{"id":2685,"date":"2026-09-04T05:05:24","date_gmt":"2026-09-04T05:05:24","guid":{"rendered":"https:\/\/www.plura.ai\/articles\/voicemail-detection-improves-agent-utilization"},"modified":"2026-09-04T05:05:24","modified_gmt":"2026-09-04T05:05:24","slug":"voicemail-detection-improves-agent-utilization","status":"publish","type":"post","link":"https:\/\/www.plura.ai\/articles\/voicemail-detection-improves-agent-utilization","title":{"rendered":"How Voicemail Detection Improves Agent Utilization"},"content":{"rendered":"<p><em>Written by: Matt Beucler, CEO, Plura AI<\/em><\/p>\n<h2 id=\"key-takeaways\">Key Takeaways for Outbound Leaders<\/h2>\n<ul>\n<li>Voicemail detection (AMD) filters out answering machines so agents only connect with live humans, removing 15-30 seconds of dead time per machine call.<\/li>\n<li>Modern AI-based AMD analyzes acoustic signatures in real time, improving speed and accuracy compared with traditional rule-based systems.<\/li>\n<li>AMD lifts agent utilization across five areas: more talk time, more conversations per hour, less idle time, lower cost per contact, and better morale.<\/li>\n<li>Key metrics for tracking AMD impact include talk time percentage, conversations per hour, and agent occupancy, with utilization gains that can reach 100% or more.<\/li>\n<li>Plura AI&#8217;s <a href=\"https:\/\/plura.ai\/ai-predictive-dialer\" target=\"_blank\" rel=\"noindex nofollow\"><strong>AI Predictive Dialer<\/strong><\/a> delivers carrier-grade AMD on its own FCC-licensed infrastructure to extend these utilization gains across every campaign.<\/li>\n<\/ul>\n<h2>How Voicemail Detection Works in Your Dialer<\/h2>\n<p>Answering machine detection (AMD) is a call-progress analysis layer that runs between the moment a call is answered and the moment an agent is connected. Its job is to classify what picked up: a live human or a machine.<\/p>\n<p>Traditional AMD relied on simple audio heuristics, such as measuring the length of the initial greeting, listening for a beep that often follows a voicemail prompt, and analyzing silence patterns. A short greeting followed by a beep signals a machine. A short greeting with no beep, followed by ambient noise, signals a human.<\/p>\n<p>Modern AI-based AMD adds a machine-learning layer on top of those heuristics. The system analyzes the acoustic signature of the greeting in real time and compares it against trained models built from millions of classified call samples. This allows the system to distinguish a live human who answers with a long greeting from a voicemail system that opens with a short one. As a result, classification errors fall compared with older rule-based systems.<\/p>\n<p>The goal of AMD is narrow and specific. The dialer connects agents only to live answers. Every call that reaches a machine is either dropped silently, routed to a voicemail-drop workflow, or logged for callback, depending on campaign configuration. The agent never hears it.<\/p>\n<h2>How Voicemail Detection Lifts Agent Utilization<\/h2>\n<p>Voicemail detection protects agent time by stripping out unproductive machine connects before they ever reach the floor. Without AMD, every answered call that reaches a machine costs an agent 15 to 30 seconds of dead time across the ring cycle, recorded greeting, beep, and manual disconnect. At high call volumes, that dead time accumulates into a significant share of the shift.<\/p>\n<p>An agent who spends a third of their answered calls on machines spends a meaningful portion of their call-handling time on interactions that never had a chance to convert. That drag on productivity shows up directly in talk time, conversations per hour, and cost per contact.<\/p>\n<p>With AMD, the dialer classifies the call before the agent is connected. Machines are filtered out at the platform level. The agent&#8217;s queue fills only with live answers.<\/p>\n<figure style=\"text-align: center\"><img decoding=\"async\" src=\"https:\/\/cdn.aigrowthmarketer.co\/1779338793506-2d33c5dff8e8.png\" alt=\"Plura Predictive Dialer dashboard displaying AI-powered outbound call pacing, transfer analysis, and dialing performance insights.\" style=\"max-height: 500px\" loading=\"lazy\"><figcaption><em>Plura Predictive Dialer automates outbound calling with AI-powered pacing, transfer optimization, and real-time performance analytics.<\/em><\/figcaption><\/figure>\n<p>The utilization gains from AMD operate across five dimensions:<\/p>\n<ul>\n<li><strong>Increased talk time:<\/strong> Agents spend more minutes per hour in live conversation because dead-end calls never reach their headset.<\/li>\n<li><strong>More conversations per hour:<\/strong> Filtering machines means the dialer cycles through dead ends faster and delivers more live connects per shift.<\/li>\n<li><strong>Reduced idle time:<\/strong> Agents are not waiting through rings and recorded greetings before the next dial. The dialer moves immediately to the next number.<\/li>\n<li><strong>Lower cost per contact:<\/strong> When talk time rises, the cost of each completed live conversation falls in line with that improvement.<\/li>\n<li><strong>Improved agent morale:<\/strong> Agents who spend their shift talking to prospects rather than hanging up on machines report higher job satisfaction, which can help reduce attrition. Outbound call center agents typically see annual turnover rates exceeding 50%, sometimes reaching 60-100%, while inbound centers see 30-45%.<\/li>\n<\/ul>\n<h2>The Accuracy vs. Speed Tradeoff in AMD<\/h2>\n<p>AMD introduces a tradeoff that every dialer administrator needs to understand before tuning settings. Accuracy and speed pull in opposite directions.<\/p>\n<p>Faster detection generally means the system makes its live-or-machine classification earlier in the greeting. When the fast-path heuristic for short greetings followed by silence can run, detection happens quickly. If voice-activity-detection boundaries are missing or the transcript is empty, detection can be delayed to the timeout, often around 20 seconds. Faster decisions reduce the delay between answer and agent connection, which protects talk time. Earlier classification can also be less accurate because the system has less audio to analyze. The result can be a higher false positive rate, where live humans get classified as machines and the call drops before an agent connects. A prospect who picks up and hears nothing experiences a wasted call and a damaged relationship.<\/p>\n<p>In traditional timing-based AMD, higher accuracy generally requires the system to wait longer to collect more audio before classifying. Modern AI-based detection can reach high accuracy with much shorter analysis times. That approach reduces false positives but can add a small connection delay. The agent connects a fraction of a second later. Callers rarely notice this delay, but across thousands of calls it can add up to measurable talk-time loss.<\/p>\n<p>Modern AI-based AMD compresses this tradeoff by improving classification accuracy at shorter audio windows. The machine-learning models recognize machine signatures earlier and with greater confidence than rule-based systems. The system can be tuned for both speed and accuracy at the same time rather than trading one for the other. The tradeoff does not disappear entirely, but it narrows significantly. Tuning still matters, and the right threshold depends on the call population, the voicemail systems in use, and the campaign&#8217;s tolerance for false positives versus connection delay.<\/p>\n<h2>Measuring the Impact on Agent Utilization<\/h2>\n<p>Agent utilization in an outbound contact center rests on three core metrics.<\/p>\n<ul>\n<li><strong>Talk time:<\/strong> The percentage of an agent&#8217;s shift spent in live conversation. A human agent in a traditional contact center often operates at 40% talk utilization, meaning 24 minutes of every hour is live conversation and the remaining 36 minutes is ringing, waiting, and administrative time.<\/li>\n<li><strong>Conversations per hour:<\/strong> The number of live connects an agent completes in a given hour. This metric depends directly on talk time and average handle time.<\/li>\n<li><strong>Agent occupancy:<\/strong> The percentage of time an agent is either in a call or in after-call work, as opposed to idle. High occupancy with low talk time indicates the agent is busy but not productive.<\/li>\n<\/ul>\n<p>An illustrative calculation shows the impact. An agent without AMD who averages 12 minutes of live talk per hour is running at 20% talk utilization. If AMD doubles the live-connect rate by filtering machines, that agent reaches 24 minutes of talk per hour. Utilization improves by 100% with no change in headcount or shift length.<sup data-disclaimer-id=\"24\" data-disclaimer-index=\"3\">3<\/sup><\/p>\n<p>The cost impact compounds quickly. <a href=\"https:\/\/plura.ai\/calculator\" target=\"_blank\">Plura&#8217;s ROI calculator<\/a> models a 15-agent operation at $20 per hour with standard taxes, benefits, and commissions at 40% talk utilization, producing a monthly cost of $60,000. Plura agents operating at 100% talk utilization with 6 agents replacing 15 humans cost $14,400 per month, a monthly savings of $45,600. Over 12 months, that gap reaches $547,200.<sup data-disclaimer-id=\"24\" data-disclaimer-index=\"3\">3<\/sup> To see how these figures scale to your own operation, run your numbers through <a href=\"https:\/\/plura.ai\/calculator\" target=\"_blank\">Plura&#8217;s ROI calculator<\/a>.<\/p>\n<p><a href=\"https:\/\/plura.ai\/plura-webchat\" target=\"_blank\" rel=\"noindex nofollow\"><strong>See AMD-driven utilization gains in action<\/strong><\/a> with a live demo tailored to your call volume and team size.<\/p>\n<h2>Best Practices for Tuning Voicemail Detection<\/h2>\n<p>Voicemail detection settings that perform well for one campaign can create unacceptable false positive rates on another. The call population matters. A campaign targeting mobile numbers in a market with aggressive carrier voicemail systems behaves differently from a campaign targeting landlines in a business-to-business context.<\/p>\n<p>Dialer administrators can use the following practical tuning guidelines.<\/p>\n<ul>\n<li><strong>Set AMD sensitivity thresholds based on your false positive tolerance.<\/strong> If your campaign cannot afford to drop live humans, start with a conservative threshold that favors accuracy over speed. Tighten it as you gather data on your specific call population.<\/li>\n<li><strong>Monitor false positive rates continuously.<\/strong> Track the percentage of calls classified as machines where the prospect later reports they answered. A false positive rate above 2 to 3% signals a need to adjust thresholds.<\/li>\n<li><strong>Adjust based on call answer patterns.<\/strong> In markets with high mobile penetration, voicemail greetings tend to be shorter and more variable. AI-based AMD handles this better than rule-based systems because the model generalizes across greeting styles instead of relying on fixed timing rules.<\/li>\n<li><strong>Use AI-based detection for better accuracy at scale.<\/strong> Rule-based AMD degrades as carrier voicemail systems evolve. AI-based AMD improves over time as the model trains on new call outcomes.<\/li>\n<\/ul>\n<p>Plura&#8217;s <a href=\"https:\/\/plura.ai\/ai-predictive-dialer\" target=\"_blank\" rel=\"noindex nofollow\">AI Predictive Dialer<\/a> includes advanced AMD that learns from call outcomes. Classification accuracy improves continuously without manual threshold adjustments after initial configuration.<\/p>\n<h2>Voicemail Detection vs. Live Answer Detection<\/h2>\n<p>Voicemail detection and live answer detection (LAD) address related but distinct classification problems. AMD focuses on distinguishing live humans from machines. LAD covers a broader set of call outcomes.<\/p>\n<p>In practice, systems such as LiveKit&#8217;s Answering Machine Detection classify an outbound call into categories like human, machine-ivr, machine-vm, machine-unavailable, or uncertain.<sup data-disclaimer-id=\"25\" data-disclaimer-index=\"4\">4<\/sup> AMD handles the live-human-versus-machine decision within that broader framework.<\/p>\n<p>The distinction matters operationally because a dialer that only detects voicemails still needs to handle fax tones, special information tones (SIT) for disconnected numbers, and operator intercepts without routing them to agents. Plura&#8217;s AI Predictive Dialer is designed to handle the full range of call outcomes so agents receive only verified live human connects.<\/p>\n<h2>Common AMD Pitfalls and How to Avoid Them<\/h2>\n<p>Operators who deploy AMD without ongoing monitoring tend to see the same issues repeat across campaigns.<\/p>\n<ul>\n<li><strong>Inconsistent detection across carrier environments.<\/strong> A threshold tuned for one carrier&#8217;s voicemail system may misclassify calls on another. Segment AMD performance reporting by carrier or number prefix and tune thresholds per segment where volume justifies it.<\/li>\n<li><strong>False positives that damage prospect relationships.<\/strong> A prospect who picks up and hears silence or a click before the call drops is less likely to answer the next attempt. Monitor false positive rates weekly and treat any spike as a campaign-level alert.<\/li>\n<li><strong>Over-reliance on AMD without measuring downstream outcomes.<\/strong> AMD improves utilization, but utilization is an input metric. The output metric is conversations that convert. If AMD filters aggressively and talk time rises while conversion rate stays flat, the false positive rate may be eliminating high-intent prospects. Measure both utilization and conversion.<\/li>\n<\/ul>\n<p>A dialer with carrier-grade AMD, real-time call outcome logging, and configurable sensitivity thresholds helps surface these issues quickly. The data becomes visible and actionable instead of buried in raw call logs.<\/p>\n<h2>Why Plura AI&#8217;s AI Predictive Dialer Fits Enterprise Operations<\/h2>\n<p>Most predictive dialers treat AMD as a bolt-on feature that runs on top of rented CPaaS infrastructure from providers such as Twilio.<sup data-disclaimer-id=\"25\" data-disclaimer-index=\"4\">4<\/sup> That architecture creates two problems: the vendor cannot enforce AMD at the carrier level because they do not own the carrier, and the detection model is static, trained on a fixed dataset that does not adapt as carrier voicemail systems evolve.<\/p>\n<p>Plura&#8217;s <a href=\"https:\/\/plura.ai\/ai-predictive-dialer\" target=\"_blank\" rel=\"noindex nofollow\">AI Predictive Dialer<\/a> is built on Plura&#8217;s own FCC-licensed audio bridging carrier. AMD runs at the carrier level, not as a software layer on top of rented infrastructure. Branded caller ID is issued directly under Plura&#8217;s carrier identity, and STIR\/SHAKEN authentication runs on every outbound call. Real-time DNC (Do Not Call) scrubbing and TCPA (Telephone Consumer Protection Act) compliance support sit in the core platform rather than as third-party add-ons.<sup data-disclaimer-ids=\"22,23\" data-disclaimer-indexes=\"1,2\">1,2<\/sup> Customers remain responsible for their own regulatory obligations and should consult qualified counsel for specific guidance.<\/p>\n<p>The dialer uses stateful conversion signals to decide who to call next, including historical answer rates, prior conversation outcomes, and prior offer-acceptance data from Plura&#8217;s Stateful Conversation Database. AMD feeds into that prioritization model as one signal among many. The result is a dialer that maximizes talk time and focuses agents on contacts most likely to convert, not just contacts most likely to answer.<\/p>\n<figure style=\"text-align: center\"><img decoding=\"async\" src=\"https:\/\/cdn.aigrowthmarketer.co\/1779339309900-eefcb08741d1.png\" alt=\"Plura Predictive Dialer dashboard showing AI-powered outbound dialing, intelligent call routing, and performance analytics.\" style=\"max-height: 500px\" loading=\"lazy\"><figcaption><em>Plura Predictive Dialer uses AI-powered outbound dialing, intelligent routing, and real-time analytics to maximize call performance.<\/em><\/figcaption><\/figure>\n<p>For a 100-seat contact center, <a href=\"https:\/\/www.plura.ai\/guides\/ai-communications-strategy\" target=\"_blank\">traditional operations can cost $4 million to $7 million annually<\/a>, while AI-powered communications using platforms like Plura often fall in the $300,000 to $700,000 range.<sup data-disclaimer-id=\"24\" data-disclaimer-index=\"3\">3<\/sup> AMD-driven utilization improvement represents a significant share of that difference.<\/p>\n<p><a href=\"https:\/\/plura.ai\/plura-webchat\" target=\"_blank\" rel=\"noindex nofollow\"><strong>Walk through AMD configuration in a live session<\/strong><\/a> built around your campaign structure and routing rules.<\/p>\n<h2>Frequently Asked Questions<\/h2>\n<h3>How much agent time does voicemail detection actually save?<\/h3>\n<p>The savings depend on the voicemail rate in your call population and your current AMD configuration. In an outbound campaign where 40% of answered calls reach voicemail, each such machine call costs an agent the same 15-30 seconds of dead time described earlier. At high call volumes, that wasted time accumulates quickly. With effective AMD, that time shifts back into live conversation.<\/p>\n<p>The practical improvement in talk utilization can be substantial. Some sources report gains from 30-40% up to 75-85% with predictive dialing, while others note over 40 minutes per hour of talk time with AI-assisted transfers. AI agents operating on Plura&#8217;s platform run at 100% talk utilization because the dialer never connects an agent to a non-live call.<\/p>\n<h3>What is a good false positive rate for voicemail detection?<\/h3>\n<p>A false positive in AMD is a live human classified as a machine and dropped before an agent connects. There is no universal standard, and acceptable rates vary by business context and cost tradeoffs. Some regulatory frameworks reference abandonment rates around 3%, and many call centers operate at 2.5-3% abandonment, but the appropriate false positive threshold depends on your campaign&#8217;s tolerance for dropped calls versus wasted agent time.<\/p>\n<p>A false positive rate below 1% is achievable with modern AI-based detection on a well-tuned campaign, as shown in published benchmarks such as DeepTrust at 0.98% FPR and Lastline reporting zero false positives in an NSS Labs study. Performance varies by methodology and context. The cost of a false positive includes both the lost call and the prospect&#8217;s experience of picking up and hearing nothing, which reduces the probability they answer the next attempt. False positive rate should sit alongside talk time and conversations per hour as a first-class campaign metric.<\/p>\n<h3>How quickly can AMD improve agent utilization after implementation?<\/h3>\n<p>AMD improvements appear immediately at the call level. The first call the dialer filters is the first call an agent does not waste time on. At the campaign level, measurable utilization gains typically appear quickly, often within the first day of operation.<\/p>\n<p>The more meaningful timeline concerns tuning AMD to the optimal threshold for a specific call population. Initial tuning usually requires collecting live campaign data to identify the false positive rate at the default threshold and adjust accordingly. AI-based AMD systems that learn from call outcomes continue improving beyond that initial window without manual intervention.<\/p>\n<h3>Does AMD work differently for AI agents versus human agents?<\/h3>\n<p>For human agents, AMD functions as a filter. Machines drop before the agent connects, which protects talk time. For AI agents, the dynamic shifts.<\/p>\n<p>An AI agent can handle both live humans and voicemail drops within the same call flow, so AMD serves a routing function rather than a pure filter. When the dialer detects a machine, the AI can execute a pre-recorded voicemail drop and move to the next number. When the dialer detects a live human, the AI agent engages in a full conversation. This approach lets AI agents extract value from calls that would be pure waste for human agents while still prioritizing live conversations for the highest-value interactions. Plura&#8217;s AI Predictive Dialer supports both routing paths within a single campaign workflow.<\/p>\n<h2>Conclusion and Next Steps<\/h2>\n<p>Voicemail detection operates as a core utilization lever for outbound operations. Without AMD, agents in high-volume environments burn a measurable share of every shift on dead ends such as rings, recorded greetings, and manual disconnects that create no revenue and frustrate the team. With AMD tuned correctly, those dead ends filter out at the platform level and agents spend their shift in live conversation.<\/p>\n<p>The measurement framework stays simple. Track talk time percentage, conversations per hour, and false positive rate. If talk time sits below roughly 40% on a human-agent floor, AMD configuration deserves a close look. If false positive rate stays high, threshold adjustment may be overdue. When human agents run at 40% utilization against a platform that can deliver 100%, the ROI gap becomes clear.<\/p>\n<p>Check your specific numbers with <a href=\"https:\/\/plura.ai\/calculator\" target=\"_blank\">Plura&#8217;s ROI calculator<\/a>. Compare plans and rates side by side at <a href=\"https:\/\/plura.ai\/pricing\" target=\"_blank\">plura.ai\/pricing<\/a>.<\/p>\n<p><a href=\"https:\/\/www.plura.ai\/plura-webchat\" target=\"_blank\"><strong>Request a live demo<\/strong><\/a> to review AMD configuration, utilization reporting, and carrier-grade compliance support for your outbound operation.<\/p>\n<hr data-disclaimer-divider=\"true\">\n<div data-disclaimer-footer=\"true\">\n<p data-disclaimer-id=\"22\" data-disclaimer-type=\"content_based\"><sup data-disclaimer-index=\"1\">1<\/sup> Plura AI maintains SOC 2, HIPAA, ISO, and GDPR posture as part of its platform infrastructure. References to compliance frameworks in this article describe Plura\u2019s platform capabilities and do not constitute a guarantee that any customer using Plura will themselves be compliant with applicable laws or standards. Customers remain solely responsible for their own regulatory obligations, certifications, consent management, recordkeeping, and the claims they make to their own end users. Consult qualified legal counsel for guidance specific to your use case.<\/p>\n<p data-disclaimer-id=\"23\" data-disclaimer-type=\"content_based\"><sup data-disclaimer-index=\"2\">2<\/sup> This article describes regulatory frameworks at a general level and does not constitute legal advice. Laws and regulations vary by jurisdiction, change over time, and apply differently depending on facts and circumstances. Readers should consult qualified legal counsel before making compliance decisions.<\/p>\n<p data-disclaimer-id=\"24\" data-disclaimer-type=\"content_based\"><sup data-disclaimer-index=\"3\">3<\/sup> Performance figures, customer outcomes, and industry statistics referenced in this article are drawn from cited third-party sources or Plura customer case studies. Individual results vary based on implementation, use case, industry, audience, and execution. Past or aggregate performance is not a guarantee of future results.<\/p>\n<p data-disclaimer-id=\"25\" data-disclaimer-type=\"content_based\"><sup data-disclaimer-index=\"4\">4<\/sup> References to third-party products, services, companies, or research are made for informational and comparative purposes only. Plura AI is not affiliated with, endorsed by, or sponsored by any third party named in this article unless explicitly stated. Trademarks and product names referenced remain the property of their respective owners.<\/p>\n<p data-disclaimer-id=\"21\" data-disclaimer-type=\"fixed\">This article is provided for informational purposes only and reflects Plura AI\u2019s understanding at the time of publication. Product capabilities, integrations, and specifications are subject to change. For the most current information, visit plura.ai.<\/p>\n<p data-disclaimer-id=\"27\" data-disclaimer-type=\"fixed\">This article was produced with the assistance of AI tools and reviewed by Plura AI prior to publication.<\/p>\n<\/div>\n<section data-read-next=\"true\">\n<h2>Read Next<\/h2>\n<ul>\n<li><a href=\"https:\/\/www.plura.ai\/articles\/predictive-dialer-voicemail-detection\" target=\"_blank\">Predictive Dialer Voicemail Detection: How AMD Works<\/a><\/li>\n<li><a href=\"https:\/\/www.plura.ai\/articles\/live-answer-vs-voicemail-detection\" target=\"_blank\">Live Answer vs Voicemail Detection in AI Predictive Dialers<\/a><\/li>\n<li><a href=\"https:\/\/www.plura.ai\/articles\/voicemail-detection-integration\" target=\"_blank\">Voicemail Detection Integration for Contact Centers<\/a><\/li>\n<li><a href=\"https:\/\/www.plura.ai\/articles\/voicemail-detection-best-practices\" target=\"_blank\">Voicemail Detection Best Practices for AI Predictive Dialers<\/a><\/li>\n<li><a href=\"https:\/\/www.plura.ai\/articles\/voicemail-detection-voice-ai\" target=\"_blank\">What Is Voicemail Detection in Voice AI?<\/a><\/li>\n<\/ul>\n<\/section>\n","protected":false},"excerpt":{"rendered":"<p>See how AMD keeps agents on live calls and off voicemail. Plura AI&#8217;s predictive dialer is built for enterprise outbound teams that need real results.<\/p>\n","protected":false},"author":106,"featured_media":2684,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"inline_featured_image":false,"footnotes":""},"categories":[2],"tags":[],"class_list":["post-2685","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai-contact-centers"],"_links":{"self":[{"href":"https:\/\/www.plura.ai\/articles\/wp-json\/wp\/v2\/posts\/2685","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.plura.ai\/articles\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.plura.ai\/articles\/wp-json\/wp\/v2\/types\/post"}],"replies":[{"embeddable":true,"href":"https:\/\/www.plura.ai\/articles\/wp-json\/wp\/v2\/comments?post=2685"}],"version-history":[{"count":0,"href":"https:\/\/www.plura.ai\/articles\/wp-json\/wp\/v2\/posts\/2685\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.plura.ai\/articles\/wp-json\/wp\/v2\/media\/2684"}],"wp:attachment":[{"href":"https:\/\/www.plura.ai\/articles\/wp-json\/wp\/v2\/media?parent=2685"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.plura.ai\/articles\/wp-json\/wp\/v2\/categories?post=2685"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.plura.ai\/articles\/wp-json\/wp\/v2\/tags?post=2685"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}