Beyond LLMs: Exploring Six Use Cases that Can Benefit from AI’s Other Technologies and Approaches
By Eshwar Eswaran and Raman Pugalumperumal
As the world navigates the complexities of workforce, policy, demographic, and generational shifts, Artificial Intelligence (AI) is emerging as a transformative tool to address these challenges. While much of the focus recently has been on Large Language Models (LLMs), AI's potential extends far beyond these domains.
This blog highlights several impactful use cases from the world of job quality, education, and climate where AI can drive progress.
Use Case #1: Estimating job quality of emerging occupations and jobs
We need to determine the quality of jobs in emerging occupation clusters in the labor market even as factors such as schedule and work hours, benefits, safety commitments, others change or lack coverage.
Actors: Workforce leaders, employers, job seekers
Inputs: Job postings, Job quality, employee voice platforms, social sentiments, employer reports
Outputs: AI powered models to predict / estimate job quality for emerging occupations
Benefits: Provides consumers with near real-time information about job quality in emerging industries
Role of AI:
To estimate quality scores for newly created occupations using existing job category benchmarks.
To align new job types with established metrics based on similar skills, industries, responsibilities, and other key factors
Application of AI:
K-Means clustering to group similar jobs based on skills, industries, responsibilities, and other factors.
Hierarchical clustering to create structured job categories aligned with established metrics.
Resources to learn more:
WattTime: Uses clustering-based machine learning techniques to identify electricity grids with similar characteristics in regions lacking detailed data, enabling the estimation of marginal emissions rates for applications like carbon-aware load shifting and renewable energy siting.
OpenAQ’s AQAI: Hierarchical machine-learning model works by integrating sensor and satellite information to get a context-aware picture of pollution over your facilities.
JFF’s Quality Jobs Framework: Outlines standards and actionable steps for improving job quality—focusing on fair pay, benefits, career growth, and supportive environments.
Use Case #2: Overcoming challenges to accessibility, privacy, and diversity of jobs, credentials, and learner records training data
We are facing significant challenges in acquiring, managing, and utilizing real-world learner and worker data for analytics, machine learning (ML), and decision-making. Synthetic data generation involves creating artificial datasets that address the above challenges by providing scalable, privacy-preserving, and customizable datasets.
Actors: Product companies, Data scientists, Labor market researchers
Inputs: Statistical properties and patterns of real-world data
Outputs: Artificial datasets from synthetic data generation mimicking real-world statistical properties
Benefits: Enhance privacy, reduce costs, overcome scarcity, and improve model performance
Role of AI:
To validate product performance across diverse scenarios without relying on production data.
To generate synthetic learner records for AI-based diagnostics while preserving student confidentiality
Application of AI:
Generative Adversarial Networks (GANs) to generate synthetic student data that resembles real-world statistical properties of student records including diversity, coverage and scope.
Resources to learn more:
In machine learning, synthetic data can offer real performance improvements: MIT researchers demonstrated that machine learning models trained on synthetic video datasets can outperform those trained on real data for certain tasks, offering a scalable and ethical alternative for action recognition.
The Synthetic Data Vault: The SDV is an ecosystem of synthetic data models, benchmarks, and metrics, with standalone libraries for specific needs.
Meedan: SynDy to create synthetic datasets for training local classifiers to access accurate information and countering misinformation.
Use Case #3: Actionable Insights for Regional Climate Investments
We need insights to allocate the right funding, grants, and investment capital efficiently for regional / local economic development and job growth in the climate sector.
Actors: Policymakers, Investors, Grant makers
Inputs: Regional climate risk and readiness, green job creation data, labor market indicators
Outputs: Scenario analysis systems estimate impact and guide investment decisions.
Benefits: Funders and investors have a clear idea of the benefit and impact from every dollar invested in the region in support of sustainable jobs growth.
Role of AI:
To assess the economic impact of sustainable jobs adoption on local communities.
To identify regional economic risks and job readiness, helping policymakers prioritize funding.
To guide investors in funding high-impact brown job transitions across industries.
Application of AI:
Spatial Analysis to identify regions with high potential for green job creation.
Time-series forecasting (Prophet) to predict economic growth impacts from green job initiatives.
Predictive analytics to estimate impact of access to funds for green job adoption on local communities.
Resources to learn more:
Berkeley Earth's new High Resolution Data Set: The machine learning component of the new dataset extracts high-resolution weather patterns and uses them to improve estimates of small-scale structures in regions and time periods with limited data availability.
Earth Genome’s Earth Index: Leverages satellite imagery and geospatial AI foundation models to generate actionable environmental insights within a day, using image embeddings to streamline searching and classification workflows.
Use Case #4: Determine Workplace Carbon Footprint (CFP)
We need to determine carbon footprint data of a workplace across geographic locations. Not all CFP data is publicly available or the available data may be incomplete.
Actors: Employers, Regional leaders, policymakers
Inputs: Company disclosures, industry reports, government databases
Outputs: Enhanced workplace CFP datasets
Benefits: Provides actionable intelligence for workplace decarb investments
Role of AI:
To estimate missing CFP data based on industry standards, regional benchmarks, and company disclosures.
To collect real-time insights from job seekers and employees about workplace sustainability practices (e.g., energy use, commuting, resource efficiency).
To analyze satellite imagery and infer emissions from industrial zones and business locations.
Application of AI:
Regression models to estimate missing carbon footprint data.
Clustering to group similar companies or industries.
Computer vision to identify energy-efficient equipment and pollution sources in satellite imagery.
Image Classification to classify land use and identify areas with high carbon emissions.
Resources to learn more:
Open Climate Fix’s Forecasting solar photovoltaic (PV) power production: Machine learning models to forecast solar PV by predicting cloud movement and evolution using inputs like satellite images, weather predictions, cloud profiles, and geographical data.
Open Earth’s City Catalyst: A platform that helps cities manage climate data, starting with Greenhouse Gas Inventories, using AI, satellite, and IoT data to simplify integration, cut costs, and accelerate climate action.
SIRUM (Supporting Initiatives to Redistribute Unused Medicine): SIRUM implemented a computer vision model designed for pill counting and extracting information from the labels (preparation date, type of medication, others)
Use Case #5: Measuring and Quantifying Worker Voice
We need to be able to understand the real voice of employees and workers beyond traditional means such as surveys, focus groups, and others, to reduce bias and improve quality of collection.
Actors: Employers, Labor leaders, Workforce organizations
Inputs: Non traditional sources such as social media platforms, anonymous professional communities (Blind), and staff engagement platforms
Outputs: Insights into worker sentiments and real-time intelligence for policy scenario planning.
Benefits: Employers and labor leaders have improved understanding of employee perspectives to tailor support and interventions effectively.
Role of AI:
To understand the true voice and sentiment of workers and employees impacted by a policy or upcoming change
To understand and tailor support and change based on categories of opinions and reactions
Application of AI:
Sentiment Analysis to gauge the tone, opinion, and emotion of workers using NLP and ML techniques.
Text Classification to categorize personas and sentiment clusters.
Resources to learn more:
Giga Fact’s Parser AI: A civic platform for news professionals uses AI to transcribe and analyze audio and video recordings of public officials' comments, revealing their claims and talking points.
Polaris Project: Search Trafficking Hotline case notes to determine accuracy of social media narratives warning about abduction-based recruitment.
AImpower: Collaborate with marginalized communities to design speech-based AI products, such as co-developing stutter-friendly videoconferencing tools and creating the first large-scale Mandarin stuttered speech dataset to improve AI systems and empower users.
Use Case #6: Advanced Credential Discovery via Language-Based Insights
We need to be able to make complex queries and discoveries of credentials and courses including quality, gains, growth, etc matching specific needs and use cases
Actors: Job Seekers, Career Counselors, Students and Learners
Inputs: Credential repositories, Quality information and frameworks
Outputs: An interface platform enabling natural language-based queries for credential discovery.
Benefits: Career counselors and students will have a powerful tool in their hand to run discoveries based on their specific scenarios and not be restricted by traditional search and filter tools.
Role of AI:
To offer advanced analysis and insights through natural language queries
Application of AI:
AI powered chat and query bot that interfaces with the credential repository to provide advanced natural language-based analysis and insights discovery.
Resources to learn more:
Jacaranda Health PROMPTS: AI-enabled digital health service that, through two-way SMS exchange, empowers women to seek care at the right time and place.
Khan Academy’s Khanmigo: AI-powered personal tutor and teaching assistant that offers engaging, on-topic, and effective learning for students.
AI is reshaping the landscape of the workforce and labor market by offering scalable solutions across diverse domains. As we embrace these technologies, several questions arise: How can we ensure ethical use of AI in equitable decision-making? What steps should be taken to democratize access to these tools globally? And how can industries collaborate more effectively to maximize AI's potential for social change?
Looking forward to learning and exploring these possibilities together—because innovation is key to building a sustainable future.
References
McGovern’s AI Use case Library
Data.Org’s AI2AI Challenge Awardees
Glossary of AI Terms Used
-
Statistical techniques used to predict outcomes based on relationships between variables (e.g., estimating missing carbon footprint data).
-
Machine learning methods that group data points into clusters based on similarities (e.g., grouping jobs or industries).
-
A field of AI that enables machines to interpret visual data from images or videos (e.g., analyzing satellite imagery).
-
Natural Language Processing (NLP) technique used to determine opinions or emotions expressed in text (e.g., gauging worker sentiments).
-
Categorizing text into predefined groups using machine learning techniques (e.g., identifying personas or sentiment clusters).
-
Geospatial techniques used to analyze geographic patterns or trends (e.g., identifying regions for green job creation).
-
Predictive models used for analyzing trends over time (e.g., forecasting sustainability score trends).
-
Interfaces that allow users to ask questions in plain language while receiving relevant insights from datasets (e.g., credential discovery via chatbots).
-
GANs consist of two neural networks—a generator that creates synthetic data and a discriminator that evaluates its authenticity. This adversarial process drives the generator to produce increasingly realistic outputs.
Eshwar Eswaran is a seasoned technology and data leader with extensive experience in product incubation, technology modernization, and data strategy. As the Director and Head of Product Incubation at Jobs for the Future (JFF), Eshwar focuses on leveraging data and product driven solutions to drive equitable economic advancement. He leads innovative initiatives that integrate cutting-edge technologies into scalable products, fostering impactful change across industries.
Raman Pugalumperumal works at the World Bank Group as a lead for AI and ML platforms. He is responsible for developing and implementing advanced AI systems, including enterprise-ready Retrieval Augmented Generation (RAG) systems, intelligent process automation solutions, and custom AI models. His work focuses on leveraging AI technologies like Azure OpenAI, Google Vertex AI, and open-source platforms to enhance productivity and efficiency within the organization
