. This doesn’t mean that they have to start from scratch or spend a fortune. Businesses can tweak or build models as needed. Importantly, the value accumulated is specific to the business. This is how to truly own and differentiate an AI solution.
Each of these approaches has its pros and cons, and most businesses should use a mix of all three. But the platform approach is where the magic happens. Rather than using another company’s pre-built solutions, businesses are in control of the entire AI creation process.
The platform approach makes it conducive to aggregating business data from diverse sources. Over the period with an accumulation of business data, the richness of context increases, making the output of LLMs both contextually relevant and factually accurate. Hence, the businesses can accumulate value over time giving them a competitive advantage. The platform approach uses Retrieval Augmentation Generation (RAG) to act as a bridge, connecting large language models (LLMs) to vast enterprise knowledge sources.
Platform (Your Data + Your Model) = Your Unique Value
Other key advantages of the platform approach include the following:
- Choice of Model. There are two main categories of pre-trained large language models (LLMs): proprietary and open source. Proprietary LLMs are owned by other companies and tend to feature extensive parameter sizes, reaching into the billions. Conversely, open source LLMs provide greater transparency, allowing a deeper understanding of their operations and source data, and encouraging contributions from the community. Among businesses, open source LLMs are gaining recognition as valuable assets due to their accessibility/collaborative capabilities, ability to assist with audits, and support of ethical and legal compliance.
- Choice of Hosting. The platform approach offers greater flexibility for provisioning and hosting the Gen AI stack, either on private, public, or hybrid cloud environments. It acknowledges the significance of low latency and high bandwidth in scenarios requiring real-time decision-making and interactive user experiences. The flexibility extends to crafting platform variants tailored for both horizontal and vertical use cases, ensuring relevance across a spectrum of applications and scenarios.
- Data Governance and Control. Control over sensitive data is a priority for any enterprise. With the platform approach, teams know exactly how the model is built, what data was included, and how it's being used. This level of control is crucial for preventing issues such as unethical outputs due to bad quality data, hallucinations, discrimination, or inadvertent use of sensitive content.
- Long-Term Sustainability. The platform approach is about long-term thinking. It's not a quick fix; it's an investment. While it may require effort and strategy upfront, the goal is to avoid starting over every time there's a shift in the wind. It ensures sustainability and success over the long term.
As businesses navigate the evolving landscape of Generative AI, recognizing the transformative potential of foundation models is key. Shifting from the traditional task-centric AI approach to a versatile platform approach heralds a new era where businesses don't just use AI but actively create value, defining the future contours of the AI landscape.
Crucially, the platform approach focuses on creating and owning value unique to businesses, transforming them into AI value creators. In essence, the platform approach empowers businesses to be their own AI architects, tailoring solutions to their specific needs and creating a sustainable and valuable AI ecosystem in the process. It's the route to not just participating in the AI revolution but leading and shaping it.
