Third-party research. These entries are summaries of publicly available company reports, vendor case studies, and peer-reviewed publications. They are not Ajentik customer references. Every numeric claim links back to its original source; entries marked vendor-sourced reflect figures published by the vendor and have not been independently audited.
Industry Research
Sourced summaries of how leading organisations are deploying AI in healthcare and adjacent industries — drawn entirely from public reporting.
Healthcare Industry
メイヨークリニックAIファクトリー：プラットフォーム駆動型臨床AI
世界最大の非営利統合医療システムが最も積極的な AI 採用者となる
チャレンジ
Mayo faced the need to democratize AI development across their organization with 76,000 staff, enable clinicians to create AI solutions without extensive technical expertise, manage a large concurrent portfolio of AI projects and ensure regulatory compliance for medical AI applications while maintaining patient safety.
Reported approach
Mayo は、Google の Vertex AI 上で AI Factory プラットフォームを開発し、AI ツールの「シチズン開発」を可能にしました。彼らの自律エージェントには、ECG 読み取り値からの心臓病の検出 (FDA 承認)、ICU キャパシティ管理システム、および癌研究のための仮説駆動型 AI が含まれます。このプラットフォームには、ガバナンスのための医療機器としてのソフトウェア審査委員会が設置されています。
Reported outcomes
- ·アヌマナなどのスピンオフ企業を通じて商業化に成功
- ·左室機能障害を対象とした Apple Watch の AI 心電図研究に 2,454 人が参加
- ·次世代を育成する医療AI学位プログラムを設立
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:Mayo Clinic operates its own AI programs (Mayo Clinic Platform); no Ajentik partnership exists or is implied. Corrected against primary sources: Anumana (the Mayo/nference spinoff) has three FDA-cleared ECG-AI algorithms, not five (low ejection fraction, pulmonary hypertension, cardiac amyloidosis); the Apple Watch reference is the 2,454-participant AI-ECG study, not 'millions'; and the unsourced '10,000+ staff trained' and internal-reporting figures were removed. Last verified 2026-07-21.
Documented Clinical Documentation and Medical Coding Deployments
Publicly documented Ambience Healthcare and AKASA deployments
チャレンジ
Clinical documentation and medical coding are the only deployment areas retained after source review.
Reported approach
Public source verification supports Cleveland Clinic Ambience clinical-documentation and AKASA medical-coding deployments. Earlier predictive-analytics, synthetic-data, AI phone-system, and integrated-ecosystem claims were removed pending evidence.
Reported outcomes
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:Cleveland Clinic runs its own AI programs through Cleveland Clinic Innovations; no Ajentik partnership exists or is implied. Four call-automation figures previously shown here were traced to a vendor blog about an unrelated UK NHS GP surgery and have been removed. Only two verified deployments remain: Ambience Healthcare ambient documentation (Cleveland Clinic newsroom, 19 Feb 2025) and AKASA coding, ~100+ documents in 1.5 minutes (newsroom, 29 Apr 2025). The 13.7 million figure refers to annual outpatient encounters.
オンコロジー AI オーケストレーター: 高度な臨床アプリケーション
AI を活用したがん治療で年間 4,000 人の腫瘍ボード患者にサービスを提供
チャレンジ
スタンフォード大学は、医師が画像、病理学、ゲノミクス、臨床記録の検討に患者 1 人あたり 1.5 ～ 2.5 時間を費やし、情報過多に直面していました。急速に進化するがん研究（30秒ごとに新しい論文が発表される）に追いつくのが難しい。時間のかかる腫瘍ボードの準備、および 400,000 以上のアクティブな臨床試験から適切な臨床試験に患者をマッチングするという課題があります。
Reported approach
Stanford deployed Microsoft-powered Healthcare Agent Orchestrator featuring specialized agents: tumor board agent analyzing multimodal data, care coordination agent providing treatment recommendations, research literature agent processing medical updates and clinical trial matching agent. The system uses secure Azure infrastructure with comprehensive FURM assessment for AI fairness.
Reported outcomes
- ·年間 4,000 人の腫瘍ボード患者をサポート
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:This is a research pilot, not a clinical deployment — per Microsoft's May 2025 Build announcement, Stanford is researching the Healthcare Agent Orchestrator, not running it in production. No Ajentik partnership exists or is implied. The 60% figure (which actually belongs to JM Family), plus the 25%, 300% ROI, '10,000 papers/month', and 'days to hours' claims were removed; only the ~4,000 tumor-board patients/year volume and the FURM safety framework remain. Last verified 2026-07-21.
新型コロナウイルス感染症の創薬: 研究の加速
AI を活用した創薬により、記録的な速さで新型コロナウイルス感染症の治療法を特定
チャレンジ
新型コロナウイルス感染症（COVID-19）のパンデミック中、研究者らは毎日50万件を超える症例、圧倒的な量の科学文献（毎週5,000件以上の新型コロナウイルスに関する論文が発表）、従来の10～15年に及ぶ医薬品開発スケジュール、長期にわたる治験を避けるために迅速に転用できる安全な既承認薬を特定する必要性など、治療選択肢の緊急の必要性に直面した。
Reported approach
BenevolentAI deployed autonomous agents including biomedical literature mining agent processing millions of scientific papers, drug-target interaction agent predicting novel relationships using knowledge graphs, clinical trial optimization agent for patient selection and ADME prediction agent for drug properties. The system identified baricitinib as potential treatment in just 48 hours.
Reported outcomes
- ·バリシチニブは2020年11月までにFDA緊急使用許可を取得
- ·AAK1阻害によりサイトカインストームの軽減に成功
- ·一般的な 10 年間のタイムラインを 9 か月に圧縮
- ·Analyzed 1M+ scientific papers in days vs years
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:BenevolentAI identified baricitinib (used in the baricitinib + remdesivir EUA) in about 48 hours. The '71% reduced recovery time' figure was fabricated — the ACTT-2 trial showed roughly one day's improvement (median 7 vs 8 days). The '$800M Alzheimer's deal' is unrelated and misdated, and the '20+ disease areas' claim predates a 2023 pipeline cut to ~5 assets — all removed. Company status: BenevolentAI was delisted from Euronext Amsterdam in March 2025 following a wind-down.
予測型高齢者ケア: AI を活用した転倒予防と健康モニタリング
Wearable AI platform revolutionizes senior care with continuous behavioral monitoring and predictive analytics
チャレンジ
高齢者向け住宅施設は、健康問題が事故発生後にのみ対処されるという事後対応型のケアモデルに直面していました。転倒は65歳以上の成人の傷害死亡の主な原因であり、年間医療費が500億ドルに達している。施設は、スタッフ不足、一貫性のないモニタリング、緊急事態前に健康状態の悪化を予測できないことに悩まされていました。
Reported approach
CarePredict は、食事、睡眠、歩行、トイレの使用など 18 以上の日常活動を継続的に監視する手首に装着する AI デバイスである Tempo を開発しました。このシステムは機械学習を使用して個々のベースラインを確立し、健康事象に先立つ微妙な逸脱を検出します。 AI アルゴリズムは、尿路感染症を症状の 3.5 日前に予測し、尿路感染症が起こる前に転倒し、活動パターンの変化によるうつ病の発症を予測します。
Reported outcomes
- ·導入された施設全体で転倒が 69% 削減
- ·監視対象住民の入院が 39% 減少
- ·臨床診断の2～3週間前にうつ病を予測
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:Vendor-authored, retrospective study (JMIR Aging 2020, 6 communities / 490 residents). The 69% lower fall rate and 39% fewer hospitalizations are retained with that caveat; the unsourced '80% UTI accuracy', '2 hrs/day', and '18+ activities' figures were removed. Not independently audited. Last verified 2026-07-21.
AI 高齢者向けコンパニオン: 孤独と大規模に闘う
共感力のある AI を搭載したソーシャル ロボットは、毎日 30 回以上のやり取りで高齢者の孤立を軽減します
チャレンジ
社会的孤立は65歳以上の高齢者の4人に1人に影響を及ぼし、死亡リスクは26％、認知症リスクは50％増加する。定期通話やコミュニティ プログラムなどの従来のソリューションは、孤立した高齢者の一部にしか届きません。新型コロナウイルス感染症のパンデミックは孤立を悪化させ、多くの高齢者が有意義な人間関係のない日々を過ごしている。
Reported approach
ElliQ は、会話を開始したり、活動を提案したり、薬のリマインダーを提供したり、家族とのビデオ通話を促進したり、健康運動を指導したりする、プロアクティブな AI コンパニオンです。コマンドを待つ受動的なデバイスとは異なり、ElliQ は共感力のある AI を使用して、気分を感知し、好みを学習し、一日を通して高齢者と関わります。このロボットは、会話的な AI と、感情的なつながりを生み出す物理的な存在を組み合わせています。
Reported outcomes
- ·ユーザーの 95% が 30 日後に孤独感が軽減されたと報告
- ·高齢者は毎日 30 回以上 ElliQ とやり取りしています
- ·ニューヨーク州は Area Agencies on Aging を通じて 800 人以上の孤立した高齢者に提供
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:The 95% figure is a self-report from the New York State Office for the Aging (NYSOFA, Aug 2023), whose own headline reads '95% report reduced loneliness' — retained as a self-report, not a clinical outcome. The unsourced '90% retention', '~$10K/yr savings', and '3x video calls' figures were removed, and '800+ to Medicaid recipients' was corrected to 800+ isolated older adults reached via Area Agencies on Aging. Last verified 2026-07-21.
AI を活用した退院: 患者の流れのボトルネックを解消
機械学習プラットフォームにより超過入院日数が削減され、数百万ドルの運用コストが節約されます
チャレンジ
退院は、医師、看護師、ソーシャルワーカー、薬剤師、急性期後施設間の調整が必要となる複雑なことで知られています。 OhioHealth は、14 の病院で年間 150 万人の患者にサービスを提供していますが、退院の遅延に悩まされており、超過 1 日あたり 800 ドル以上の費用が発生しています。患者は保険の承認や熟練した看護師の配置などの非臨床プロセスを待つために入院したままになることがよくありました。
Reported approach
Qventus deployed AI agents that predict discharge readiness, automate milestone tracking and orchestrate multi-team workflows. The system analyzes 100+ variables including clinical status, social determinants and post-acute bed availability to predict and accelerate safe discharges. AI identifies barriers early and automatically routes tasks to appropriate team members.
Reported outcomes
- ·初年度に 8,554 の超過患者日数を削減
- ·スループットの向上により年間 170 万ドルを節約
- ·急性期後のケアの必要性をより早期に特定（2 日以上前）
- ·事務負担の軽減によりスタッフの満足度が向上
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:Vendor-reported (Qventus–OhioHealth case study), not independently audited. The 8,554 figure is an annualized projection and the $1.7M is cumulative to date. The '0.5-day LOS reduction', '+25% discharge efficiency', '30% ED boarding', and '14 hospitals/1.5M patients/$800 per day' figures could not be verified and were removed (OhioHealth operates 15 hospitals). Last verified 2026-07-21.
AI を活用した遠隔患者モニタリング: 在宅での病院レベルのケア
FDA の認可を受けた AI プラットフォームは、継続的なバイタルサイン監視により再入院を 削減します
チャレンジ
退院後の患者は医療の過程で最もリスクの高い時期に直面しており、メディケア患者の 5 人に 1 人が 30 日以内に再入院し、年間 260 億ドルの費用がかかります。電話やオフィス訪問といった従来のフォローアップでは、問題を発見するのが遅すぎます。心不全、COPD、その他の慢性疾患を患う患者は、警告サインがケアチームに届くことなく自宅で悪化します。
Reported approach
Biofourmis deploys FDA-cleared wearable biosensors combined with AI algorithms that continuously analyze 20+ physiological parameters. The Biovitals platform detects subtle deterioration patterns 8+ hours before clinical symptoms appear, enabling proactive intervention. AI personalizes alert thresholds based on each patient's baseline, dramatically reducing false alarms while catching true deterioration.
Reported outcomes
- ·臨床チームは生データではなく実用的な洞察を受け取ります
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:Vendor-reported, not independently audited. The FDA 510(k) clearance K183282 (Biovitals Analytics Engine, 2019) is verified; the unsourced '8+ hours early detection', '97%', '$12K/patient', '89% mortality reduction', and '20+ parameters' figures were removed. Company status: Biofourmis merged with CopilotIQ (Oct 2024), and its life-science arm was acquired by ActiGraph (Jan 2025). Last verified 2026-07-21.
AI とマッチングしたコンパニオン ケア: 人間関係の拡大
テクノロジープラットフォームが高齢者と「パパパル」仲間をマッチングし、医療費を9%削減
チャレンジ
Healthcare plans struggled to address social determinants of health—transportation barriers, social isolation and daily living challenges—that drive costly medical utilization. Traditional home care focused on clinical tasks, missing the companionship and practical support that prevents health decline. Seniors needed both social connection and help with errands, technology and appointments.
Reported approach
パパは AI を使用して、高齢者と「パパ パル」をマッチングします。「パパ パル」は、移動手段、同伴者、テクノロジー支援、簡単な家事を提供する、厳選された仲間です。プラットフォームのアルゴリズムは、性格、興味、言語、特定のニーズを考慮して、最適なマッチングを作成します。 AI は、訪問パターンと結果を監視して、マッチングを継続的に改善し、新たな健康リスクを特定します。
Reported outcomes
- ·熱心なメンバーの総医療費を 9% 削減
- ·入院数が 18% 減少
- ·平均会員満足度評価は 4.8/5
- ·200万人以上のコンパニオン訪問が完了しました
- ·ピーク時に約 70 の医療プランと提携（Humana、Aetna、Centene を含む）、2023 年以降は縮小
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:Vendor-reported (Papa's May 2024 Medicare Advantage claims study): the 9% cost reduction and 18% fewer inpatient admissions are retained as vendor claims. '150+ health plans' was corrected to about 70 at peak, contracting since 2023. Reputational context: a May 2023 Bloomberg Businessweek investigation detailed abuse allegations, after which major insurers (Humana, Aetna, Molina) declined to renew. Last verified 2026-07-21.
AI を活用した慢性疾患管理: 大規模な糖尿病管理
AI コーチングを備えたコネクテッド デバイス プラットフォームは、集中的な対面ケアに匹敵する臨床成果を達成します
チャレンジ
糖尿病は世界中で 5 億 3,700 万人の成人に影響を与えており、その管理には血糖、食事、運動、投薬に常に注意を払う必要があります。従来のケア（四半期ごとの医師の診察）では、99% の確率で患者はサポートを受けられません。管理が不十分な場合、合併症が発生し、米国だけで年間 3,270 億ドルの損失が発生します。患者には、一時的な予約ではなく、継続的な指導が必要です。
Reported approach
Livongo provides a connected blood glucose meter that uploads readings in real-time to an AI platform. When readings fall outside personalized parameters, AI triggers immediate coaching interventions—sometimes automated messages, sometimes live certified diabetes educators. The system learns each member's patterns, providing proactive guidance before problems occur.
Reported outcomes
- ·A1c 平均 0.9% (臨床的に有意)
- ·会員の 86% が月に 16 回以上血糖値をチェックしています
- ·メンバー 1 人当たりの年間平均コスト削減額は 1,908 ドル
- ·糖尿病関連の医療費を 21% 削減
- ·糖尿病に関連する救急外来受診が 30% 減少
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:Vendor-reported, not independently audited. The ~0.8–0.9% A1c reduction is Livongo's own sustained figure; the '$1,908 savings' traces to the Livongo S-1 filing; '1.2M members' is Teladoc's total, not Livongo's. The unsourced '86% engagement' was removed and '2,000+ employers' corrected toward the ~1,328 actually reported. Context: Teladoc recorded multibillion-dollar goodwill impairments after its 2020 Livongo acquisition. Last verified 2026-07-21.
AI 介護者の評価: 燃え尽き症候群を未然に防ぐ
証拠に基づいたプラットフォームは AI を使用して家族の介護者の燃え尽き症候群のリスクを特定し、それに対処します
チャレンジ
家族の介護者（5,300万人のアメリカ人）は、60％を超える燃え尽き症候群に直面しており、自らの健康上の問題を引き起こし、介護を継続できなくなる。介護中の従業員が勤務時間を短縮したり、仕事を休んだり、退職したりすることで、雇用主は年間 330 億ドルの損失を被っています。従来のサポート プログラムは、燃え尽き症候群を防ぐというよりも、すでに危機に瀕している介護者に役立ちます。
Reported approach
TCARE uses an AI-powered assessment algorithm developed from 20+ years of academic research to quantify caregiver burden across multiple dimensions: identity discrepancy, care burden, relationship quality and health. The platform then generates personalized action plans with specific interventions proven to address identified risks. AI continuously learns from outcomes to improve recommendations.
Reported outcomes
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:Vendor-reported, not independently audited. The '+35% retention' is an employee-retention figure (vendor framing), and the unsourced 42%, +25%, '$200-400', and '15-min' figures were removed. The evidence base is the peer-reviewed Montgomery et al. 2011 randomized controlled trial (266 caregivers, 4 states), not a 40-state deployment. Last verified 2026-07-21.
バーチャル看護師アバター: AI の取り組みを通じて再入院を削減
共感力のある AI アバターが退院後のフォローアップを実施し、再入院率 5% 未満を達成
チャレンジ
Post-discharge patient follow-up is critical but difficult to scale. Hospitals face 20% readmission rates with $26 billion in annual penalties. Phone call follow-up reaches only 30-40% of patients, with nurses spending hours on unsuccessful calls. Patients forget instructions, don't recognize warning signs and delay seeking care until emergencies occur.
Reported approach
Sensely の AI を活用した仮想看護師アバターであるオリビアは、スマートフォンを介してチェックインを行い、自然な会話で症状、服薬遵守、懸念事項について質問します。このシステムは、音声認識、感情分析、臨床プロトコルを使用して、悪化のリスクがある患者を特定します。高リスク患者は、完全な会話コンテキストを備えた臨床スタッフに自動的にエスカレーションされます。
Reported outcomes
- ·NHS、メイヨークリニック、主要な医療システム全体に展開
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:Vendor-reported, not independently audited. The original four tiles (<5% / 85% / 93% / 4.5) had no source and were removed; the entry is rebuilt on the real 'Ask NHS' pilot (Sensely/PRNewswire, Apr 2021). Company status: Sensely was acquired by Mediktor in June 2024. Last verified 2026-07-21.
AI-Optimized Hospital Operations: From Bottlenecks to Flow
Machine learning platform optimizes infusion centers ORs and bed management across 600+ hospitals
チャレンジ
病院の各部門は、がんの輸液、手術、入院が 1 日を通して予測不可能に発生するため、非常に変動する需要に直面しています。静的なスケジューリングでは、午前中のラッシュと午後の落ち着きが生じ、患者はキャパシティが空いている間待たされることになります。その結果、ボトルネックが連鎖的に発生します。ER への入院、手術の遅延、スタッフの残業によりコストがかさみ、患者は苦しみます。
Reported approach
LeanTaaS iQueue uses machine learning to predict demand patterns and optimize scheduling across infusion centers, operating rooms and inpatient beds. The system analyzes historical patterns, treatment durations and real-time data to create optimal appointment templates. AI continuously learns from actual flow to improve predictions and identifies scheduling opportunities in real-time.
Reported outcomes
- ·患者の待ち時間を 50% 削減
- ·大規模な医療システムで年間 1,000 万ドル以上を節約
- ·スタンフォード大学、UCSF、クリーブランド クリニックを含む 1,000+ 以上の病院に導入
- ·患者のスループットが 15% 増加
- ·スタッフの残業時間を25%削減
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:Vendor-reported, not independently audited. The '50%' utilization figure was corrected toward the ~31% in LeanTaaS's own case studies, and the unsourced +20%, $10M, and +30% figures were removed. '600+ hospitals' was updated to the published figure: LeanTaaS serves 1,000+ hospitals and centers across 185 health systems (Dec 2023). Last verified 2026-07-21.
Historical Current Health Hospital-at-Home Platform
Historical overview; Best Buy completed the sale of Current Health in June 2025
チャレンジ
This entry is retained only as historical context for Best Buy ownership of Current Health.
Reported approach
Best Buy acquired Current Health in 2021 and completed its sale in June 2025. This entry describes the former ownership period and does not represent a current Best Buy Health deployment.
Reported outcomes
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:The 38% cost-reduction and <10% readmission figures previously shown were outcomes from Levine et al. (Annals of Internal Medicine, 2020), a Brigham & Women's home-hospital RCT of 91 patients — not Best Buy Health / Current Health results — and the 95% preference, 30%-faster, 'Equal+', and reach figures were unsourced; all removed. Company status: Best Buy divested Current Health in June 2025 and cut 161 roles in July 2025; this entry is retained as history.
Cross-Industry
顧客サービス革命: 医療コミュニケーションの教訓
世界中で 1 億 5,000 万人以上のユーザーにサービスを提供するスウェーデンのフィンテック大手が AI で顧客サービスを変革
チャレンジ
Klarna は平均 11 分の解決時間に苦労しており、35 以上の言語にわたる毎日の数百万件のトランザクションに対して 24 時間年中無休の多言語サポートを必要としていました。同社は、高い顧客満足度と規制順守を維持しながら、機密の財務情報を処理できるソリューションを必要としていました。
Reported approach
Klarna deployed an OpenAI-powered autonomous AI assistant that handled 2.3 million conversations in its first month (February 2024). The system processes natural language queries, makes autonomous decisions about refunds and payments and seamlessly escalates complex cases to human agents. It maintains context across conversations and provides personalized responses based on customer history.
Reported outcomes
- ·解決時間が 11 分から 2 分未満に短縮されました (82% 改善)
- ·AI は 700 人の常勤エージェントに相当する作業を実行します
- ·繰り返しの問い合わせが 25% 減り、顧客満足度が向上
- ·年間 4,000 万ドル以上の利益改善が見込まれる
- ·ネイティブレベルの流暢さで35以上の言語をサポート
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:Figures are Klarna's own press release (company-reported, not independent third-party research). The 2.3M conversations figure covers Klarna's first month (February 2024), not a full year. For balance: Klarna publicly rebalanced toward human support in 2025 after the initial rollout.
Erica: ヘルスケア仮想アシスタントの青写真
Serving 40+ million mobile banking customers with personalized AI guidance
チャレンジ
この銀行は、日常的な問い合わせに人間のエージェントが多大な時間を費やすという、圧倒的な顧客サービスの需要に直面していました。顧客は、人間の支援を待たずに、アカウント情報、取引履歴、財務上のアドバイスに即座にアクセスできる必要がありました。従来の IVR システムは顧客に不満を抱かせ、高い離脱率を引き起こしていました。
Reported approach
Erica, launched in 2018, uses predictive analytics and natural language processing to provide comprehensive financial assistance. The AI agent autonomously handles balance inquiries, payment scheduling, spending analysis, fraud detection and personalized financial insights. It integrates seamlessly with mobile banking and learns from each interaction.
Reported outcomes
- ·リリース以来 30 億件以上のインタラクションを処理
- ·4,200 万人の顧客 (モバイル ユーザーの 50%) にサービスを提供
- ·毎日 150 万件のインタラクションを処理
- ·クライアントの98%以上が44秒以内に回答を得る（人間へのエスカレーションあり）
- ·Reduced call center volume by 50% for routine inquiries
- ·顧客満足度 4.7/5 を達成
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:Figures are Bank of America's own disclosures (company-reported, not independent third-party research). The '+19%' figure was a bank-wide Q2-2023 net-income headline never attributed by BofA to Erica and has been removed. '98% autonomous resolution' was a speed metric, not autonomy: BofA reports that over 98% of clients get answers within 44 seconds, with human escalation. Totals updated to the August 2025 release (3B+ interactions, ~50M users).
Devin AI 実装: ヘルスケア業務のスケーリング
ブラジル最大のフィンテック企業が自律型 AI でレガシー システムを変革
チャレンジ
同社は、8 年前から 600 万行のモノリシック ETL システムで 100,000 を超えるデータ クラスの実装をリファクタリングし、レガシー システムの複雑な相互依存関係を管理し、従来の移行による大量のリソース割り当て (18 か月間で 1,000 人以上のエンジニア) を回避し、変革中のシステムの安定性を維持する必要がありました。
Reported approach
Nubank deployed Cognition Labs' Devin AI, an autonomous software engineering agent. Devin analyzed the monolithic codebase, created migration strategies, generated modular sub-components and executed systematic refactoring. The AI learned from each task, improving performance over time.
Reported outcomes
- ·エンジニアリング時間を 12 倍に短縮し節約
- ·手動移行と比較して 20 倍のコスト削減
- ·タスク完了時間を 40 分から 10 分に短縮
- ·移行を数か月または数年ではなく数週間で完了
- ·移行中の本番インシデントはゼロ
- ·1,000 人以上のエンジニアをイノベーション活動に解放
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:Outcome figures originate from Cognition AI (Devin) marketing about a Nubank engineering engagement — vendor-reported, not independently audited. Nubank is a digital bank, not a healthcare organization (the earlier 'Healthcare Operations' headline was wrong). The unsourced 'zero production incidents' claim was removed, and 'freed 1,000+ engineers' is the vendor's counterfactual, not an audited outcome. Last verified 2026-07-21.
Salesforce Agentforce: 保険とヘルスケアの融合
50 か国以上で 5,000 万人の顧客にサービスを提供する 150 年の会社
チャレンジ
従業員 38,000 人のプルデンシャルは、州ごとの複雑な保険規制、複数の事業部門にまたがる時間のかかる保険金請求処理、包括的なサービスを妨げる断片化された顧客データ、50 以上の異なる規制枠組みのナビゲーションを必要とする顧客サービス業務における多大な手作業に苦労していました。
Reported approach
Prudential implemented Salesforce Agentforce for Financial Services, deploying autonomous agents for customer identification, contract analysis, policy retrieval and claims processing. The system features human-in-the-loop oversight for regulated operations and multi-LLM architecture for specialized functions.
Reported outcomes
- ·カスタマー サービス担当者 1 人当たり、週に少なくとも半日を節約
- ·何百もの手動ルーティングワークフローを排除
- ·管理負担の軽減による顧客共感の向上
- ·卸売業の生産性が大幅に向上
- ·年間 1,200 万ドルの運用コストの節約
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:Vendor-reported (Salesforce customer story), and a pilot/beta rather than a deployment — Agentforce launched in late 2024. Prudential is a financial-services firm, not healthcare. The 'half a day' saving applies to annuity wholesalers and is a projected figure ('up to half a day per week'), not a realized CSR outcome; the unsourced 'hundreds of workflows', '100% compliance', and '$12M' figures were removed. Last verified 2026-07-21.
UiPath Enterprise Automation: Verified PACE Evidence
Global bank reports a vendor-sourced efficiency benefit across its PACE automation portfolio
チャレンジ
This entry is limited to published evidence about the Barclays PACE automation portfolio; earlier mortgage-processing and document-agent details were removed pending evidence.
Reported approach
Barclays uses UiPath for enterprise automation and is exploring agentic workflows. The retained vendor-reported evidence is a 30%+ efficiency benefit across its PACE automation portfolio; earlier document-agent deployment claims were removed pending evidence.
Reported outcomes
- ·Vendor-reported 30%+ efficiency benefit across PACE implementations
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:The 98% straight-through and 12,000-hours figures shown in earlier versions were Fiserv's numbers from the same UiPath On Tour London coverage, not Barclays'; the days-to-minutes figure was a UiPath product demo and the £25M figure was unsourced — all removed. Barclays is exploring agentic workflows; the only correctly scoped figure retained is the 30%+ efficiency benefit across its PACE automation portfolio (vendor-reported, not independently audited).
Microsoft Copilot Studio: サプライ チェーン インテリジェンス
世界的な材料会社が AI オートメーションでサプライ チェーンを変革
チャレンジ
Dow struggled with manual processing of 100,000+ PDF invoices yearly from 5,000+ suppliers, difficulty detecting billing inaccuracies and anomalies costing $3-5M annually, time-consuming freight rate investigations taking weeks or months per dispute and lack of visibility into cost optimization opportunities across global supply chain.
Reported approach
Using Microsoft Copilot Studio, Dow deployed autonomous invoice scanning agents for billing analysis and a natural language "Freight Agent" for investigation. The system features automatic anomaly detection, pattern recognition for cost optimization, dashboard integration for employee review and conversational interface for deep analysis.
Reported outcomes
- ·最初の 1 年以内に 500 万ドル以上の節約が見込まれる
- ·調査時間を数週間から数分に短縮
- ·スタッフを追加せずに 100,000 件以上の請求書を処理できるように拡張
- ·第 1 四半期に 200 万ドルの超過請求を特定
- ·技術者以外のスタッフでも複雑な分析を実行できるようになりました
- ·手動データ入力を 95% 削減
What this signals for the sector
This entry is included as a research note about the broader healthcare AI landscape. It does not describe an Ajentik product, customer, or partnership. Readers can use the source links below to inspect the original reporting.
Editorial note:Vendor-reported (Microsoft WorkLab 'AI impact at Dow'), not independently audited. The '$5M+ first year' figure was softened to 'millions (anticipated, once scaled)' to match the source, and the unsourced '99.5% accuracy' and '$2M first quarter' figures were removed. Last verified 2026-07-21.
Want to discuss your own deployment?
These research notes describe other organisations work. If you would like to talk to us about your own healthcare AI plans, contact our team.