Identifying coherent strategies for Enterprise AI Enablement is challenging. Many AI projects emerge from the bottom up based on specific use cases or AI features made available in existing enterprise software. However, a unified strategy can guide AI project development across the enterprise, ensuring each project contributes to the larger transformation effort.

Generative AI can increase the value of information across the enterprise by making it easier to access and understand, more accessible and valuable to a broader range of users, and integrating data from different sources to generate new insights.

This feature of Generative AI can be used as an incentive to drive digital transformation and automation across the enterprise, driving the development of data APIs to make data available to AI applications and other automation, driving the development of standardized data models and documentation to make data more accessible and valuable to AI applications, and providing an interactive platform for exploring and developing new fully automated processes.

I'll outline some strategies for deploying generative AI across an organization. These include using Generative AI to build an enterprise semantic index, leveraging new Generative AI functionality in existing platforms, using platform APIs and Generative AI to expand data accessibility, and using Generative AI to integrate data from multiple platforms, generating new insights. Each step emphasizes the incremental benefits and contributions to a cohesive digital transformation.

AI Can Drive Digital Transformation

Generative AI makes enterprise data more accessible and valuable to a broader range of users, from experts to novices. It offers a natural language interface, enabling users to ask questions across entire datasets or individual documents. Generative AI can guide users through complex information, summarizing, explaining, inferring, translating, and transforming data as needed. It integrates and correlates data across systems, extracting structure from unstructured data or transforming structured data into more usable forms. These capabilities can drive broad digital transformation by incentivizing incremental transformation projects that lay the foundation for the larger automation effort.

One challenge in digital transformation is ensuring each incremental step provides immediate value. If many steps are required before realizing any benefit, it can be challenging to justify and maintain momentum.

Immediate Benefits from Incremental Improvements

Generative AI projects rely on the same infrastructure improvements as other automation projects (e.g. data models, APIs, authentication systems). With Generative AI, each incremental project can deliver perceivable value. Even making a single data source available through an API can yield significant benefits when combined with an AI Copilot that makes the data more accessible and valuable to users throughout an organization. Each additional data source increases in value non-linearly, as Generative AI can help discover new relationships and insights across datasets.

ROI

Quantifying the Return on Investment for AI enablement projects is challenging.

AI projects can increase customer satisfaction by providing faster and better responses from Customer Service Representatives:

• Utilizing customer interaction history and summarization for contextually relevant responses.
• Implementing runbook copilots for efficient troubleshooting.
• Enabling billing invoice Q&A for quick resolution of common queries.

However, increased customer satisfaction is difficult to quantify in terms of ROI.

AI projects can also increase employee productivity:

• Enterprise Copilots:
  • Facilitating faster access to necessary information.
• Incident Copilots:
  • Speeding up troubleshooting to mitigate issues affecting customer satisfaction and revenue.
• Software Engineering Productivity:
  • Tools like GitHub Copilot suggest code snippets, reducing coding time and errors.
• Onboarding New B2B Customers and Partners:
  • Contract and partner onboarding copilots streamline and expedite processes.

Again, the ROI for increased productivity is difficult to quantify. Despite this, the value these projects bring to efficiency, innovation, and business growth is undeniable.

Copilots and Autopilots

Generative AI solutions can be implemented as copilots or autopilots. Enterprise Copilots make existing information more valuable to more people, while Enterprise Autopilots take action based on new analytic and operational data, performing tasks faster and at a scale beyond what is possible with humans alone.

Enterprise Copilots assist users in completing tasks, such as:

• Chatbots integrated with enterprise data and APIs, providing answers based on organizational information and assisting with tasks (e.g. Glean, Mechanician).
• Enhanced programming editors that suggest context-based code (e.g. GitHub Copilot).
• AI-enhanced user interfaces in enterprise systems (e.g. Salesforce Einstein Copilot).
• Automated self-service support for internal employees, partners, and customers.

Enterprise Autopilots interact with automated processes, carrying out tasks typically done by humans, such as:

• Content analysis for message and event routing.
• Extracting structured data from text for automated processes.
• Generating content and messages in response to real-time data and events.

AI Enablement Strategies

Below are some strategies for applying Generative AI across an organization. The overall goal is to increase the value of the information spread throughout the enterprise by making existing data platforms more effective for existing users, making existing data more usable to more people, and integrating siloed data across more systems.

1. Build an enterprise semantic index with an LLM front-end to make enterprise knowledge more accessible and valuable to a broader group of users. Start with data sources that are easily indexable using Embedding vectors that encode data using a semantic representation similar to how LLMs encode information. These Embedding vectors will typically be stored in a separate database, usually called a Vector Database. You can assemble your own Vector Database and LLMs or use a third-party platform like Glean that provides a no-code/low-code solution.

These data sources should be text-heavy and relatively static, including:

• Easily Indexable Knowlege Bases
  • Wiki Platforms (e.g. Confluence)
  • Issue Tracking Systems (e.g. Jira)
  • Cloud Drives (e.g. SharePoint)
  • Email Systems (e.g. Outlook)
  • Messaging Platforms (e.g. Slack)
  • Data Catalogs (e.g. Alation)
  • Code Repositories (e.g. GitHub)
  • Enterprise Architecture Management Platforms (e.g. LeanIX)

Over time, expand the index to include data sources that are not as easily indexable and contain data that must be retrieved at runtime using APIs using a Retrieval Augmented Generation (RAG) or Function Calling approach. Strive to expose the data from each source in a manner that can be reused in future AI and automation projects.

• Considerations
  • What data sources are easily indexable using vector embeddings that provide a semantic representation of the data?
  • What data sources are not easily indexable and contain data that must be retrieved at runtime using APIs using a Retrieval Augmented Generation (RAG) or Function Calling approach?
  • How is authentication handled for enterprise users' data sources during indexing and runtime?

2. Leverage Generative AI capabilities from existing platform vendors to enhance their value to core users.

   Every enterprise platform has a core set of users who rely on it for their regular work; enabling Gen AI within the platform makes the most sense for these users. The platform's data will be available to a Generative AI, and the standard user interface will usually be enhanced with Generative AI functionality.

• Considerations:
  • Who are the core users of these platforms?
  • What Gen AI capabilities are available in our current software platforms?
  • Are there capabilities that we want/need and are missing?
  • Can the platform be extended to support these missing capabilities, or must we use a third-party approach?

3. Use platform APIs to expose data to the Enterprise Semantic Index and other AI applications, increasing data accessibility beyond core users. This will increase the value of the data by expanding its user base beyond each platform's core users and making it more accessible and valuable to those who are not experts in the data domain but who can benefit from it with the assistance of an AI; LLMs offer a user-friendly interface for non-experts to access and comprehend data that was previously challenging to utilize effectively.

Use the opportunity to expose data from each platform in a manner that can be reused in future AI and automation projects. Consider industry-standardized APIs and data models (e.g. TOGAF, TM Forum ODA) that can be leveraged by third-party systems to access the platforms and their data.

• Considerations:
  • Does the platform's licensing make it cost-effective or technically feasible for this expanded user base?
  • Can third-party systems use the platform's data through APIs?

4. Use Generative AI to combine data from different platforms to generate insights unavailable from a single source, further expanding the user base and value.

Generative AI can correlate data from multiple sources through inference, translation, and transformation. By interacting with users to explore and express relationships between diverse data sets, it reveals insights that single sources cannot provide.

Additionally, it offers a platform for interactive experimentation with data from various systems, uncovering new automation opportunities.

• Considerations:
  • Can any of the overlapping platforms support these use cases, or are third-party solutions required?

Conclusion

These strategies for Enterprise AI Enablement focus on incremental improvements that deliver immediate benefits while laying the foundation for future projects. This approach makes enterprise data more accessible and valuable while driving further organizational transformation efforts.