02 / OPEN QUESTIONS
Industry challenges AI has not fully solved
Browse by industry, then explore by difficultyNot untouched—just not solved yet
The hardest problems
often live in reality.
Explore 30 concrete challenges across industries. Each topic opens into a field guide. AI is already involved; the focus here is what still makes reliable, accessible, and scalable adoption difficult.
30 topics
Healthcare
3 topics · Easier to Frontier01Turn clinical records into information doctors can actually useOrganize chart summaries, follow-up notes, and test results to reduce duplicate entry while keeping original sources available for clinicians to verify.↗02Make chronic-care advice fit patients? real livesConnect ongoing monitoring, visit records, and patient feedback in an actionable care plan rather than generating health advice that merely looks complete.↗03Build clinical decision support that can be validated across hospitalsEnable support systems to explain their basis, surface uncertainty, and defer decisions when evidence is insufficient across different hospitals, devices, and patient populations.↗
Education
3 topics · Easier to Frontier01Turn learning materials into formats that work for diverse learnersCreate easy-to-read text, audio, captions, and tiered exercises for the same concept, while helping teachers quickly check that facts and learning objectives remain aligned.↗02Move beyond right-or-wrong answers to effective personalized practiceUse students? mastery and error patterns to choose next exercises, focusing on transferable understanding rather than doing students? work for them.↗03Bring reliable AI support to classrooms using under-resourced languagesProvide teachers and students with explanations in local languages, bilingual materials, and classroom assistance while preserving community control over language data and cultural content.↗
Agriculture and food
3 topics · Easier to Frontier01Make crop and pest diagnosis work in the field, not just in demo photosUse phone images and brief questions to help farmers identify common problems and suggest next checks validated by local agricultural advisers.↗02Turn weather, soil, and crop conditions into water-saving irrigation decisionsCombine weather forecasts, soil moisture, and crop growth stages to help growers schedule irrigation instead of relying on a single sensor and a fixed threshold.↗03Help harvesting robots handle varied ripeness and delicate fruitDevelop robots that identify ripe fruit, plan movements, and harvest safely in open fields despite foliage, changing light, and differences between individual fruits.↗
Manufacturing
3 topics · Easier to Frontier01Build reviewable visual quality inspection with few defect samplesFor a specific product on one production line, detect scratches, dimensional deviations, or missing assembly steps and send suspicious samples to inspectors for confirmation.↗02Move equipment maintenance from fault alerts to actionable advance planningCombine equipment sensors, maintenance work orders, and production plans to predict potential failure windows and recommend suitable inspections or maintenance actions.↗03Help small-batch production lines adapt quickly to new products and processesExplore flexible systems that can quickly understand new workpieces, tools, and production constraints, reducing the need for manual reprogramming in low-volume, high-mix manufacturing.↗
Energy
3 topics · Easier to Frontier01Make building energy forecasts serve comfort as well as efficiencyUse historical load, weather, and building usage to help managers spot energy anomalies and plan low-risk efficiency adjustments.↗02Improve local accuracy in distributed renewable generation forecastsCombine weather forecasts, terrain, and equipment conditions to provide local grids with timely interval forecasts of wind and solar output.↗03Keep the grid safe and restore service quickly during extreme weatherExplore decision support that combines weather, equipment, load, and field reports to help dispatch and repair teams identify risks and recovery options faster.↗
Ecology and environment
3 topics · Easier to Frontier01Turn camera-trap data into ecological monitoring leads fasterAutomatically filter blank camera-trap images and identify species clues and timestamps, so researchers can focus on verification and ecological interpretation.↗02Connect scattered monitoring data into traceable pollution leadsCombine sensor, remote-sensing, weather, and field-sampling data to identify possible sources of pollution changes and help regulators and remediation teams plan follow-up sampling.↗03Make ecological restoration plans account for long-term change and community needsHelp compare the long-term effects of restoration measures on water, soil, biodiversity, and local livelihoods instead of optimizing only short-term or single environmental indicators.↗
Transportation and logistics
3 topics · Easier to Frontier01Optimize picking and replenishment sequences for small and midsize warehousesUse orders, inventory locations, and worker routes to offer adjustable picking suggestions that reduce walking and stockouts without changing existing warehouse safety rules.↗02Make city delivery plans adapt to changes and fairness constraintsDynamically reschedule deliveries as orders, traffic, weather, and service windows change, while accounting for driver safety, service quality, and access in different areas.↗03Make autonomous driving predictable in rare road eventsHandle long-tail situations such as construction detours, occluded pedestrians, unusual vehicles, and negotiations among road users, so autonomous systems can slow safely or request takeover when uncertain.↗
Government and legal services
3 topics · Easier to Frontier01Make service guides fit individual circumstances and local rulesOrganize public policies and procedures into conditional guidance that explains required documents, where to apply, and when information was updated, with links to the original sources.↗02Help the public understand legal documents without replacing a lawyer?s judgmentTurn contracts, notices, and procedural explanations into plain-language summaries, flag deadlines and questions to verify, and clearly distinguish explanation from legal advice.↗03Make public-sector AI decisions explainable, appealable, and auditableEstablish transparent decision-support processes for areas such as public resource allocation and risk prioritization, so affected people can understand the rules and challenge decisions.↗
Construction and disaster response
3 topics · Easier to Frontier01Use site footage to help spot construction safety hazardsFlag visible risks in construction-site footage, such as missing protective equipment or entry into hazardous areas, to help safety officers inspect faster without replacing those responsible on site.↗02Identify project delays and supply-chain risks earlierCombine progress records, material deliveries, weather, and site issues to flag risks that may affect the critical path and help project teams adjust resources in advance.↗03Turn diverse information at disaster sites into trustworthy leads for actionCombine sensor, satellite, public-report, and rescue-team information to help incident commanders build a situation picture and prioritize tasks, clearly showing sources and uncertainty.↗
Scientific research
3 topics · Easier to Frontier01Find reliable evidence in vast research literature, not just similar wordingHelp researchers map topics, methods, and disagreements, with precise links back to source passages, without mistaking generated summaries for systematic reviews.↗02Make data analysis reproducible, not just produce polished chartsUse AI to help clean data, write analysis code, and check statistical assumptions while retaining full records of the environment, parameters, and human edits.↗03Build an autonomous science loop from hypothesis to reproducible experimentExplore whether AI can propose falsifiable hypotheses, design executable experiments, operate instruments, and revise plans after failure, rather than merely tell plausible-sounding research stories.↗
No topics in this industry.
Difficulty reflects data availability, real-world complexity, validation time, and the cost of errors. It is a guide for exploration, not a ranking of industry value or technical maturity.