The Seven Intelligences of AI
A Framework for Understanding AI Capabilities in the Public Sector
As artificial intelligence becomes a central component of digital transformation, many public sector teams struggle to distinguish what AI can realistically deliver today. Agencies understand the potential, but they often lack a practical vocabulary for framing capabilities, use cases, and adoption paths.
Spruce proposes a simple and operational framework called The Seven Intelligences of AI. It organizes applied AI into seven capability categories that reflect real and emerging government use cases. The goal is to help leaders evaluate how AI can improve service delivery, decision support, and operational efficiency in ways that are measurable and aligned with mission needs.
1. Conversational Intelligence
Conversational intelligence refers to AI systems that understand and generate natural language in interactive, context-aware dialogue. These systems power virtual assistants, citizen self-service portals, employee helpdesks, and inquiry triage tools. High-performing models handle multilingual inputs, summarize context, escalate when necessary, and provide transparent citations or references when appropriate.
Example
New York City’s MyCity Chatbot supports small businesses by guiding users through complex licensing requirements and regulations in multiple languages. It reduces wait times, improves access for non-native English speakers, and shifts call center load toward higher-value inquiries.
Citation: NYC MyCity Chatbot
2. Visual Intelligence
Visual intelligence covers the family of computer vision capabilities that extract information from images and video. This includes object recognition, anomaly detection, scene interpretation, geospatial tagging, license plate analysis, and automated inspection. These tools reduce reliance on manual review and enable proactive detection and monitoring.
Example
The U.S. Department of Transportation conducted a pedestrian detection pilot that used AI with existing traffic cameras to identify unsafe crossings and near misses. The pilot provided insights for engineering interventions without requiring new hardware deployments.
Citation: USDOT ITS JPO, “Leveraging Existing Infrastructure and Computer Vision for Pedestrian Safety.”
3. Retrieval Intelligence
Retrieval intelligence allows AI systems to search and retrieve content using semantic meaning rather than keywords. Retrieval-augmented systems unify content from many sources, including PDFs, knowledgebases, email archives, policies, and legacy applications. This category often underpins applied AI in regulated environments because it provides verifiable, source-grounded responses.
Example
Northwestern University built a multilingual semantic search tool using Amazon Kendra and Amazon SageMaker. Students use natural language queries to find information across policy documents, administrative pages, and knowledge repositories in multiple languages.
4. Generative Intelligence
Generative intelligence refers to AI that produces functional output from natural language instructions. This includes generating code, workflows, rules, queries, configuration files, and other structured logic needed to build or modify digital systems. These capabilities reduce technical barriers and enable teams to create or update processes without extensive programming expertise.
Example
Researchers at the Massachusetts Institute of Technology developed a tool called SCIGEN that guides generative AI models to propose new materials with exotic quantum properties. By constraining model outputs to specific geometric lattice structures, the system successfully generated two novel material candidates that were subsequently synthesized in the lab. The approach demonstrates how generative AI can reliably produce domain-specific, technically valid outputs.
Citation: MIT News, “New tool makes generative AI models more likely to create breakthrough materials.”
5. Insight Intelligence
Insight intelligence refers to AI models that interpret, summarize, categorize, compare, or transform large volumes of content in order to surface meaning. These systems handle legislative text, public comments, meeting transcripts, historical records, case files, and research materials. They reduce cognitive load on analysts and support evidence-based decision making by distilling many inputs into more actionable insight.
Example
The U.S. Government Accountability Office documents use cases where natural language processing tools help agencies categorize and summarize thousands of public comments submitted during federal rulemaking. These systems reduce manual review time and help agencies identify themes and potential impacts with greater accuracy.
Citation: GAO AI Use Case Library, “NLP for Summarizing Public Comments.”
6. Prediction Intelligence
Prediction intelligence identifies patterns in historical and real-time data and generates forecasts or risk scores. It informs planning, fraud detection, resource allocation, risk assessments, and public safety or mobility decisions. These models typically combine machine learning with domain expertise and structured datasets.
Example
The city of Tampere in Finland operates an AI-driven predictive analytics service called Tampere Pulse that forecasts visitor flows in the city centre. The system uses data from the city’s Internet of Things platform including traffic cameras with machine vision, weather conditions, and event schedules to produce a month-long forecast of visitor volumes. With prediction accuracy above 79 percent for the exact visitor flow category and more than 97 percent for the correct or adjacent category up to three weeks ahead, businesses can optimize staffing and inventory, and city teams can better plan maintenance and security resources.
Citation: OECD OPSI, “Tampere Pulse: AI-powered visitor flow forecasting in Finland.”
7. Execution Intelligence
Execution intelligence involves AI systems that not only understand or predict but also perform tasks. These capabilities often combine AI with robotic process automation, autonomous agents, or physical robotics. They support automated scheduling, form completion, data integration, case movement, and other high-volume operational activities.
Example
Omega Healthcare Management Services, which provides revenue-cycle management services to more than 350 healthcare organizations, partnered with UiPath to automate document-heavy processes such as medical billing and insurance claims. Using AI-powered document understanding and robotic process automation, Omega processes over 100 million transactions, saves more than 15,000 employee hours per month, cuts documentation time by about 40 percent, and reduces turnaround time by roughly 50 percent while maintaining approximately 99.5 percent accuracy.
Conclusion
The Seven Intelligences framework provides a common language for discussing AI readiness and capability adoption. It helps agency leaders understand which categories align with their mission, where to begin, and how to prioritize investments.
Spruce will continue refining this framework with visual materials and additional sector-specific examples. It will also be integrated into AI readiness assessments, capability roadmaps, and organizational adoption planning.