The advent of generative AI (gen AI) is empowering enterprises around the world to transform the way they work by integrating AI directly into their fundamental infrastructure, systems and processes. The same goes for IBM where the IBM Chief Information Officer (CIO) organization was challenged to build a gen AI experience, now known as AskIBM, to assist IBM employees with routine business activities so they can focus on more strategic work. The goal was to offer every IBM employee an AI-first digital assistant that could automate day-to-day business tasks such as drafting an e-mail, translating a document or searching for information.
Given how fast the gen AI space is evolving, the CIO organization took quick action and leveraged IBM’s own AI technologies to build the alpha release of AskIBM in just 60 days using a rapid development and onboarding approach. The 60-day period included the IBM watsonx™ path to production as well as critical legal, privacy, cybersecurity and AI ethics reviews to meet the recommended AI standards and guardrails. During the process, the CIO organization collaborated with IBM Software and IBM Research® to deliver an application that was ready for enterprise release. As a result, access to the alpha release was made available to the more than 280,000 IBM employees globally in early January 2024.
The AskIBM foundation was built with watsonx.ai™ and IBM Granite™-class large language models (LLMs)1, and augmented with strategic internal IBM content—excluding IBM confidential information, sensitive personal information and customer information. As part of the development and deployment, they sought collaboration from IBM employees to help train and tune the AI foundation models by testing AskIBM. IBM employees were also encouraged to provide feedback to the product teams about how to improve watsonx to scale for IBM enterprise clients globally. AskIBM was designed to automate normal activities, such as drafting a client email, creating a product brief or quickly summarizing a long document to prepare for a meeting, call or other need using gen AI.
About 5,000 documents in multiple languages were ingested in the internal alpha release of AskIBM to train the Granite-class LLMs—and that number has increased to over 30,000 as of July 2024. The Retrieval-Augmented Generation (RAG) technique enables AskIBM to use the ingested contents to deliver tailored and appropriate responses to IBM employees by creating new responses from the most relevant information found. See figure 1 for the AskIBM answer flow.
The CIO organization plans to continue collaborating with IBM Software and other AI teams at IBM to use watsonx Orchestrate™ to help further enhance the unified user experience by automating other tasks and simplifying complex workflows.
1 watsonx.ai includes IBM-built foundation models referred to as IBM Granite. These multi-size foundation models apply generative AI to both language and code.
The AskIBM beta release was launched in May 2024. With the beta release the CIO organization delivered a unified user experience by having a single-entry point for all IBM employees to automate tasks to help increase productivity. Users can now go to one place to summarize, translate and create content, draft documents, chat with documents, find information and access other strategic digital assistants. Turning this type of work over to AskIBM helps to focus attention on more high-value activities.
Interestingly though, initial results showed that AskIBM performed well, but a roadblock to additional success was behavioral. The key to getting the most value out of AskIBM is using natural language to create clear and intentional prompts that provide the context needed to generate effective answers. However, at first many IBM employees searched using single or double keyword-based queries rather than natural language queries. In fact, upon release, natural language queries only accounted for 2–4% of the questions asked, but as a result of education and training, this amount increased to around 10% by the end of January 2024. Through continuing education, training and hands-on use, IBM employees are steadily adapting to this new way of working.
The CIO organization’s experience designing, developing and deploying AskIBM has yielded valuable lessons we can share with IBM clients worldwide. These include:
- Being adaptable and open to managing and incorporating approved changes while in development
- Making effective decisions by having the right people involved to expedite decision-making and foster collaborative teamwork to support a scaled enterprise initiative
- Creating a scalable solution by scoping a larger effort and employing a manageable, phased approach to development and user onboarding based on performance results and stress testing
- Prioritizing organizational change management from the beginning to plan education and training to help users adopt new AI technology and ways of working
The IBM Chief Information Officer (CIO) organization leads IBM’s internal IT strategy and is responsible for delivering, securing, modernizing and supporting the IT solutions that IBM employees, clients and partners use to do their jobs every day. The CIO organization’s strategy encompasses creating an adaptive IT platform that makes IT tools, applications and systems easier to access across the enterprise, accelerates problem-solving and serves as an innovation engine for IBM, catalyzing business growth.
Legal
© Copyright IBM Corporation 2024. IBM, the IBM logo, Granite, IBM Research, watsonx, watsonx.ai, and watsonx Orchestrate are trademarks or registered trademarks of IBM Corp., in the U.S. and/or other countries. This document is current as of the initial date of publication and may be changed by IBM at any time. Not all offerings are available in every country in which IBM operates.
Client examples are presented as illustrations of how those clients have used IBM products and the results they may have achieved. Actual performance, cost, savings or other results in other operating environments may vary.