AI and Automation Glossary
Table of Contents
This glossary provides definitions for key terms related to Artificial Intelligence (AI) and Machine Learning (ML), designed to help you understand the concepts and terminology used in our AI-powered and automation features and services.
Term |
Definition |
| Aggregation (or aggregated data) | Data where multiple individuals (N>10) is combined into a summary format (for example: 75% of users complete more than 10 tasks a month). |
| AI Content | Collectively, the Inputs and Outputs (see these definitions below) |
| AI Features | The products, features, or tools made available to Subscriber as part of the Services that are powered by artificial intelligence, machine learning, or similar technologies. |
| AI Service Providers | Any third-party that Sprout Social relies on to provide certain AI Features (whether customer has entered into separate terms and conditions with such third-party or the third-party is considered a subprocessor of Sprout). |
| AI system | A machine-based system that uses varying levels of autonomy to infer outputs based on provided inputs. |
| Anonymization (or anonymized data) | Data where all personally identifiable information (like name or email address) is permanently and irreversibly removed to prevent the reidentification of the individual, even with additional effort or external data sources. |
| Classification AI | AI systems that analyze data and assign it to predefined categories or classes, which may or may not incorporate the use of LLMs. This includes our features that classify messages based on priority or positive/negative sentiment. |
| Clickstream data | Data related to events that are generated by Sprout Social users’ use of the platform (e.g., “User 123 completed a message in the Inbox on January 1, 2025 at 1:30pm CT”). |
| De-identification (or de-identified data) | Data where personally identifiable information is removed, but the data may still be identifiable through additional effort or external data sources. |
| Deployer | According to the EU AI Act, “a natural or legal person, public authority, agency or other body using an AI system under its authority except where the AI system is used in the course of a personal non-professional activity”. |
| Derived data | Data that is created or generated by extracting, transforming, processing, and combining existing data sets through algorithms, statistical methods, or business logic. |
|
First party AI (aka “in-house AI models”) |
Models created, trained, and/or fine-tuned by Sprout Social and hosted within Sprout’s infrastructure. This includes the use of AWS Bedrock for hosting local versions of pre-trained models such as Claude. |
| Fine-tuning | The process of taking a pre-trained model and iterating on it with smaller, task-specific data sets to apply the model to a specific use case. |
| Generative AI | AI systems that generate new content (text, images, etc.) by learning patterns from training data. This includes our features to suggest responses in customer use cases, create captions for images, suggest new output posts or content, and generate summaries of conversations or reports. |
| General-purpose AI model | According to the EU AI Act, “an AI model that displays significant generality and is capable of competently performing a wide range of distinct tasks”. |
| High risk AI systems | According to the EU AI Act, an AI system that either (i) a product or safety component of product is already regulated under product safety laws, or (ii) used in high impact areas such as critical infrastructure, biometric and emotion recognition, education and vocational training, employment, essential private and public services, law enforcement, migration, or administration of justice or elections. |
| Inputs | Any Subscriber Data, content or materials that Subscriber submits to the AI Features such as an audio file, video file, document, image, or text to receive the Output. |
| Internal business use models | First party AI models that are used internally by Sprout to help us understand customer fit, churn risk, features most heavily used in the platform, etc. (go-to-market analysis). |
| Machine learning (ML) | Machine learning is a sub-field of AI. Using machine learning, computers can make predictions based on past behavior and “learn” from new information. |
| Model cards | Documentation that provides details on how a model was developed and how it performs; model cards may include information about the intended use, excluded uses, ethical considerations, training and evaluation data, and metrics. |
| Natural language processing (NLP) | Natural language processing is a sub-field of AI that focuses on the recognition and analysis of speech and text. NLP enables computers to parse context and generate meaningful text. |
| Open-source model | A model that is available for public use, training, and fine-tuning under open source licenses. |
| Outputs | The resulting image, text, reports, summaries, insights, or any other content, which is generated by the AI Features based on the Inputs and provided to Subscriber within the Services. |
| Provider | According to the EU AI Act, “a natural or legal person, public authority, agency or other body that develops an AI system or a general-purpose AI model or that has an AI system or a general-purpose AI model developed and places it on the market or puts the AI system into service under its own name or trademark, whether for payment or free of charge”. |
| Pseudonymization (or pseudonymized data) | Data where personally identifiable information (like name or email address) is replaced with a placeholder or pseudonym (e.g., “janedoe@sproutsocial.com viewed the pricing page” is replaced with “User XAF156 viewed the pricing page”) that conceivably could be reversed with additional effort or by matching up with external data sources. |
|
Third party AI (aka “third-party AI models”) |
Models created by third parties and hosted outside Sprout’s infrastructure. This includes our use of OpenAI for the AI Assist features. |
| Training | The process of teaching a model to make predictions by providing it with training data sets and adjusting the model’s parameters. |
| Training data | Data sets used to teach a model. |
Was this article helpful?
