{
  "id": "2026-10-04-youtube-customer-stories",
  "parent": "2026-10-04-zendesk-ai-agents-deep-analysis",
  "date": "2026-10-04",
  "kinds": ["evidence"],
  "title": {
    "en": "Customer stories from YouTube for eleven agents",
    "es": "Casos de clientes en YouTube para once agentes"
  },
  "summary": {
    "en": "We added 34 claims from customer interviews on YouTube to Agentforce, Decagon, Fin, Forethought, Gorgias AI Agent, Kore.ai, LivePerson, Lyro, NiCE Cognigy, PolyAI and Zendesk AI Agents, each linked to the moment in the video it comes from. Thirty-three come from videos the vendor published and are labelled vendor-reported; Cloudbeds describing its use of Forethought, on Gainsight's channel, is the only independent report. They add named-customer results alongside the limits those customers describe: knowledge articles that had to be rewritten, daily review of answers, and wrong answers until guidance changed. No evidence state moves.",
    "es": "Añadimos 34 afirmaciones procedentes de entrevistas con clientes en YouTube a Agentforce, Decagon, Fin, Forethought, Gorgias AI Agent, Kore.ai, LivePerson, Lyro, NiCE Cognigy, PolyAI y Zendesk AI Agents, cada una enlazada al momento del vídeo del que procede. Treinta y tres proceden de vídeos publicados por el propio proveedor y se etiquetan como declaradas por el proveedor; el relato de Cloudbeds sobre su uso de Forethought, en el canal de Gainsight, es el único informe independiente. Aportan resultados de clientes con nombre junto a los límites que esos mismos clientes describen: artículos de conocimiento que hubo que reescribir, revisión diaria de las respuestas y respuestas erróneas hasta que se cambiaron las instrucciones. Ningún estado de evidencia cambia."
  },
  "patch": [
    {
      "op": "add",
      "path": "/agentforce/strengths/21",
      "value": {
        "id": "agentforce-yt-remarkable-orders",
        "text": "Tine Engberg of reMarkable reports on Salesforce’s channel that its external Agentforce agent began answering questions about orders and tracking after integration with Salesforce Order Management. This describes reMarkable’s configured deployment, not an out-of-the-box action.",
        "sourceType": "vendor-reported",
        "sourceTitle": "How reMarkable Transformed their Customer Service | Salesforce Agentforce (YouTube, Salesforce Product Center)",
        "sourceUrl": "https://www.youtube.com/watch?v=HvIACIq2fMM&t=490s",
        "publishedAt": "2026-04-08",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/agentforce/weaknesses/21",
      "value": {
        "id": "agentforce-yt-remarkable-knowledge-rework",
        "text": "Tine Engberg of reMarkable reports on Salesforce’s channel that its existing knowledge articles needed case-by-case reworking for Agentforce: “AI is not reading them as we do”. The team adjusted articles for questions its agent Mark could not answer.",
        "sourceType": "vendor-reported",
        "sourceTitle": "How reMarkable Transformed their Customer Service | Salesforce Agentforce (YouTube, Salesforce Product Center)",
        "sourceUrl": "https://www.youtube.com/watch?v=HvIACIq2fMM&t=326s",
        "publishedAt": "2026-04-08",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/agentforce/weaknesses/22",
      "value": {
        "id": "agentforce-yt-remarkable-resource-burden",
        "text": "Tine Engberg of reMarkable reports on Salesforce’s channel that the initial Agentforce rollout was resource-intensive: “initially it was really tough generally just on resource”. She says it subsequently became part of normal customer-support development.",
        "sourceType": "vendor-reported",
        "sourceTitle": "How reMarkable Transformed their Customer Service | Salesforce Agentforce (YouTube, Salesforce Product Center)",
        "sourceUrl": "https://www.youtube.com/watch?v=HvIACIq2fMM&t=643s",
        "publishedAt": "2026-04-08",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/decagon/strengths/24",
      "value": {
        "id": "decagon-yt-faire-return-initiation",
        "text": "In a customer interview on Decagon’s channel, Faire’s customer-experience and central-operations leader reports that initiating a return previously took time and that, after implementing Decagon, “customers are now able to do it seamlessly”. No completion-rate or time-saving measurement is supplied.",
        "sourceType": "vendor-reported",
        "sourceTitle": "Faire Eliminates Support Roadblocks with Decagon’s Fast, Feedback-Driven AI (YouTube, Decagon AI )",
        "sourceUrl": "https://www.youtube.com/watch?v=Zf8kfA75U0g&t=22s",
        "publishedAt": "2025-02-20",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/decagon/weaknesses/19",
      "value": {
        "id": "decagon-yt-faire-workflow-bottleneck",
        "text": "In a Decagon-hosted interview, Faire VP of Global Operations Shan Hu says expansion of support through its virtual assistant was constrained by “how many workflows we can build”. Training helped the team begin writing AOPs; this is a deployment-capacity limitation, not an assertion of general model failure.",
        "sourceType": "vendor-reported",
        "sourceTitle": "Faire's VP, Global Ops Shan Hu | Decagon Dialogues '25 (YouTube, Decagon AI )",
        "sourceUrl": "https://www.youtube.com/watch?v=4pcGV2gOVtU&t=352s",
        "publishedAt": "2025-10-20",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/fin/strengths/17",
      "value": {
        "id": "fin-yt-nuuly-subscription-actions",
        "text": "In an Intercom-published customer account, Natalie Hurst, Senior Director of Customer Success at Nuuly, says Fin has been taking subscription pause/cancel actions for two months, replacing tickets where an associate previously checked account and return status. At 00:43 she reports a 10% increase in Fin resolution rate, equating to about 20,000 conversations monthly, and at 01:04 a 5% drop in conversations handled by associates.",
        "sourceType": "vendor-reported",
        "sourceTitle": "Fin Handles Subscription Issues for Nuuly Customers (YouTube, Intercom)",
        "sourceUrl": "https://www.youtube.com/watch?v=GMFAwbrcxGU&t=21s",
        "publishedAt": "2025-09-16",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/fin/strengths/18",
      "value": {
        "id": "fin-yt-nuuly-tasks-csat",
        "text": "In a Fin-published interview, Natalie Hurst of Nuuly reports that Fin customer satisfaction was around 40% before Tasks could take the actions associates could, and reached 80% after Tasks were enabled and monitored. These are Nuuly-specific results, not a general Fin benchmark.",
        "sourceType": "vendor-reported",
        "sourceTitle": "Live from Pioneer: AI is handling complexity we never thought it could (YouTube, Fin)",
        "sourceUrl": "https://www.youtube.com/watch?v=YiPW_sWMelo&t=460s",
        "publishedAt": "2025-10-21",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/fin/weaknesses/24",
      "value": {
        "id": "fin-yt-nuuly-daily-monitoring",
        "text": "In a Fin-published interview, Natalie Hurst of Nuuly cautions that Fin cannot simply be switched on and assumed to work perfectly: Nuuly dedicated someone to checking responses and actual actions, tweaking daily, and maintaining articles, guidance and snippets.",
        "sourceType": "vendor-reported",
        "sourceTitle": "Live from Pioneer: AI is handling complexity we never thought it could (YouTube, Fin)",
        "sourceUrl": "https://www.youtube.com/watch?v=YiPW_sWMelo&t=270s",
        "publishedAt": "2025-10-21",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/forethought/strengths/20",
      "value": {
        "id": "forethought-yt-cloudbeds-deflection-deployment",
        "text": "In a Gainsight interview, Colin Slade, Senior Vice President of AI Strategy and Customer Success at Cloudbeds, says Cloudbeds continues to use Forethought for front-end deflection and considers recreating and maintaining that capability internally difficult. He describes Forethought as an existing component alongside Cloudbeds’ separate internally built AI workflows; the interview’s broader time-saving and automation metrics are not attributed specifically to Forethought.",
        "sourceType": "independent-report",
        "sourceTitle": "How a 5-Person Team Built 150 AI Workflows That Changed an Entire Company ft. Colin Slade(Cloudbeds) (YouTube, Gainsight)",
        "sourceUrl": "https://www.youtube.com/watch?v=cHCdr_kGadw&t=2405s",
        "publishedAt": "2025-11-12",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "replace",
      "path": "/gorgias/factsCheckedAt",
      "value": "2026-10-04"
    },
    {
      "op": "add",
      "path": "/gorgias/strengths/6",
      "value": {
        "id": "gorgias-yt-pajar-bilingual-email",
        "text": "Pajar’s customer-service manager reports that its Gorgias AI Agent, called Robin, had been answering emails in French and English for several months. The customer says: “she's answering emails in French and in English”.",
        "sourceType": "vendor-reported",
        "sourceTitle": "How Pajar Uses AI Agent to Automate 45% of Support Tickets (YouTube, Gorgias)",
        "sourceUrl": "https://www.youtube.com/watch?v=TOVTzEqhF4g&t=152s",
        "publishedAt": "2025-01-28",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/gorgias/strengths/7",
      "value": {
        "id": "gorgias-yt-pajar-human-handoff",
        "text": "Pajar’s customer-service manager reports that Gorgias AI Agent hands the ticket to a human agent when it is unsure of information or a process.",
        "sourceType": "vendor-reported",
        "sourceTitle": "How Pajar Uses AI Agent to Automate 45% of Support Tickets (YouTube, Gorgias)",
        "sourceUrl": "https://www.youtube.com/watch?v=TOVTzEqhF4g&t=172s",
        "publishedAt": "2025-01-28",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/gorgias/strengths/8",
      "value": {
        "id": "gorgias-yt-pajar-combined-automation",
        "text": "Pajar’s customer-service manager reports that AI Agent and Automate together take 45% of its customer-service time; this is a combined-product result, not AI Agent alone. The customer says: “AI agent and automate are taking 45% of our customer service time”.",
        "sourceType": "vendor-reported",
        "sourceTitle": "How Pajar Uses AI Agent to Automate 45% of Support Tickets (YouTube, Gorgias)",
        "sourceUrl": "https://www.youtube.com/watch?v=TOVTzEqhF4g&t=319s",
        "publishedAt": "2025-01-28",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/gorgias/strengths/9",
      "value": {
        "id": "gorgias-yt-pajar-return-links",
        "text": "Pajar’s customer-service manager says Gorgias AI Agent sends customers a Loop return link for the correct Shopify store. The customer says: “the agent is just going to send the loop link directly in the right Shopify store”.",
        "sourceType": "vendor-reported",
        "sourceTitle": "How Pajar Uses AI Agent to Automate 45% of Support Tickets (YouTube, Gorgias)",
        "sourceUrl": "https://www.youtube.com/watch?v=TOVTzEqhF4g&t=448s",
        "publishedAt": "2025-01-28",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/gorgias/strengths/10",
      "value": {
        "id": "gorgias-yt-delta-galil-automation",
        "text": "Delta Galil customer-service director Ajit Patel reports over 60% automation with Gorgias AI handling questions such as order status and return tracking. The customer says: “we're at a 60 plus% rate of automation right now”.",
        "sourceType": "vendor-reported",
        "sourceTitle": "Behind the Inbox Podcast | Inside Delta Galil's 60% automation playbook | Episode 1 (YouTube, Gorgias)",
        "sourceUrl": "https://www.youtube.com/watch?v=bO_9eDSuvzg&t=608s",
        "publishedAt": "2026-07-21",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/gorgias/strengths/11",
      "value": {
        "id": "gorgias-yt-glamnetic-automation",
        "text": "Glamnetic’s CX leaders report a 55% automation rate for their Gorgias AI Agent program, compared with about 25% before its optimization work; more than 60% remains a goal. The customer says: “a very clear 55% automation rate”.",
        "sourceType": "vendor-reported",
        "sourceTitle": "Double the automation rate, same CSAT: Glamnetic's road to 55% (YouTube, Gorgias)",
        "sourceUrl": "https://www.youtube.com/watch?v=h92nTcyTff0&t=549s",
        "publishedAt": "2026-08-13",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/gorgias/strengths/12",
      "value": {
        "id": "gorgias-yt-glamnetic-order-address",
        "text": "Glamnetic CX manager Jackie Briones describes a Gorgias AI Agent ticket that edits the shipping address in Shopify after confirming the order and new address with the customer. The customer says: “then it goes into editing that shipping address within Shopify”.",
        "sourceType": "vendor-reported",
        "sourceTitle": "Double the automation rate, same CSAT: Glamnetic's road to 55% (YouTube, Gorgias)",
        "sourceUrl": "https://www.youtube.com/watch?v=h92nTcyTff0&t=809s",
        "publishedAt": "2026-08-13",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/gorgias/weaknesses/9",
      "value": {
        "id": "gorgias-yt-delta-galil-guidance-errors",
        "text": "Delta Galil senior customer-service manager Nicole Parker reports that initial Gorgias AI guidance passed her tests but subsequently gave incorrect answers to customers, requiring iterative guidance changes and review of low-CSAT tickets. The customer says: “This isn't correct. Why is AI answering like this?”.",
        "sourceType": "vendor-reported",
        "sourceTitle": "Behind the Inbox Podcast | Inside Delta Galil's 60% automation playbook | Episode 1 (YouTube, Gorgias)",
        "sourceUrl": "https://www.youtube.com/watch?v=bO_9eDSuvzg&t=955s",
        "publishedAt": "2026-07-21",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/gorgias/weaknesses/10",
      "value": {
        "id": "gorgias-yt-delta-galil-address-error",
        "text": "Delta Galil’s customer-service leaders report that Gorgias AI gave a corporate address that should not have been given for a licensed brand; Gaia detected the issue and the ecommerce team was asked to correct the source page. The customer says: “it was giving out the corporate address.”.",
        "sourceType": "vendor-reported",
        "sourceTitle": "Behind the Inbox Podcast | Inside Delta Galil's 60% automation playbook | Episode 1 (YouTube, Gorgias)",
        "sourceUrl": "https://www.youtube.com/watch?v=bO_9eDSuvzg&t=1275s",
        "publishedAt": "2026-07-21",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/gorgias/weaknesses/11",
      "value": {
        "id": "gorgias-yt-glamnetic-discount-code-gap",
        "text": "Glamnetic CX manager Jackie Briones says discount-code handling has not yet worked for its Gorgias AI Agent deployment and remains challenging to control. The customer says: “the one thing that just did not work for us”.",
        "sourceType": "vendor-reported",
        "sourceTitle": "Double the automation rate, same CSAT: Glamnetic's road to 55% (YouTube, Gorgias)",
        "sourceUrl": "https://www.youtube.com/watch?v=h92nTcyTff0&t=1193s",
        "publishedAt": "2026-08-13",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/gorgias/weaknesses/12",
      "value": {
        "id": "gorgias-yt-glamnetic-conversion-gap",
        "text": "Glamnetic VP of CX Mia Chapa says its 50/50 product-page AI shopping-assistant experiment increased automation and reduced handovers, but the conversion lift was still not fully there at the interview. The customer says: “the lift is still not fully there”.",
        "sourceType": "vendor-reported",
        "sourceTitle": "Double the automation rate, same CSAT: Glamnetic's road to 55% (YouTube, Gorgias)",
        "sourceUrl": "https://www.youtube.com/watch?v=h92nTcyTff0&t=1376s",
        "publishedAt": "2026-08-13",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/gorgias/weaknesses/13",
      "value": {
        "id": "gorgias-yt-glamnetic-qa-effort",
        "text": "Glamnetic CX manager Jackie Briones describes daily ticket, guidance and action checks for Gorgias AI Agent as tedious but worthwhile; the deployment still requires active human QA. The customer says: “piece by piece on a daily basis. Again, yes, tedious, worth it, 100%.”.",
        "sourceType": "vendor-reported",
        "sourceTitle": "Double the automation rate, same CSAT: Glamnetic's road to 55% (YouTube, Gorgias)",
        "sourceUrl": "https://www.youtube.com/watch?v=h92nTcyTff0&t=1685s",
        "publishedAt": "2026-08-13",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/kore-ai/strengths/17",
      "value": {
        "id": "kore-ai-yt-nova-kbm-slovenian",
        "text": "Nova KBM representatives report choosing Kore.ai because it could support Slovenian within the bank’s ambitious timeframe where competing proposals could not. The customer says: “the ability to support the Slovenian language”.",
        "sourceType": "vendor-reported",
        "sourceTitle": "Nova KBM Offers Seamless Customer Experience with AI-Powered Contact Centers-Kore.ai Client Stories (YouTube, Kore.ai)",
        "sourceUrl": "https://www.youtube.com/watch?v=lMx2cffzm6k&t=41s",
        "publishedAt": "2022-09-23",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/liveperson/strengths/25",
      "value": {
        "id": "liveperson-yt-hard-rock-concierge",
        "text": "In a LivePerson-published customer interview, a Hard Rock International representative reports that its recently launched Roadie conversational concierge bot lets guests request room service and towels and make restaurant dining reservations.",
        "sourceType": "vendor-reported",
        "sourceUrl": "https://www.youtube.com/watch?v=_uw4UwDXgTw&t=66s",
        "sourceTitle": "How Hard Rock Innovates with AI & Digital Transformation Strategies (YouTube, LivePerson)",
        "publishedAt": "2024-04-18",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/lyro/strengths/23",
      "value": {
        "id": "lyro-yt-tidio-internal-quality-review",
        "text": "In Tidio’s own deployment interview, Tidio support practitioners say they sample Lyro conversations monthly and report an answer success rate rising from about 70% to about 83% over a year while maintaining the FAQ. This is vendor self-use, not an external customer result or the separate 58% automation figure in the video title.",
        "sourceType": "vendor-reported",
        "sourceUrl": "https://www.youtube.com/watch?v=LEVv0OpUXOw&t=104s",
        "sourceTitle": "Tidio Support Achieves 58% Automation with Lyro AI Chatbot - New Claude 3.0 Model (YouTube, Tidio)",
        "publishedAt": "2024-05-28",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/lyro/weaknesses/22",
      "value": {
        "id": "lyro-yt-tidio-offline-handoff",
        "text": "In Tidio’s own deployment interview, Tidio support practitioners report that Lyro previously promised human handoff when nobody was available on chat, producing negative satisfaction ratings; they say they subsequently fixed the conversation flow and satisfaction did not drop. This describes a historical, reportedly resolved issue in Tidio’s internal deployment.",
        "sourceType": "vendor-reported",
        "sourceUrl": "https://www.youtube.com/watch?v=LEVv0OpUXOw&t=345s",
        "sourceTitle": "Tidio Support Achieves 58% Automation with Lyro AI Chatbot - New Claude 3.0 Model (YouTube, Tidio)",
        "publishedAt": "2024-05-28",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/lyro/weaknesses/23",
      "value": {
        "id": "lyro-yt-tidio-knowledge-maintenance",
        "text": "In Tidio’s own deployment interview, Tidio support practitioners say combining historical conversations, help articles and manually written Q&A made Lyro’s knowledge base hard to manage because information repeated across sources; they recommend a knowledge owner and one source of truth, and describe correcting wrong answers found in escalations and periodic checks.",
        "sourceType": "vendor-reported",
        "sourceUrl": "https://www.youtube.com/watch?v=LEVv0OpUXOw&t=187s",
        "sourceTitle": "Tidio Support Achieves 58% Automation with Lyro AI Chatbot - New Claude 3.0 Model (YouTube, Tidio)",
        "publishedAt": "2024-05-28",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/nice-cognigy/strengths/23",
      "value": {
        "id": "nice-cognigy-yt-bosch-technical-answers",
        "text": "In a NiCE Cognigy-published customer interview, Michael, senior digital product owner for Bosch Mobility Aftermarket in North America, reports that Bosch Answers reduced customers’ time spent searching documents or piecing together technical answers by 80%; he reports about 12,000 conversations to date. This is Bosch’s deployment-specific report, not an independently measured benchmark.",
        "sourceType": "vendor-reported",
        "sourceUrl": "https://www.youtube.com/watch?v=QiPqlc4o_-A&t=106s",
        "sourceTitle": "Bosch: Agentic AI Agent for Reducing Complaints by 70% (YouTube, NiCE Cognigy)",
        "publishedAt": "2026-06-17",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/nice-cognigy/weaknesses/10",
      "value": {
        "id": "nice-cognigy-yt-bosch-stale-documentation",
        "text": "In a NiCE Cognigy-published customer interview, Michael of Bosch Mobility Aftermarket says incorrect answers from its AI agent can reflect outdated source data, and that operating the agent led Bosch to revisit and improve its documentation.",
        "sourceType": "vendor-reported",
        "sourceUrl": "https://www.youtube.com/watch?v=QiPqlc4o_-A&t=127s",
        "sourceTitle": "Bosch: Agentic AI Agent for Reducing Complaints by 70% (YouTube, NiCE Cognigy)",
        "publishedAt": "2026-06-17",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/polyai/strengths/15",
      "value": {
        "id": "polyai-yt-melting-pot-after-hours",
        "text": "In a PolyAI-published customer interview, Dan Mullins, The Melting Pot’s manager of technology, says its PolyAI voice assistant answers opening-hours and reservation calls and attributes more than $300,000 in sales over the prior year to calls outside standard hours: \"over the last year it well clears $300,000 in sales\". The figure is his deployment report, not audited incremental revenue.",
        "sourceType": "vendor-reported",
        "sourceTitle": "How the Melting Pot generates over $300k from after hours bookings with PolyAI (YouTube, PolyAI)",
        "sourceUrl": "https://www.youtube.com/watch?v=-cqS5Vmel5k&t=85s",
        "publishedAt": "2026-01-23",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/polyai/strengths/16",
      "value": {
        "id": "polyai-yt-melting-pot-opentable",
        "text": "In a PolyAI-published customer interview, Dan Mullins of The Melting Pot describes its voice assistant’s integration with OpenTable and telephone systems, saying \"The integration with OpenTable, of course, is make or break\" and that it supports reservations; he subsequently describes OpenTable experiences/package bookings that previously needed restaurant staff.",
        "sourceType": "vendor-reported",
        "sourceTitle": "How the Melting Pot generates over $300k from after hours bookings with PolyAI (YouTube, PolyAI)",
        "sourceUrl": "https://www.youtube.com/watch?v=-cqS5Vmel5k&t=122s",
        "publishedAt": "2026-01-23",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/polyai/strengths/17",
      "value": {
        "id": "polyai-yt-audibel-abandonment",
        "text": "In a PolyAI-published customer interview, Audibel representatives report that after deploying PolyAI their \"abandonment rate dropped from 43% down to 2%\" within days and hold times fell from 12–15 minutes to two minutes or less. This is a customer-reported deployment result, not an independent measurement.",
        "sourceType": "vendor-reported",
        "sourceTitle": "How Audibel reduced abandonment rates by 44% and increased revenue by deploying PolyAI (YouTube, PolyAI)",
        "sourceUrl": "https://www.youtube.com/watch?v=YjG7fOtT8ok&t=85s",
        "publishedAt": "2025-11-14",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/polyai/weaknesses/12",
      "value": {
        "id": "polyai-yt-audibel-verification-boundary",
        "text": "In a PolyAI-published customer interview, Audibel representatives explain that in their hearing-healthcare deployment there are questions the AI agent cannot ask, so \"that phone call is handed over to a live agent\" for patient verification and compliance. This describes Audibel’s deployment boundary, not a finding that PolyAI is noncompliant.",
        "sourceType": "vendor-reported",
        "sourceTitle": "How Audibel reduced abandonment rates by 44% and increased revenue by deploying PolyAI (YouTube, PolyAI)",
        "sourceUrl": "https://www.youtube.com/watch?v=YjG7fOtT8ok&t=64s",
        "publishedAt": "2025-11-14",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/zendesk-ai-agents/strengths/15",
      "value": {
        "id": "zendesk-ai-agents-yt-lucid-pilot-optimization",
        "text": "In a Zendesk-published customer interview, Keyvan Sadigh of Lucid Software says the team manually built AI agents across use cases and improved pilot accuracy from 50–60% to above 90% after a few weeks of effort and optimization. He does not define the accuracy metric or provide an evaluation sample, so this is a customer-reported pilot result rather than an independently verified resolution rate.",
        "sourceType": "vendor-reported",
        "sourceTitle": "Lucid Software drives AI-powered support transformation with Zendesk | Zendesk customer story (YouTube, Zendesk)",
        "sourceUrl": "https://www.youtube.com/watch?v=eQa1OaC_uxY&t=106s",
        "publishedAt": "2026-09-10",
        "checkedAt": "2026-10-04"
      }
    },
    {
      "op": "add",
      "path": "/zendesk-ai-agents/strengths/16",
      "value": {
        "id": "zendesk-ai-agents-yt-qantas-automated-resolutions",
        "text": "In a Zendesk-published customer interview, Luca Giaccardi of Qantas Loyalty reports a 30–35% initial automated-resolution target, 60% within the first two weeks and peaks of 85%; general-knowledge inquiries occasionally reached the high 90s. He conditions these results on providing the information, clear intent and ongoing case review; they are not a general Zendesk benchmark.",
        "sourceType": "vendor-reported",
        "sourceTitle": "Qantas Loyalty soars past automation goals with Zendesk AI Agents | Zendesk customer story (YouTube, Zendesk)",
        "sourceUrl": "https://www.youtube.com/watch?v=uJe6u8Sqm88&t=127s",
        "publishedAt": "2026-06-26",
        "checkedAt": "2026-10-04"
      }
    }
  ],
  "digest": "sha256:f1856137d8e88313c68b207d700629f0d328eed8962e84a121f3bb474dc6d6f9"
}
