aIQ glossary

Autonomous entities within AI systems designed to perform tasks, make decisions, and interact with their environment based on predefined goals or instructions.

Frameworks to ensure AI systems are ethical, fair, and compliant with regulations.

The study of methods for making computers mimic human decisions to solve problems, including tasks such as learning, reasoning, planning, perception, language understanding, and robotics.

The use of technology, particularly artificial intelligence and machine learning, to perform tasks or processes without human intervention, improving efficiency, accuracy, and speed while reducing operational costs and freeing up human resources for more complex and creative tasks.

In the context of language models, bias refers to the tendency of a model to favor or exhibit certain stereotypes, prejudices, or imbalances based on the training data it was exposed to. Efforts are made to reduce and mitigate biases in language models.

AI tools designed for structured, task-based customer interactions (e.g., FAQs, basic support).

A language model developed by OpenAI, based on the GPT-3.5 architecture, capable of generating human-like responses in conversational contexts.

A subset of artificial intelligence that simulates human thought processes in complex scenarios. These systems use self-learning algorithms, data mining, natural language processing, and pattern recognition to understand, reason, learn, and interact with humans in a way that is intuitive and contextually aware, assisting in decision-making and problem-solving tasks.

The information or conversation history that provides meaning and relevance to the current prompt. Generative AI uses context to generate responses that align with the ongoing conversation.

Advanced AI tools capable of handling multi-turn conversations with contextual understanding (e.g., virtual assistants).

Refers to a collection or dataset of text, images, or any other type of data used to train generative models. The corpus serves as the source of information from which generative models to learn the patterns, distribution, and characteristics of the data. It is used to train models for tasks such as text generation, image synthesis, or multi-modal generation. Examples: A text corpus can consist of books, articles, websites, or any collection of textual data. An image corpus may include images from online databases, custom datasets, or image repositories.

Models/agents that produce the same output for a given input.

The process of enhancing or augmenting data with additional information to improve its quality, relevance, and utility for AI models. This can involve adding context, metadata, or external data sources to the existing dataset.

The concept of designing AI systems to ensure fairness, transparency, and accountability.

It is crucial to be aware of the ethical considerations associated with generative models. As they have the potential to generate realistic fake content, issues such as misinformation, bias, and privacy concerns should always be considered. Business professionals should understand these implications and use generative models responsibly by always checking any, and all output generated by these models.

The degree to which an AI model's decision-making process can be understood and interpreted by human stakeholders, making it clear and accountable, which is crucial for compliance and trust.

The principle of ensuring that AI models do not produce discriminatory or biased outputs in areas such as lending, hiring, and law enforcement, often measured by metrics like group representation, economic impact, and prediction accuracy.

AI models that create new content based on input data, such as text, images, or music (e.g., ChatGPT, DALL-E)

Generative models are machine learning models designed to learn and generate new data that resembles a given training dataset. They capture the underlying patterns and distribution of the data to create new samples. In Generative AI, the term "LLM" (Large Language Model) refers to the type of language model which is the Transformer-based language model. Transformers have gained significant popularity due to their effectiveness in capturing long-range dependencies and generating coherent and contextually relevant text. The most famous Transformer-based language model is OpenAI's GPT (Generative Pre-trained Transformer) series, which includes models like GPT-2 , GPT-3 & GPT-4.

The initial input or instruction given to a Generative AI model to generate a response. It can be a question, statement, or any text that provides context for the conversation.

Prompt engineering refers to the process of designing and refining prompts or instructions given to a language model to elicit desired responses. It involves crafting specific cues, providing context, or utilizing techniques to guide the model's output towards desired outcomes. Prompt engineering aims to improve the quality, relevance, and control of generated responses by shaping the input prompts appropriately.

AI models that are designed to perform logical and analytical reasoning to solve complex problems and make decisions based on rules, constraints, and available data. Unlike traditional generative AI models, which focus on creating new content based on learned patterns, reasoning models prioritize logical consistency and context-awareness, often providing explanations for their conclusions.

A unit of text that represents the smallest meaningful or discrete element, typically a word or a character, used as input or output in generative models.

Tokens are the individual units that make up the input data, such as words, characters, or subwords. The token limit refers to the maximum number of tokens that can be processed by the model in a single inference or generation step.

The process of breaking down a sequence of text into individual tokens, allowing for better representation and processing by generative models.

The large dataset used to train Generative AI models. It consists of diverse text sources, such as books, articles, websites, and other written content, to enable the model to learn patterns of language usage.

In the context of language models, hallucination refers to the generation of incorrect or nonsensical information in the model's responses. It can occur when the model produces text that may sound plausible but is not accurate or supported by factual evidence.

Lighthouse is a technology organization focused on innovation, education, and evolution. COE stands for Center of Excellence.

KPMG’s strategic approach and commitment to responsible and ethical AI usage.

Top 20 member firms at KPMG.

Global Digital Steering Group. Key stakeholders and decision-makers for digital. Governance body that meets monthly.

Top eight member firms at KPMG. Includes Australia, Canada, China, France, Germany, Japan, the UK, and the US.

Bringing human intelligence and AI into daily operations.

Country entities at KPMG. For example, Australia, the UK, or New Zealand. Each country has its own CEO and operates like a business.

Certifies that all AI solutions built on the platform adhere to KPMG’s Trusted AI framework.

Tax, Audit, and Advisory.

KPMG's proprietary approach to using the global KPMG Member Firm network as the first client in KPMG's global AI transformation journey. This journey begins with awareness, adoption, enablement, and evolution.

Developers Conference event in November 2024.

AI tools within Workbench that provide everyday or specialized capabilities. Specialties include ESG Compliance and Credential Builder.

The user interface (UI) and design standards.

Firm-wide strategy to be the most trusted professional services firm. Prioritizes AI and technology. Released in 2021 and reshaped as LEAP 2.0.

Designed to leverage AI for business transformation and sustained competitive advantage, it outlines key objectives and activities across several strategic pillars and critical enablers to ensure a consistent approach to AI across all member firms. 

Structured Global AI program sub-team entities comprised of networks of SMEs and technical experts that collaborate to conceptualize, develop and deliver services, artefacts, support and resources to achieve defined AI transformation goals.

This pillar focuses on enhancing market presence and sales capabilities through AI. It includes organizing the Global PI team for success, amplifying the brand, establishing thought leadership, and enhancing market and sales capabilities.

This pillar aims to innovate and diversify AI client services and transform existing service offerings. It involves the deployment of AI in various functions such as Audit, Tax, and Advisory, and includes initiatives like the KPMG Workbench and Country Enablement with AI Leads.

This pillar focuses on deploying AI accelerators to improve service delivery and client experience. It aims to refine the service delivery model and introduce innovative tools.

This pillar involves reimagining and future-proofing the business model to meet evolving market demands. It includes defining business model disruption and driving ideas and incubating new service offerings.

This pillar focuses on modernizing business operations and driving efficiency through AI adoption. It includes modernizing the business model and driving efficiency and reducing costs.

Essential components and support mechanisms that facilitate the successful implementation and adoption of an AI strategy within an organization. These enablers are crucial for ensuring that the AI initiatives are effectively integrated into the business processes and that they deliver the intended value. 

Grouping of activities intended to equip KPMG's member firms with the necessary tools, support, and resources to effectively implement and deliver the Global AI Strategy at a local level. (See also Immersion and Adoption and Country Activation)

Activity to drive the change management and educational efforts needed to upskill employees and build a culture that embraces AI. It includes training programs, communication strategies, and support systems to help employees adapt to new AI tools and processes.

Activity focused on partnering with AI Country Leads to align local and global ambitions, ensuring continuous engagement and communication, performance measurement and reporting and facilitation of exchange of insights and best practices among member firms to build a strong network of AI expertise.

Work to develop and deploy KPMG Workbench as a global collaboration platform that streamlines efforts, reduces costs and drives revenue, while providing access to global AI tools for everyday and specialized use.

The  Global AI Ninjas and Navigator Network is an initiative at KPMG designed to drive the adoption of AI tools and foster a culture of innovation and collaboration across the organization. This network consists of champions who are not only tech-savvy but also passionate about AI and committed to promoting its use within the company. The network includes representatives from various levels across 56 Member Firms, and they meet regularly to share knowledge and learn from each other.

The primary objectives of the Global AI Ninjas and Navigator Network are to:

• Educate colleagues across all roles and functions about AI.

• Enhance KPMG's digital presence and improve customer engagement.

• Drive ROI for the Global Member Firm network through upskilling, innovation, and adoption/enablement

KPMG aIQ Chat is a globally available Gen AI Chat tool that provides an expansive library of language models. It is hosted on KPMG’s secure Microsoft Azure platform, thereby protecting KPMG and client data. These language models can be used to help with various tasks, like creating engaging content, summarizing long documents, brainstorming ideas, and even translating natural language to code.

A one-stop-shop for Global AI info and news. It is the home for all things global AI including, use cases, credentials, learning pathways, and transformation assets.

Powered by Microsoft Azure, KPMG Digital Gateway is a single platform cloud-based solution that gives clients access to the full suite of KPMG Tax technologies. It provides organizations with direct access to our extensive investments in machine learning, data analytics, powerful visualizations and AI technologies all in one place.

KPMG Clara is an automated, agile, intelligent, and scalable platform. It empowers and enables our auditors by connecting data, information, and a growing ecosystem of advanced capabilities – through a single, streamlined interface. It is how we are fully digitizing the audit to drive global consistency and audit quality and bringing deeper actionable insights to our clients.

KPMG Velocity is an AI-enabled platform designed to integrate our thinking, methods, tools, and techniques to drive organizational change. It combines our industry insights, AI, data, ESG, Cyber transformation assets, and more with our leading digital transformation businesses: Connected, Powered, and Trusted.

Monthly active users.

Net Promotor Score.

The G20 AI Country Leads are responsible for the delivery of the AI Strategic Framework within their member firm. They lead the AI culture change and equip the firm with new AI capabilities.

A network of AI Change and Adoption leads from the G20 member firms. The C&A Lead Network drives the implementation of the AI change and adoption strategy and the successful activation of AI tools within their respective member firms.

An advanced AI-powered code completion tool designed for enterprise developers. It streamlines development, enhances collaboration, and empowers teams to build better software faster by automatically generating code snippets, functions, and entire classes based on natural language commands.

An AI-powered assistant integrated into Microsoft Power BI that transforms how organizations visualize and analyze data. It assists users in generating reports, creating tailored visualizations, and analyzing data more efficiently.

A platform that enables the customization of existing copilots or the creation of new ones tailored to unique business contexts. It empowers developers to create low-code/no-code, AI-powered, multi-lingual, multi-modality text and voice copilots that can converse with users on any business topic.

An end-to-end data analytics platform that unifies data engineering, warehousing, science, and business intelligence. It simplifies transforming raw data into insights by integrating technologies like Azure Data Factory, Azure Synapse Analytics, and Power BI into a single product.

Part of Microsoft Purview, a unified data governance solution. Purview DLP helps organizations protect sensitive information across various environments, while Purview Audit supports forensic and compliance investigations.

An AI-powered assistant integrated within SharePoint that enhances productivity and collaboration. It helps users create, edit, and manage content more efficiently by providing intelligent suggestions and automating repetitive tasks and makes it easier for teams to collaborate and share knowledge.

A tool to help build better work habits and behaviors by providing personalized, actionable recommendations around setting aside focused time, managing breaks, using personalized insights, sharing recognition, and leveraging AI technology to reduce manual workload.

Project to use AI to enhance the existing KPMG Enterprise Search capabilities, improving accuracy, simplifying the user experience and reducing costs to develop and maintain the service.

Global Management Team

Grouping of Critical Enabler activities that provide tools, processes, and methodologies that support the effective implementation and adoption of AI across the organization. The intent is to ensure that AI initiatives are seamlessly integrated into the daily operations and workflows of the firm, thereby maximizing their impact and value. (See also AI Accelerators, Infrastructure and Proof of Concepts (PoCs)).

A comprehensive initiative designed to ensure that AI systems are developed and deployed in a trustworthy and ethical manner. This program encompasses a set of frameworks, controls, processes, and tools aimed at accelerating AI adoption while aligning with the Trusted AI principles, global regulations, and the firm's goals.

An advanced AI-powered assistant integrated into the Microsoft 365 suite. It enhances productivity, streamlines workflows, and supports decision-making by combining the power of large language models with the data and tools within Microsoft 365 applications like Word, Excel, Outlook, and Teams.

Formerly known as Bing Copilot and Microsoft Copilot, this AI-powered partner maximizes productivity, boosts creativity, and keeps users informed. It offers a comprehensive suite of tools and services designed to empower users to achieve more, faster.

Work to harness opportunities available under KPMG’s strategic alliance with Microsoft to be an early access partner to preview/pilot, deploy and enable new technologies such as Microsoft 365 Copilot (as a flagship product), AI Accelerators and Azure OpenAI Service products. It also includes introduction of technologies and platforms that enhance our AI infrastructure, informed by proof of concept (POC) testing. (See also AI Accelerators, Infrastructure and Proof of Concept (PoCs) Functionality Summary).