Transform how you govern with data and AI

Use this AI agent to unlock value from untapped government data. With computer vision, machine learning and text analytics, intelligent document processing extracts contextual information from scanned document images for downstream decisioning and analytics.

The value of this solution:

• Greater productivity.
• Higher cost savings.
• Reduced complexity.

AI techniques used in this solution:

• Computer vision models.
• Machine learning models.

How AI helps:

• Modernizes government filing and information systems.
• Significantly reduces the hours required to research and find information.
• Expands the amount of useful information available for insights.
• Positions an organization to be proactive in accomplishing its mission, whether its mission is citizen services, investigations or operations.

The AI models provide:

• The extraction of key information from images or scanned documents is automated by this AI model into a structured format, enabling downstream analytics and decisioning.
• Supplement any current OCR/RPA processes to significantly improve the accuracy and quality of the information extraction, especially with tougher forms like blurry documents, forms with checkboxes or handwriting.
• At this point, data can be used for investigations, alerts can be generated, extracted entities can be made searchable or the newly structured data can feed existing processes.

Answer powerful questions with a specialized copilot that digs through the piles of intelligence data that has been pre-filtered with analytics and AI to pull out the most relevant information. With large language models (LLMs) and investigation analytics, the task of identifying valuable information from intelligence becomes more targeted, faster and easier.

The value of this solution:

• Greater productivity.
• Improved public safety.
• Trustworthy insights.

AI techniques used in this solution:

A GenAI copilot consumes information pre-filtered through an analytics platform to further pinpoint elements of interest and surface opportunities to collect or consider additional information.

How AI helps:

• Improves the investigative process by going beyond summarizing information to highlight potential leads, identify missing data and make associations between data.
• Enhances the usage of an analytics platform that reconciles information and maps entities from various records and data sources.
• Identifies gaps in the investigative process by generating narratives and describing key details using organizational best practices in an explainable way.
• Surfaces information for an investigator while maintaining his or her role as the interpreter of facts.

The AI models provide:

The LLM on top of an analytics platform:

• Sits on top of an investigations system to tease out information that has been collected, resolved, mapped and related.
• Coordinates queries and responses between the search engine and the LLM, which then uses the new knowledge and its training data to create better responses.
• Provides responses that are explainable, auditable and accountable, which point to the specific citations and documentation that guided it.

Identify social benefit cases at high risk for error to support quality assurance, triage overpayment referrals and automate benefit recovery. SAS solutions support the Supplemental Nutrition Assistance Program (SNAP) and Temporary Assistance for Needy Families (TANF), US-based programs that help low-income families.

The value of this solution:

• Faster decision making.
• Better outcomes.
• Stronger fraud detection and prevention or faster issue resolution.
• Improved customer service.

AI techniques used in this solution:

• Multi-variate regression by machine learning to identify and correlate key input variables that affect SNAP eligibility case errors.
• AI modeling for identifying appropriate peer groups.
• Machine learning decision trees for risk scoring and ranking high-risk-for-error cases.
• AI modeling for fraud referral triage by risk scoring for potential error, overpayment and possible fraud versus case processing errors.

How AI helps:

• Gives SNAP quality assurance teams the ability to assess all cases without sampling to identify emerging patterns leading to improper payments and correct the error before more issuances are released.
• Gives SNAP benefit recovery teams the ability to effectively triage overpayment referrals to increase effectiveness and ROI.
• Gives SNAP benefit recovery teams the ability to automate overpayment determinations to maximize efficiency.

The AI models provide:

• Risk-scoring for all active SNAP cases to spot high-risk case errors faster.
• Risk-scoring of overpayment referrals to more effectively triage high-value work.
• SNAP case peer grouping, with identification and correlation of key input variables.

Use natural language processing, text analytics and an LLM to categorize, synthesize and summarize large volumes of written feedback. This approach identifies themes, groups feedback and generates a concise summary of key points, making the process of handling written feedback easier, faster and more accurate.

The value of this solution:

• Greater productivity.
• Trustworthy insights.
• Reduced complexity.

AI techniques used in this solution:

• Natural language processing (NLP).
• LLMs.

How AI helps:

Combining NLP and text analytics with an LLM can help by:

• Avoiding hallucinations: Pre-filters relevant data, ensuring accurate outputs.
• Enhancing time to value: Uses smaller LLMs like Llama 2 by processing less data.
• Reducing costs: Cuts down data sent to LLMs, minimizing API calls and computational resources.
• Ensuring privacy and security: Uses local vector databases for fine-tuning and protecting sensitive data.
• Supporting verification: Enables traceability of LLM outputs, enhancing transparency and trust.

The AI models provide:

The combination of NLP and LLM:

• Read tens of thousands of pieces of feedback.
• Identifies recurring themes among commentary.
• Identifies sentiment, such as negative reactions.
• Compiles recommendations.
• Summarizes the commentary by recurrent, similar or customer-defined themes.

Eliminate the guesswork in flood prediction and preparation using machine learning algorithms and synthetic data for digital twins. Developed with historical data, this AI-powered model can be deployed even where hyper-localized data is scarce, effectively filling in data gaps.

The value of this solution:

• Improved public safety.
• Better outcomes.
• More accurate forecasting.

AI techniques used in this solution:

Machine learning models create an early warning detection system, and digital twin technology supported by synthetic data enables a forecasted flood inundation model.

How AI helps:

• Protect: Improve citizen safety and emergency response services with real-time visibility and flood predictions.
• Respond: Reduce the impact of flooding incidents on property and infrastructure by automating and streamlining response.
• Improve: Enhance emergency planning with improved situational awareness and historical insights.

The AI models provide:

• Real-time insights and situational awareness, including current condition monitoring with automated alerting.
• Future flooding predictions with automated altering.
• Historical and forensic analysis.
• Forecasted flood inundation modeling.
• Simulations for emergency planners, improving modeling for various disasters.

Increase transparency and trust between the tax agencies and citizens with the Non-Invasive Compliance and Enforcement (NICE) solution. This customer-facing system analyzes real-time input from taxpayers when they file their declarations, comparing it with data available to the tax agency. The analysis identifies potential educational prompts based on applicable laws and policies, providing recommendations to taxpayers. This real-time feedback enhances voluntary compliance and scores and accounts for potential fraud risk, giving taxpayers the opportunity to correct their returns without additional enforcement.

The value of this solution:

• Maximized operational efficiency.
• Greater customer engagement.
• Improved regulatory compliance.

AI techniques used in this solution:

• Machine learning models, including supervised and unsupervised machine learning models.
• Taxpayer segmentation.
• Entity resolution.
• Predictive analytics and prescriptive analytics.

How AI helps:

• Gain comprehensive insights from siloed income tax, value-added tax, customs tax data and data from international partners and institutions.
• Enhance compliance.
• Enforce regulations.
• Optimize resources.
• Build respect between taxpayers and the tax authorities.

The AI models provide:

• A fully integrated analytical system that provides an extensive, in-depth view of tax compliance and automatically identifies potential fraud risks and possible errors in declaration filings.
• Comprehensive analysis of available data, including third-party data from domestic and international partners. The system automatically identifies and flags potential compliance discrepancies.
• A holistic view of a taxpayer's history and appropriate tax education based on their taxable activities, predictive analytics to ensure optimization of compliance efforts by the agency and comprehensive fraud risk scoring on an ongoing basis.

Reassess the value of residential property using property and sales data every day with greater speed, ease and accuracy than the traditional method. The system provides the true value of every factor considered for a property, in accordance with other factors present in the property.

The value of this solution:

• Trustworthy, real-time insights.
• Increased revenue.
• Reduced complexity.

AI techniques used in this solution:

• Advanced regression tree analysis by machine learning.

How AI helps:

• Strengthen government decision making with an integrated, real-time view of data.
• Optimize resources and improve government effectiveness.
• Increase public trust while maintaining data privacy and ensuring AI transparency.
• Be better prepared to quickly respond in times of disruption and uncertainty.

The AI models provide:

• Daily reconsideration of every sale. Thousands of decision trees are run daily using data from property sales. From these decision trees, the algorithms extract the value of every factor for every property.
• Factor ranking by importance. The algorithms identify the factors that are most influential in the value of properties.
• Reappraisal of all properties. The machine-learning algorithm calculates the value of every property in the community. This allows the assessor’s office to understand market trends for homes, neighborhoods or the entire community.

Identify and flag anomalous or manipulated data feeds to enhance data truthfulness and trust. In national security, our solution approaches data skeptically because adversaries often actively alter data with the intention to deceive and mislead.

The value of this solution:

• Data governance.
• Improved safety.
• Trustworthy insights.

AI techniques used in this solution:

The solution uses a layered analytic approach embedded in SAS Intelligent Decisioning, employing multiple AI and machine learning techniques to detect and alert malicious data injections. The hybrid approach includes:

• Predictive modeling: Uses neural networks, decision trees and other models to uncover new patterns from legacy data.
• Text analytics: Extracts meaningful information from vast unstructured text data.
• Anomaly detection: Applies techniques like regression, clustering and sequence analysis to identify abnormal patterns.
• Automated business rules: Filter data feeds based on sophisticated rules reflecting suspicious patterns.
• Entity-based network analysis: Identifies links between data sources and anomalies.
• Combined learning models: Use supervised, semi-supervised and unsupervised learning models together, making it harder for adversaries to inject malicious data successfully.

How AI helps:

• Monitoring the data feeds for critical infrastructure management is essential. It enhances resiliency by preventing system compromises. Critical Infrastructure is a known target for adversarial nation-states.
• Early detection enhances customer satisfaction and mission effectiveness. Anomalous data can indicate malicious intent, system instability or upcoming maintenance issues.
• Constantly monitoring data veracity and implementing improvement initiatives can lead to cost savings. Data-driven analysis helps identify the root causes of system inefficiencies, enabling quick and effective improvements.

The AI models provide:

• Constant data veracity monitoring.
• Machine learning algorithms make it easier to identify anomalous or malicious data feeds, ensuring data quality and truthfulness.
• Rapid ingestion and curation of new data sources for better mission execution and efficient system security analysis using the SAS DataOps Process.
• Real-time data veracity scorecards to detect malicious data injections and other issues, supporting downstream systems.
• Continuous performance monitoring and quick updates or replacements of models as conditions change and adversaries adapt, using the SAS ModelOps Process.