brianletort.ai

AI Transformation Library

See how AI changes the work.

102 real-world cases. What changed, what improved, and where people still decide.

CardsTable

102 cases

1–24

CaseAI valueOpen
Every mammogram flowing to two readers contrasts with a larger single-read lane and a smaller double-read lane.
Capital Region of Denmark mammography screening programme		33.5% fewer radiologist screening reads
An editorial scene for Google contrasts finds migration sites and writes context-specific edits across the monorepo. with discovers likely change locations and prompts a monorepo-trained llm to generate edits. in the llm-assisted 32-bit to 64-bit identifier migration workflow.
Google		74.45% of 595 submitted code changes and 69.46% of 93,574 edits were LLM-generated.
A family waiting for a separate eye appointment contrasts with retinal imaging performed during an existing clinic visit.
Johns Hopkins Children's Center / Johns Hopkins Medicine		Eye-exam completion: 22% → 100% in the control arm
An exhaustive field of regression tests contrasts with a model-selected subset monitored by a software engineer.
Meta (Facebook)		Runs fewer than one-third of dependency-selected tests
Sequential stroke imaging and transfer handoffs contrast with parallel clinical verification, team activation, and ambulance preparation.
Greater Houston primary stroke centers		Median transfer time: 210 → 133 minutes
An editorial scene for Swiss Plasma Center at EPFL and Google DeepMind contrasts designs a controller for a predefined plasma state and configuration. with specifies plasma shape, current, and location objectives. in the deep-reinforcement-learning magnetic control on the tcv tokamak workflow.
Swiss Plasma Center at EPFL and Google DeepMind		Across the full 0.1-1.0 second interval, RMSE was 0.62 kA for plasma current and 0.75 cm for shape.
Uniform herbicide broadcast across a field contrasts with camera-guided nozzles targeting individual weeds under a grower-set threshold.
John Deere (Deere & Company)		59% average herbicide savings
An editorial scene for The Permanente Medical Group contrasts conducts the visit while taking notes or reconstructs the encounter afterward. with transcribes the encounter without retaining audio and generates a draft clinical note. in the ambient ai draft-note generation during patient encounters workflow.
The Permanente Medical Group		3,442 physicians used the tool in 303,266 encounters; 968 used it in at least 100 encounters.
A growing queue of routine customer requests contrasts with automated resolution and expert human handling for complex conversations.
Klarna		Average resolution time: 11 → under 2 minutes
An editorial scene for Twenty-eight South-East Queensland long-term care facilities contrasts provides usual care and available psychosocial activities. with participates in an individual non-facilitated 15-minute responsive-robot session three times weekly. in the paro robotic-seal engagement sessions workflow.
Twenty-eight South-East Queensland long-term care facilities		PARO participants were significantly more verbally and visually engaged than plush-toy participants.
An editorial scene for Rising Academies schools contrasts attends normal math classes and studies with available static materials or limited human support. with uses a school-provided phone for two weekly whatsapp tutoring sessions during study hall. in the rori ai math tutoring through whatsapp workflow.
Rising Academies schools		Treatment growth was significantly higher with effect size 0.36 (p<0.001), described as roughly an extra year of learning.
A mortgage analyst surrounded by document queues contrasts with a human credit adjudicator receiving one prepared dossier.
TD Bank Group		Average preparation time: 15 hours → under 3 minutes
An editorial scene for UCLA Health contrasts conducts the visit and writes the note in the ehr. with records the conversation and generates a draft note. in the randomized deployment of dax and nabla ambient ai scribes workflow.
UCLA Health		Nabla arm fell from 4:30 to 3:49; a 9.5% greater reduction than control.
An editorial scene for University of California, San Francisco and UC Berkeley contrasts uses head tracking to select letters on a commercial assistive device. with attempts to speak while 253 cortical electrodes capture speech-motor signals. in the ai speech neuroprosthesis for text, audio, and avatar control workflow.
University of California, San Francisco and UC Berkeley		Median real-time decoding rate of 78.3 words per minute.
An editorial scene for Waymo contrasts perceives the road, plans maneuvers, and controls the vehicle for each passenger trip. with perceives, predicts, plans, and controls the vehicle with no human in the driver seat. in the fully autonomous waymo one driving workflow.
Waymo		9 Waymo claims, an 88% reduction.
An editorial scene for American Airlines contrasts manually evaluates arrivals and assigns among 136 dfw gates. with combines real-time flight and operational data to assign the nearest available gate with the shortest taxi time. in the smart gating machine-learning gate assignment workflow.
American Airlines		20% shorter, about two minutes per flight and more than 11 taxi hours saved daily.
An editorial scene for Deutsche Bahn contrasts physically walks around and climbs onto trains to inspect for damage. with capture exterior, underbody, and roof imagery as the train enters the plant. in the e-check camera-gate, ai diagnostics, and robotic inspection workflow.
Deutsche Bahn		One and a half hours with the digital E-Check; camera-gate traversal takes about five minutes.
An editorial scene for Ocado Group / Ocado Retail contrasts retrieves and picks grocery items through the established cfc process. with selects eligible grocery items robotically from the storage grid. in the on-grid robotic pick at luton cfc workflow.
Ocado Group / Ocado Retail		227 units per labor hour in FY24, a 9.1% improvement.
An editorial scene for Google DeepMind and EMBL-EBI contrasts search the much smaller experimental structure corpus or initiate a bespoke structure-determination project. with serves a precomputed predicted structure and confidence metrics for a queried protein sequence. in the alphafold protein structure database precomputed structure access workflow.
Google DeepMind and EMBL-EBI		More than 214 million predicted structures, covering nearly the full UniProt database
An editorial scene for Audi Neckarsulm contrasts uses ultrasound to inspect a random sample of resistance spot welds. with analyzes process data for approximately 1.5 million welds across 300 vehicles each shift. in the wps analytics full-population inspection workflow.
Audi Neckarsulm		AI analyzes about 1.5 million welds on 300 vehicles each shift
An editorial scene for Be My Eyes contrasts contacts a sighted volunteer or trusted person and shares a camera view. with captures an image and asks be my ai for a description or follow-up detail. in the conversational multimodal visual assistance through be my ai workflow.
Be My Eyes		Fourteen users described Be My AI support across personal and social contexts.
An editorial scene for Lawrence Berkeley National Laboratory and University of California, Berkeley contrasts select a target and design a synthesis recipe from literature and thermodynamics. with propose recipes and execute powder dosing, heating, and xrd characterization. in the a-lab closed-loop autonomous materials synthesis workflow.
Lawrence Berkeley National Laboratory and University of California, Berkeley		36 of 57 target compounds synthesized, a 63% success rate
An editorial scene for Boston Public Schools contrasts build and maintain routes manually for each school over multiple weeks. with generates district-wide student stops, route sequences, and multi-school bus reuse in about 30 minutes. in the bird system-level school-bus route optimization workflow.
Boston Public Schools		50 buses removed in the first implementation year and about $5 million saved, with average ride time remaining about 23 minutes
An editorial scene for Sixty-eight English secondary schools in the EEF Teacher Choices trial contrasts creates questions, quizzes, activities, and differentiated resources without generative ai. with uses an implementation guide and prompts the model for questions, quizzes, activity ideas, or tailored materials. in the teacher-controlled chatgpt lesson planning with implementation guide workflow.
Sixty-eight English secondary schools in the EEF Teacher Choices trial		56.2 minutes in the ChatGPT group, 25.3 minutes or 31% lower.
Curated collections →Evidence methodologyPublication history
AI Transformation Library | Brian Letort