Meta's Muse is already calling contact centers, and Five9 finds trust in AI depends on a human in reach - Weekly

Last week the industry was drafting standards for personal AI agents that act for customers. This week the data showed those agents are already on the phone. In a CX Today analysis published on 15 Oct, Pindrop said it detected the first calls from Meta's Muse on 9 Sep, one week before Meta announced its outbound call beta, and that weekly Muse call volume is now eight times the level of that first week. Financial institutions receive the most Muse calls, followed by healthcare and retail organizations. Pindrop found that 47 percent of Muse calls navigate IVR menus with keypad digits and 64 percent reach a human agent, after a median of two minutes and one second in the phone menu and two minutes and 18 seconds with the person who picks up.

The same piece quotes NiCE, which notes that Muse reached No. 1 among free apps in Apple's U.S. App Store a week after launch and was downloaded 2.8 million times in the United States and Canada within 12 days. NiCE's Neeraj Verma put the problem this way: "Neither blocking nor welcoming is a strategy on its own. Enterprises need a controlled front door that can tell who or what is engaging." Retailers have already split. Amazon has blocked Muse from shopping on its platform, while Shopify integrated it through Shop Pay.

Five9 chief marketing and growth officer Jay Lee made the argument that will matter most for staffing plans: "If an AI agent will sit on hold for you, a lot of calls that never got made start getting made." In other words, personal agents could add contact volume rather than remove it. Pindrop's wider numbers show why a simple block list will not work. It identified 727,000 confirmed AI agent interactions in two months and labeled 94 percent of them benign and 6 percent malicious, a mix that runs from personal assistants and payer bots to account takeover attempts. Recognizing a synthetic voice is not the same as knowing that the customer authorized the call.

For buyers of CCaaS and call center AI software, the practical questions are now concrete: how the platform will forecast machine-initiated volume, which actions a personal agent may take, and how authentication changes for high-risk requests.

Five9 survey: trust in AI depends on a person within reach

On 14 Oct, Five9 released its 2026 Consumer CX Report, based on a Hanover Research survey of 3,000 consumers in the U.S., U.K., and Germany. The headline finding is that trust that AI can resolve an issue rises from 26 percent to 55 percent when a human option is visible. Patience is short. Ninety-two percent of consumers give AI less than 10 minutes before asking for a person, 71 percent say it is very or extremely important to know when they are talking to an AI agent, and 48 percent say they try to use AI as little as possible.

CMSWire noted that password resets and order status checks are the only tasks where more than a third of consumers accept AI working alone, and that 46 percent say a bot that repeatedly misunderstood them is what pushes them to a person. CX Dive reported from the same data that 46 percent rated their latest AI or chatbot experience positively, compared with about two thirds for live phone, email, and in-person support.

A separate Zingtree survey of more than 1,000 U.S. consumers, covered by CX Dive, points the same way. Half of respondents said their complex issues were never resolved, and a third spent multiple days trying before giving up. Analysts quoted in the piece blamed context lost between AI and human agents. Zingtree research also ran in last week's issue.

Regulators are writing the handoff into rules. CXP Asia reported on 16 Oct that China has introduced its first national standard for coordination between AI systems and human customer service agents. AI systems must offer an easy way to reach a person, pass customer information along so people do not repeat themselves, and label AI-generated answers. For prices, discounts, refunds, compensation, or contract changes, AI may explain the rules and open a service request, but final confirmation has to come from a human agent or a standardized procedure.

Vendors are building that middle ground. At CCW Nashville, Sanas launched Supervised AI, which lets one supervisor watch up to four AI-led calls at once, guide the AI privately, take over mid-call, and hand control back without the customer repeating anything. Customers are told they are speaking with AI and that a supervisor is available. It is available by invitation, with broader availability planned for December 2026.

Salesforce takes its Contact Center to Europe and buys Loops

In September Salesforce said its native Contact Center would reach the U.K. and EU in October. It has now set the date. A Service Cloud release post on 13 Oct said the product would become generally available on 16 Oct across more than 30 European countries, including the U.K., Germany, France, Italy, and Spain, with Australia and New Zealand following in November. New headless skills let teams set up a contact center through AI coding assistants without opening the Salesforce UI. Its Casey help agent gains self-optimizing features in November, with a human approving every change. Service Rep Assistant adds agentic Slack swarming in open beta, and workforce management gets agentic recommendations. Salesforce also introduced Core, Advanced, and Max Service editions that bundle 500,000, 1 million, and 2.75 million Flex Credits.

The same week, Salesforce signed a definitive agreement to acquire Loops, an AI analytics startup that monitors business metrics, flags anomalies, and explains what drove a change. CX Today reported that Salesforce plans to fold the technology and team into Tableau after the close. CX Network's funding roundup counted Loops as Salesforce's second startup deal in two weeks, after its agreement to buy Listen Labs on 29 Sep.

Admins have something to check, too. Tech Times reported that with Winter '27 now generally available, Agentforce runs by default in eligible Enterprise, Performance, Unlimited, and Developer Edition orgs and the Setup toggle to switch it off is gone. Because agents inherit the permissions of the user contexts they run under, the piece argues that old access grants are now an audit item.

QA and conversation intelligence get another round of launches

Kustomer launched Conversation Intelligence and AI QA on 14 Oct. Conversation Intelligence uses classifiers, scores, and extractors to turn conversations into structured data that can drive routing and workflows. AI QA scores every conversation against criteria the business sets, with weights, auto-fail rules, and visible reasoning. Both are available now, and Turo and Loaf are quoted as customers.

CallMiner expanded AI Assist so it can reason over CallMiner's scores, measures, and classifiers and back each finding with counts, trends, and contact references. It also redesigned RealTime Command Post so supervisors can find live conversations that need attention. British Gas is quoted as an early user, and CallMiner plans to show the features at LISTEN West in Dallas on 20 and 21 Oct.

Verint said it was named a Leader in the IDC MarketScape for conversational intelligence and analytics software, a week after Genesys announced the same placement. Buyers comparing QA and conversation intelligence tools now have several vendors scoring 100 percent of conversations and explaining each score, so the differences are moving to how those scores feed coaching and routing.

Partners and channels

Parloa, which teamed up with Thoughtworks a week earlier, added TELUS Digital as an implementation partner. Enterprises contract with Parloa for the platform and can hire TELUS Digital for integration, change management, and managed services across North America and Europe. Toku extended its local numbering and PSTN connectivity to Zoom Contact Center across 14 Asia Pacific markets. Cisne, a workforce management startup, joined the Five9 ISV partner program with an integration that recalculates staffing as voice AI agents resolve some calls and hand others to people.

On messaging, Sinch added RCS to its SAP Engagement Cloud integration with SMS fallback. Twilio opened a private beta of Twilio Cohorts for list building and bulk sends, and made its Segment Conversation Memory integration generally available, with retrieval Twilio says runs in under 300 milliseconds. Gorgias made AI Agent on Socials generally available for Instagram DMs, Facebook Messenger, and WhatsApp.

Also this week

Intercom added Fin Voice to its pricing page at $1.99 per voice outcome, according to PricingSaaS, which counts it as Intercom's sixth pricing change in 90 days. Fin, which shipped Fin 4 the week before, lists the Fin AI Agent at $0.99 per outcome with a 90-day money-back guarantee. Intercom also added Dictate to the Inbox composer.

Freshworks signed an agreement to acquire Torii, a SaaS management platform that tracks AI consumption down to the token, to extend Freshservice. It expects the deal to close in its fourth fiscal quarter and did not disclose a price.

For compliance teams, Microsoft posted a roadmap item for LLM-based PCI redaction in Dynamics 365 Contact Center that removes card numbers, PINs, and account numbers from transcripts, recordings, logs, and analytics before they are stored, with a preview planned for December 2026 and general availability for March 2027.

Amazon Connect can now prefill after contact work guides with details pulled from the conversation, such as booking identifiers and claim numbers. Telnyx added AI-based SMS opt-out detection that treats phrases like "don't message me again" as a STOP, at $0.002 per inbound message checked. Zendesk opened an early access program for an agentic shopping assistant inside its commerce agent.

New in the CXSpy catalog

CXSpy added eight company pages this week, bringing the catalog to 321. Crescendo, Netomi, Siena, and Giga are listed under call center AI. Siena is the New York startup whose $17 million Series A led last week's funding roundup. ElevenLabs, one of the design partners named in Sierra's Personal Agent Protocol draft, now has a page under speech AI. Gupshup is listed under messaging, AVOXI under CPaaS, and Luware under CCaaS.

Sources

CX Today on Meta's Muse and the contact center (15 Oct 2026): CX Today
Five9 2026 Consumer CX Report (14 Oct 2026): Five9
CMSWire on the Five9 report (14 Oct 2026): CMSWire
CX Dive on the Five9 report (14 Oct 2026): CX Dive
CX Dive on the Zingtree survey (13 Oct 2026): CX Dive
CXP Asia on China's AI and human customer service standard (16 Oct 2026): CXP Asia
CX Today on Sanas Supervised AI (14 Oct 2026): CX Today
Salesforce on fall 2026 Service Cloud releases (13 Oct 2026): Salesforce
CX Today on Salesforce and Loops (13 Oct 2026): CX Today
CX Network funding roundup (16 Oct 2026): CX Network
Tech Times on Agentforce on by default (13 Oct 2026): Tech Times
Kustomer on Conversation Intelligence and AI QA (14 Oct 2026): Kustomer
CallMiner on AI Assist and RealTime Command Post (14 Oct 2026): CallMiner
Verint on the IDC MarketScape (14 Oct 2026): Verint
TELUS Digital on its Parloa partnership (14 Oct 2026): TELUS Digital
Toku and Zoom Contact Center, via PR Newswire (13 Oct 2026): PR Newswire
Cisne and Five9, via EIN Presswire (12 Oct 2026): EIN Presswire
Sinch on RCS for SAP Engagement Cloud (13 Oct 2026): Sinch
Twilio changelog on Twilio Cohorts (14 Oct 2026): Twilio changelog
Twilio on Segment Conversation Memory (12 Oct 2026): Twilio
Gorgias on AI Agent on Socials (14 Oct 2026): Gorgias
PricingSaaS on Fin Voice pricing (11 Oct 2026): PricingSaaS
Intercom changelog on Dictate (14 Oct 2026): Intercom changelog
Freshworks on acquiring Torii (12 Oct 2026): Freshworks
Microsoft 365 roadmap item on Dynamics 365 Contact Center PCI redaction (16 Oct 2026): Microsoft 365 roadmap
AWS on Amazon Connect after contact work guides (12 Oct 2026): AWS
Telnyx release notes on AI-based SMS opt-out (15 Oct 2026): Telnyx release notes
Zendesk on the agentic shopping assistant early access program (12 Oct 2026): Zendesk