The News: IBM announced that watsonx.governance will be generally available in early December to help businesses monitor and manage AI models. Governance is required to ensure that operational, regulatory, and legal risks are properly monitored and managed throughout the operational lifecycle of each generative AI model that is used. Specifically, watsonx governance will help organizations:
- Manage AI to meet upcoming safety and transparency regulations and policies worldwide – a “nutrition label” for AI
- Proactively detect and mitigate risk, monitoring for fairness, bias, drift, and new large language model (LLM) metrics
- Manage, monitor, and govern AI models from IBM, open source communities, and other model providers
You can read the press release announcing the news on IBM’s website.
IBM watsonx.governance Set for General Availability in December
Analyst Take: IBM announced that watsonx.governance will be generally available in early December to help businesses address the challenges with monitoring, managing, and operating generative AI models and tools. In particular, the generally available release will include various capabilities that are designed to address the use of LLMs or foundation models, in terms of managing risk, embracing transparency, and anticipating compliance with future AI-focused regulation.
The expanded governance capabilities for LLMs focus on four main areas: monitoring of inputs and outputs of LLMs; visibility into the development of models; transparency around the operation and use of LLMs; and validation of LLMs to ensure the models are working properly. Each aspect is important and particularly relevant to enterprises using generative AI, whether using open LLMs, LLMs embedded within SaaS platforms, or custom-developed LLMs.
Monitoring LLM Metrics Inputs and Outputs
One of the key issues that enterprises face when using generative AI is properly monitoring both the inputs and outputs of LLMs, then alerting administrators when pre-set thresholds are breached for quality metrics and drift, instances of toxic content (including hate speech, abuse, and profanity), and the presence of personal identifiable information (PII). This need is particularly important in the context of generative AI that is used in customer, employee, or other interactions where humans are likely to be interacting directly with AI or could be affected by the output of a generative AI system or tool.
Visibility into LLM Development
IBM says watsonx.governance will automatically collect information about the model building process while explaining decisions to mitigate hallucinations and other new risks. This process must be undertaken on an ongoing basis with any AI model to ensure that the model is grounded in relevant data and that its outputs can be refined to mitigate hallucinations. Regardless of whether a generative AI tool touches or impacts humans, this visibility will be required to instill confidence in generative AI.
Transparency of AI Lifecycle for LLMs
The latest release of watsonx.governance is designed to automatically document model facts across all stages of the lifecycle, monitor for drift for text models, and track health details such as data size, latency, and throughput to identify bottlenecks and compute-intensive workloads.
Again, organizations seeking to deploy generative AI will want to pay attention to this feature, particularly as customers, partners, and, to some extent, regulators might want to audit AI models to ensure that they are functioning as designed upon deployment and throughout their lifecycle. This consideration will be increasingly important as generative AI models become more complex and incorporate a wider range of data sources.
Validation Tools for LLMs
Watsonx.governance also enables prompt engineers to map LLM outputs to provided context/reference data for Q&A use cases to determine whether the LLM is appropriately influenced by the reference data to help ensure it is relevant to the output. This validation step is critical to the accuracy of the model and is going to be elevated in importance as third-party data models located on-premises or in the cloud are used to feed the model.
IBM says it is planning to expand watsonx.governance’s validation capabilities—as well as the other three governance capabilities—in first quarter (Q1) 2024 to allow clients to govern third-party AI models from any vendor — on cloud or on-premises — to orchestrate governance processes across their entire organizations. This feature is critical for enterprises, as most will utilize a wide range of models, both large and small, and might deploy them across a mixed data environment.
Governance Is Key to Enterprise Use of Generative AI
IBM has done an admirable job of embedding AI governance tools and capabilities within its watsonx.governance platform, no doubt to appeal to both nervous IT managers, legal professionals, and even denizens of the C-suite. The need for these types of tools is clear, but the real benefit comes from deploying and actively using them consistently.
Data—particularly ever-changing, large amounts of corporate data – must be consistently monitored, cleaned, and analyzed when it is used to train or ground an AI model. Because this data is growing continuously, edge case data points, bias, toxicity, or other unwanted elements might creep in, and if these “dirty” data make their way into an AI model, it could have negative consequences. That is why a governance platform or specific governance tools should be used in conjunction with data-monitoring solutions to ensure that AI models that refer to and are grounded by company data perform as designed across their entire deployment lifecycle.
Disclosure: The Futurum Group is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.
Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of The Futurum Group as a whole.
Other Insights from The Futurum Group:
IBM Helping Riyadh Air Get into the Skies in 2025 with Digital Tech
IBM Q3 Revenue Up 4.6% to $14.75 Billion, Beats Analyst Estimates
IBM Rolls New Granite AI Models, Continues watsonx Platform Momentum
Author Information
Keith Kirkpatrick is VP & Research Director, Enterprise Software & Digital Workflows for The Futurum Group. Keith has over 25 years of experience in research, marketing, and consulting-based fields.
He has authored in-depth reports and market forecast studies covering artificial intelligence, biometrics, data analytics, robotics, high performance computing, and quantum computing, with a specific focus on the use of these technologies within large enterprise organizations and SMBs. He has also established strong working relationships with the international technology vendor community and is a frequent speaker at industry conferences and events.
In his career as a financial and technology journalist he has written for national and trade publications, including BusinessWeek, CNBC.com, Investment Dealers’ Digest, The Red Herring, The Communications of the ACM, and Mobile Computing & Communications, among others.
He is a member of the Association of Independent Information Professionals (AIIP).
Keith holds dual Bachelor of Arts degrees in Magazine Journalism and Sociology from Syracuse University.
