← Back to the homepage
BIGGER PICTURE 99 AI stories

NASA parts shaped like bones.
Burnt scrolls read unopened.

Explore all 99 applications
Three metal telescope supports with branching, curved struts and bolt holes on a workbench.
NASA / Henry DennisNASA media guidelinesWhy this shape?
How this collection was selected

We selected 99 distinct applications with a traceable benefit for people or a demonstrated step towards one. Each links to an original study, a report by an involved institution, or an explicitly labelled provider report. The order starts with particularly tangible examples; it is not a ranking.

“In use” describes use in the documented project, not independent proof of effectiveness or a guarantee of current operation. Study evaluations and research are labelled separately. Better predictions, higher detection rates and saved working time do not automatically mean lives saved or a net social benefit. Each case explains its specific limit. This selection is not a statistical comparison with incidents in the register.

Sources reviewed: 14 November 2026

How did they do it?

Every example freely accessible

13–24 of 47 matches · 99 applications in total

AI-generated illustration: Neighbouring traffic signals coordinating vehicle flows.
One junction sees cars approaching. The next gets a head start.
Public & everyday servicesTested in a study

Traffic lights that tell the next junction what is coming

Discover the story
Result & evidence

SURTRAC coordinates traffic signals using real-time traffic information and AI scheduling. Drivers can move through connected intersections with fewer interruptions, reducing the time people lose in traffic during everyday journeys.

Documented result

A nine-intersection Pittsburgh pilot measured 25.79% shorter travel times and 40.64% less waiting than the previous signal control.

Carnegie Mellon University
Pittsburgh, Pennsylvania, United States 2013

A before-and-after pilot from 2012, not a randomized citywide result; performance depends on local traffic.

Reviewed 2026-11-14
AI-generated illustration: Conceptual data-centre cutaway showing racks, cooling pipes and controlled airflow.
The energy saving hides in how pumps and cooling towers work together.
Energy & climateTested in a study

Why running more cooling equipment can sometimes use less energy

Discover the story
Result & evidence

Machine learning recommends cooling settings using thousands of sensor readings, helping operators keep servers within operating limits while reducing the electricity required by pumps, fans and other cooling equipment.

Documented result

Google’s 2016 live-site test reported up to 40% lower cooling-energy use after its AI recommendations were applied.

Google; Google DeepMind
Google data-centre test site; location not disclosed 2016

The figure concerns cooling, not all data-centre electricity, and comes from the system’s developer.

Reviewed 2026-11-14
AI-generated illustration: An ECG waveform becomes a clue for further heart-function assessment.
Can the heart’s electrical trace reveal a reason to check its pumping power?
Earlier diagnosisTested in a study

A routine ECG offered a clue to heart weakness nobody had diagnosed

Discover the story
Result & evidence

Mayo Clinic's EAGLE system used routine ECGs to flag possible weak heart pumping, giving primary-care teams a practical reason to arrange a confirmatory heart ultrasound.

Documented result

Among 22,641 adults, AI alerts produced five additional low-ejection-fraction diagnoses per 1,000 patients compared with usual care.

Mayo Clinic
Minnesota and Wisconsin, United States 2021

The ECG flag required ultrasound confirmation; the trial established improved detection rather than longer survival.

Reviewed 2026-11-14
AI-generated illustration: A learner studies a pendulum beside a laptop and an open physics notebook.
Sometimes the useful answer is another question—at exactly the right step.
Learning & workTested in a study

The physics tutor that waits for your next step

Discover the story
Result & evidence

A purpose-built AI tutor guided university students through physics problems using structured teaching practices. Learners could ask questions and work at their own pace rather than depend solely on shared classroom timing.

Documented result

In the Harvard experiment, students achieved more than twice the immediate learning gain of the active-learning comparison, in less time.

Harvard University
Cambridge, Massachusetts, United States 2025

Two lessons in one university course; immediate test gains do not establish long-term retention.

Reviewed 2026-11-14
AI-generated illustration: Conceptual greenhouse cutaway with cucumber vines, sensors and irrigation.
Light, heat and water became decisions a growing team could send across an ocean.
Nature & foodTested in a study

Growing Dutch cucumbers from an office in Canada

Discover the story
Result & evidence

AI systems use greenhouse sensors to adjust climate and fertigation, demonstrating how growers can manage crop conditions remotely when farms expand and skilled crop-management staff are difficult to recruit.

Documented result

In Wageningen’s 2018 cucumber challenge, the winning AI-controlled compartment outperformed the grower-operated reference in production; the study was published in 2020.

Wageningen University & Research; Autonomous Greenhouse Challenge teams
Netherlands 2020

The winner used more lighting and sacrificed some resource efficiency; pest control still depended on people.

Reviewed 2026-11-14
AI-generated illustration: A highlighted fold illustrates how computer assistance can draw an endoscopist’s attention.
A fleeting fold on a moving image can be the detail that matters.
Earlier diagnosisTested in a study

A green box that helped doctors spot more polyps

Discover the story
Result & evidence

During colonoscopy, GI Genius highlighted suspicious areas in real time, helping endoscopists find more adenomas that could then be assessed and removed during the same examination.

Documented result

In COLO-DETECT, 56.6% of AI-assisted examinations found an adenoma versus 48.4% with standard colonoscopy across 2,032 participants.

Newcastle University; South Tyneside and Sunderland NHS Foundation Trust
England, United Kingdom 2024

The trial found more adenomas, but not more advanced adenomas or cancers; cancer prevention was not directly measured.

Reviewed 2026-11-14
AI-generated illustration: Neural signals guide a robotic arm toward an everyday object.
Reaching for a cup starts with a brain signal. The challenge is keeping it reliable.
Accessibility & independenceTested in a study

He opened a cabinet with a robotic arm controlled by brain signals

Discover the story
Result & evidence

An AI-supported brain-computer interface let a man paralyzed after a stroke control a robotic arm, turning intended movements into practical actions such as retrieving and positioning a cup.

Documented result

The participant moved blocks, opened a cabinet and brought a cup to a dispenser; the system supported use over several months.

University of California, San Francisco
San Francisco, United States 2025

These were supervised research tasks with implanted sensors; independent home use was still a planned next step.

Reviewed 2026-11-14
AI-generated illustration: A phone reminder connecting everyday routines with the classroom.
Before a student falls behind, a message can point to help.
Learning & workTested in a study

A text message that leads back to the classroom

Discover the story
Result & evidence

Georgia State used an academic chatbot to answer routine course questions and send timely support. This gives students another route to help with coursework and deadlines, especially when staff are unavailable.

Documented result

The June 2026 analysis found a four-percentage-point increase in earning an A or B across two undergraduate courses.

Georgia State University; National Institute for Student Success
Atlanta, Georgia, United States 2026

A working paper on two courses; effects varied by subgroup and required staff-maintained content and human follow-up.

Reviewed 2026-11-14
AI-generated illustration: A candidate molecule moves from computational selection toward clinical testing.
From a suggested molecule to a 71-person trial: how far did the idea travel?
Better careTested in a study

An AI-designed lung drug reached patients. What did the trial find?

Discover the story
Result & evidence

AI helped identify a biological target and design rentosertib, creating a new treatment candidate for people with idiopathic pulmonary fibrosis and an early human signal worth testing further.

Documented result

In a 71-person phase 2a trial, mean forced vital capacity changed by +98.4 mL with 60 mg daily versus −20.3 mL with placebo over 12 weeks.

Insilico Medicine; participating Chinese hospitals
China 2025

Small, short trial with lung function as a secondary endpoint; liver-related withdrawals occurred, and lasting clinical benefit remains unproven.

Reviewed 2026-11-14
AI-generated illustration: Conceptual aircraft changing altitude beside a layer of contrail-forming air.
The atmosphere has layers where a flight leaves a different mark.
Energy & climateTested in a study

A small change in flight path can leave less cloud behind

Discover the story
Result & evidence

AI forecasts atmospheric regions where persistent contrails may form, allowing pilots to test small route adjustments that reduce a warming effect of aviation beyond its carbon-dioxide emissions.

Documented result

A 2026 randomized trial found 11.6% less contrail formation across 1,232 eligible flights; the 112 flown as planned showed 62% less.

Google Research; American Airlines; Breakthrough Energy
United States 2026

Public 2026 preprint, revised version accepted; contrail formation is not a direct measurement of net climate impact.

Reviewed 2026-11-14
AI-generated illustration: An ultrasound sweep provides patterns used to estimate gestational age.
Could a simple probe sweep bring pregnancy dating to clinics without specialists?
Earlier diagnosisTested in a study

A simple ultrasound sweep let novices estimate a pregnancy's age

Discover the story
Result & evidence

A handheld ultrasound device used AI to interpret simple abdominal sweeps, allowing briefly trained health workers to estimate gestational age where specialist scanners and sonographers are scarce.

Documented result

In 400 pregnancies in Zambia and North Carolina, novice users achieved estimates equivalent to expert sonographers between weeks 14 and 37.

University of North Carolina; University of Zambia; Butterfly Network
Zambia and North Carolina, United States 2024

This tested pregnancy dating, not improved birth outcomes; performance after week 37 was outside the recommended range.

Reviewed 2026-11-14
AI-generated illustration: Conceptual layered aerial view comparing intact and storm-damaged roofs.
Roofs become clues to where disaster responders should look first.
Disaster preparednessTested in a study

Finding damaged buildings before every roof is checked

Discover the story
Result & evidence

AI compares satellite images before and after a disaster to help humanitarian analysts locate damaged buildings, giving relief organisations faster evidence for prioritising assessments and directing limited emergency resources.

Documented result

Across nine emergencies, UN evaluators produced initial damage findings six times faster and covered seven times more area on average.

UNOSAT; UN Global Pulse; Google Research; WFP
Global 2024

Results concern preliminary, analyst-assisted assessments; individual building damage still needs scrutiny and suitable imagery.

Reviewed 2026-11-14
A bigger picture of AI starts with concrete evidence.Back to filters ↑