Meta Muse Agent Calls Customer Support to Cut Household Bills

Peter Yang

Meta's existing reach across billions of social media users gives Muse an unmatched distribution moat to become the first mainstream consumer AI agent.

AI-assisted summary of Meta’s Muse AI Agent Saved Me $800+ a Year on My Bills (10 Real Use Cases) — original source. Check the original for full context.

Source published: 2026-09-18

Key ideas

Meta's free AI agent Muse demonstrates direct real-world utility by autonomously calling customer support lines to negotiate bills, securing an Xfinity internet rate drop from $84/month to $60/month. (13:03)
An in-app billing audit reduced an AT&T cellular plan from $200 to $160 per month, generating $480 in annual savings by transitioning autopay to a checking account with secure manual credential handoffs. (0:46)
Deep integration with Meta's consumer ecosystem allows Muse to automate Facebook Marketplace commerce, including listing items, sourcing products, and haggling with sellers over Messenger. (5:48)
Propelled by Meta's massive user distribution moat, Muse surged into the top three on the App Store within one week of its consumer launch. (15:47)

Detailed notes

## Background Published on YouTube by Peter Yang on September 18, 2026, this video demonstrates ten practical use cases for Meta's personal AI agent, Muse. The description highlights how the free agent saved Yang over $800 annually on his cable and phone bills, specifically by autonomously placing calls to customer support to negotiate lower rates on his behalf.

## TL;DR In this walkthrough, Peter Yang showcases Meta's Muse personal AI agent as a powerful tool for streamlining daily life logistics. He demonstrates how Muse delivered over $800 in annual recurring savings by reconfiguring his AT&T billing settings and autonomously calling Xfinity customer support to negotiate a discounted internet rate. Yang also covers how Muse aggregates social messages, facilitates buying and selling on Facebook Marketplace, curates an ad-free morning news feed, enforces personal habit tracking, and discovers local weekend events. Ultimately, he argues that Meta's unparalleled consumer distribution gives Muse a decisive advantage in bringing autonomous AI agents to mainstream users.

## Atmosphere & style The video feels like an engaging, hands-on desktop screenshare led by an enthusiastic and practical early adopter. Yang strikes an authentic, casual tone, effortlessly blending analytical cost-benefit breakdowns with humorous personal confessions about bedtime doomscrolling and designing animated red panda avatars. The narrative arc moves briskly from tangible financial wins to broader lifestyle optimizations, leaving viewers with a tangible sense of excitement about the near future of consumer-facing automation.

## Bulletin - [13:03](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=783s) Meta's free AI agent Muse demonstrates direct real-world utility by autonomously calling customer support lines to negotiate bills, securing an Xfinity internet rate drop from $84/month to $60/month. - [00:46](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=46s) An in-app billing audit reduced an AT&T cellular plan from $200 to $160 per month, generating $480 in annual savings by transitioning autopay to a checking account with secure manual credential handoffs. - [05:48](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=348s) Deep integration with Meta's consumer ecosystem allows Muse to automate Facebook Marketplace commerce, including listing items, sourcing products, and haggling with sellers over Messenger. - [15:47](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=947s) Propelled by Meta's massive user distribution moat, Muse surged into the top three on the App Store within one week of its consumer launch.

## Key takeaways - [00:00](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=0s) Meta's Muse sets a precedent for consumer AI agents by delivering measurable financial returns, saving over $800 annually on household bills at zero software cost. - [00:46](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=46s) Muse identified a $10 per line autopay discount on a 4-line AT&T account, cutting monthly expenses from $200 to $160 and securing $480 in recurring annual savings. - [00:46](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=46s) To protect user privacy and sensitive financial data, Muse yields control within its cloud browser so users manually input bank and routing details rather than storing them in the agent. - [02:45](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=165s) The agent features deep UI customization, allowing users to replace generic chat boxes with responsive, animated avatars that feature dynamic idle and processing animations. - [04:11](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=251s) Muse prioritizes personal consumer communications—integrating Instagram, Threads, Messenger, and iMessage—to filter promotional noise and conduct daily cross-platform triage. - [05:48](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=348s) Facebook Marketplace capabilities extend beyond basic search: Muse can create sale listings and autonomously negotiate pricing with prospective buyers and sellers via Messenger. - [06:44](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=404s) The dedicated Feed tab offers an ad-free, high-signal alternative to algorithmic social media feeds, aggregating focused topic digests and personal project updates. - [08:01](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=481s) The Goals tab implements structured habit tracking, combining diagnostic questioning, timed nightly reminders, and morning check-ins to build persistent streak data. - [10:05](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=605s) The Ideas tab bypasses the empty chat prompt problem by proactively analyzing user context to suggest actionable tasks, such as reporting municipal potholes or comparing consumer products. - [12:11](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=731s) As a local lifestyle scout, Muse automates itinerary planning, date night reservations, and recurring event discovery across local community calendars. - [13:03](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=783s) Muse's standout capability is placing outbound voice calls to live customer support agents using a transparent AI persona that discloses call transcription while advocating on the user's behalf. - [13:03](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=783s) During an autonomous Xfinity call, Muse bypassed IVR phone trees and hold queues to discover unadvertised promotional rates and autopay discounts without manual user effort. - [15:47](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=947s) Meta holds a unique structural advantage in the AI agent race, leveraging its multi-billion-user ecosystem across Facebook, Instagram, and WhatsApp to scale Muse rapidly. - [15:47](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=947s) Effective personal AI workflows are stratifying across specialized tools: ChatGPT for cognitive knowledge work, Grokbot for background cloud execution, and Muse for personal administration and phone automation.

## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=0s) Introduction: Why Muse Is a Game-Changing Personal Agent

Meta's Muse stands out as the single best personal AI agent on the market to date. The tool delivers "insane value" for software that is completely free to use, demonstrating concrete real-world utility by saving over $800 annually across cable and phone bills. Rather than merely offering suggestions, Muse's standout capability is taking direct action—including autonomously calling customer support lines to negotiate contracts and rates on the user's behalf.

Across ten primary use cases, Muse handles diverse everyday tasks that range from triaging communications across platforms like Messenger and iMessage to curating customized morning news feeds, tracking personal habits, booking dinner reservations and event tickets, and aggressively cutting recurring household expenses.

---

[00:46](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=46s) Saving $480 a Year on AT&T and Expanding into Insurance

The money-saving workflow begins inside the Muse desktop application with a dedicated finances chat thread aimed at reducing an AT&T cellular bill. To initiate the audit, Muse directs the user to log into the carrier's portal. The user inputs their credentials into Muse's built-in secure store, allowing the agent to navigate the AT&T website via a cloud browser. Muse inspects the account structure and spots an immediate billing inefficiency: switching the payment method from an existing credit card to direct bank account autopay yields a discount of $10 per line each month. With four active lines on the plan, this single tweak represents an ongoing savings of $40 per month.

```mermaid flowchart TD accTitle: AT&T Bill Optimization Flow accDescr: Muse identifies payment method savings and guides the user through secure bank linking. subgraph Audit ["Account Audit"] A["Store Login in Secure Vault"] --> B["Cloud Browser Logs Into AT&T"] B --> C["Identify Autopay Discount: $10/line/month across 4 lines"] end subgraph Execution ["Execution & Handoff"] C --> D["User Manually Enters Bank & Routing Info (Privacy/Security Protection)"] D --> E["Autopay Switched to Checking Account"] end subgraph Outcome ["Financial Impact"] E --> F["Monthly Bill Drops: $200 to $160"] F --> G["Total Annual Savings: $480/year"] end ```

After being instructed to apply the discounts without disrupting existing phone lines, the agent guides the user through the enrollment sequence. When it reaches the sensitive step requiring bank account and routing numbers, Muse intentionally yields control to the user inside the cloud browser view rather than storing or typing financial credentials itself—a deliberate security measure to prevent privacy risks. Once the user submits their Bank of America checking details, Muse finalizes the setup, confirming that the monthly statement immediately drops from $200 down to roughly $160. Within a five-to-ten-minute chat interaction, Muse locks in $480 in annual savings. Building on this momentum, the user taps the finances thread to audit their Tesla auto insurance, prompting Muse to gather vehicle details and garaging locations to begin cross-shopping competitive quotes against Geico, Progressive, and State Farm.

- **Carrier & Lines**: AT&T account with 4 lines. - **Immediate Arbitrage**: Shifting autopay from a credit card to a bank account cuts $10 per line ($40/month total). - **Net Bill Impact**: Lowers monthly payment from $200 to $160, producing $480 in recurring yearly savings. - **Security Guardrail**: Cloud browser lets the user manually type bank and routing data for complete privacy. - **Insurance Pipeline**: Automated quote discovery initiated across Geico, Progressive, and State Farm.

---

[02:45](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=165s) App Orientation and Customizing an Animated Red Panda Avatar

Muse defaults to an out-of-the-box avatar, but the desktop experience is built to be deeply customized. As a self-professed red panda enthusiast, the user begins setup by asking Muse to completely restyle its visual persona into an animated red panda. Through an interactive iterative loop in chat, Muse generates four initial concept renderings. Prompted to "make it cuter," it produces a second batch of four refined options, one of which the user selects.

Once chosen, Muse instantly applies the new look and generates reactive idle animations right away, allowing the red panda avatar to naturally look around the interface. To expand on this polish, the user prompts it for additional dynamic behaviors. By default, Muse provides animations for directly facing the user and typing while processing answers, but Muse is also able to generate standalone preview clips such as the red panda waving hello or munching on bamboo. While Muse admits it cannot currently inject custom custom-generated video clips directly into the static interactive profile slot, the seamless process of transforming an AI into an expressive, moving companion highlights the software's overall refinement. Having a responsive animated character feels markedly warmer and more personal than staring at an empty chat text box.

---

[04:11](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=251s) Connecting Everyday Apps and Unified Message Monitoring

Muse's broader software ecosystem is tailored explicitly around personal life rather than generic enterprise productivity. While common tools support email and calendar synchronization, Muse’s key competitive edge is native integration across the Meta product suite: Instagram, Threads, Facebook, and Messenger.

To manage communication overload, the user configures a dedicated "messages thread," directing Muse to review incoming communications across chatting services while excluding standard email clutter. After linking services including iMessage, Instagram, and Threads, Muse conducts an inbox scan. Although the initial DM triage reveals that inbound messages on Instagram and Threads consist almost entirely of junk and promotional spam, the utility of the feature is clear. Juggling disparate channels like WhatsApp, Messenger, and iMessage is universally exhausting. Automating a daily scheduled check-in allows Muse to silently surface truly critical notifications without requiring the user to manually open and clean out every single app. When tested against workplace tools like Linear, Muse fails to connect, but this limitation reflects an intentional design philosophy: ordinary consumers benefit far more from one-click consumer app integrations than from wrestling with developer API keys and enterprise webhooks.

- **Native Meta Ecosystem**: Direct access to Instagram DMs, Threads, Facebook profiles, and Messenger. - **Message Aggregation**: Unifies cross-app triage across iMessage, WhatsApp, and social chat apps into a single monitoring routine. - **Daily Scheduled Briefing**: Flags urgent follow-ups while filtering out social media promotional spam. - **Consumer Focus**: Prioritizes seamless plug-and-play access over developer tools or API configurations.

---

[05:48](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=348s) Automating Facebook Marketplace Buying, Listing, and Negotiating

Thanks to deep Facebook integration, Muse functions as an autonomous commerce assistant on Facebook Marketplace, handling both sides of the transaction: buying and selling.

When tasked with finding single golf clubs locally, Muse parses Marketplace listings and delivers a curated summary of three targeted options, complete with links to inspect each item's full posting. On the selling side, offloading an old iPhone 13 Pro is equally frictionless: the user provides the item photos, target listing price, and descriptive criteria, and Muse publishes the listing directly to Marketplace. Crucially, because Muse links directly into Facebook Messenger, it can negotiate deals autonomously. The agent can message sellers to confirm whether an item is still available, align on specific terms, and haggle on price to close purchases at the best possible rate.

---

[06:44](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=404s) Building an Ad-Free Personalized Morning News Feed

In contrast to the standard Facebook news feed—which has become increasingly cluttered with low-effort "AI slop junk and ads"—Muse features a dedicated **Feed** tab on the left sidebar that functions as a high-signal, bespoke intelligence dashboard. Setup requires nothing more than asking Muse to monitor core topics of interest, such as major breakthroughs from leading AI research laboratories.

Every morning, the agent compiles an uncluttered digest that can be skimmed over coffee, with clickable entries that launch straight into the browser for deeper reading. The feed is fully modular; users can blend external industry news with personal project alerts, such as subscriber trends or draft updates for a newsletter. Most importantly, the feed currently operates completely ad-free, creating an uninterrupted morning reading ritual focused entirely on what the user cares about.

---

[08:01](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=481s) Habit Tracking in the Goals Tab: The Phone-Free Bedroom Challenge

Muse pairs information consumption with personal accountability through its **Goals** tab. The agent automatically maintains a living to-do list derived from ongoing conversational context—tracking items like booking an upcoming flight to Japan, finalizing a YouTube video script about Muse, or negotiating cable contracts. Underneath these task items lies a proactive behavioral goal-tracking system designed to modify daily habits.

The user uses this system to address a persistent bad habit: bringing their smartphone into the bedroom at night. Despite maintaining disciplined routines around daily exercise and clean eating, late-night phone browsing often causes the user to wake up around 5:00 a.m. only to fall into an addictive doomscrolling loop instead of going back to sleep. When prompted to help break this cycle, Muse asks diagnostic questions about what draws the phone into the bedroom and what alternate activity could replace it. The user shares that their bedtime target is between 10:00 and 10:30 p.m. ("kind of like an old man") and that their aspiration is to replace late-night phone scrolling with reading books.

```mermaid flowchart LR accTitle: Habit Tracking Architecture accDescr: Morning and evening interactive check-ins feed a persistent streak-tracking UI artifact. subgraph Routine ["Daily Cadence"] A["10:00 PM Bedtime Reminder: Leave phone outside bedroom"] --> B["Morning Check-In: Audit adherence"] end subgraph Interface ["Sidebar Artifact"] B --> C["Log Success / Failure"] C --> D["Update Habit Charts & Streaks"] end ```

Muse converts this challenge into a dedicated interactive UI artifact in the left-hand workspace: - **Bedtime Intervention**: Muse triggers a targeted alert around 10:00 p.m. to remind the user to leave their phone outside the bedroom. - **Morning Audit**: A daily morning follow-up asks whether the user successfully honored the rule. - **Streak Visualization**: Responses populate visual habit charts and consecutive streak tallies within the interface.

Though the initial results show a rocky start—with the user admitting they still succumbed to bringing the phone to bed two nights in a row—the automated daily reminders, persistent visual artifact, and structured accountability offer an effective framework for targeting personal health and lifestyle goals.

[10:05](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=605s) Proactive Ideas and Side-by-Side Product Comparisons

When starting a new habit or routine, Muse does not leave the user staring at an empty prompt; if you indicate that you want to start a health goal, it actively guides you through selecting and structuring the right targets. This leads into what Peter considers one of the software’s standout interfaces: the proactive suggestions found in the "Ideas" tab. In stark contrast to standard ChatGPT-style interfaces—where users are confronted with an empty chat box and often have no idea what to type—Muse draws on prior conversations and personal context to surface proactive recommendations. It suggests staging video shoot packets, locking in lower utility bills, triaging email inboxes, or analyzing meal photos to calculate calories. It even surfaces creative civic actions, such as drafting a message to a local government agency to fix a nearby pothole or a broken street light.

```mermaid flowchart TD accTitle: Proactive Ideas to Automated Tasks accDescr: Flowchart illustrating how Muse turns user context into proactive ideas and automated tasks. subgraph ContextEngine["User Context Engine"] A["Chat History & Personal Context"] --> B["Ideas Tab (Proactive Engine)"] end subgraph SuggestedTasks["Suggested Automations"] B --> C["Personal Tasks • Meal calorie tracking • Inbox triaging"] B --> D["Civic Actions • Report street lights • Report potholes"] B --> E["Research & Commerce • Product comparisons • Ticket booking"] end ```

To test its shopping capabilities live, Peter asks Muse to compare electronic razors side by side: *"give me a side byside comparison of the best options and let's see what it comes up with."* After roughly five minutes of browsing product listings on Best Buy, Muse delivers a ranked list of five options with detailed rationale, giving top recommendation honors to a Panasonic model. While Peter notes a minor limitation—wishing the agent had consulted dedicated editorial review publications like *Wirecutter* in addition to retail storefronts—he emphasizes that the agent is capable of handling the entire transaction on the spot via its integrated browser tool.

- **Sample proactive prompt categories in Ideas:** - Production workflows: Staging a video shoot packet. - Financial optimizations: Locking in discounted utility and service rates. - Daily productivity: Inbox triaging and message prioritization. - Health tracking: Snapping photos of meals to log macronutrients and calories. - Civic engagement: Contacting municipal public works departments for road maintenance and street repairs.

---

[12:11](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=731s) Event Discovery, Date Night Planning, and "Weekend Scout"

Muse functions exceptionally well as a localized research assistant. Peter demonstrates how it aggregates upcoming concerts and ticketing events across the San Francisco Bay Area. Narrowing down the query, he asks it to identify highly rated performances for September 30th—his wedding anniversary—along with nearby dinner recommendations.

The agent highlights a performance at the Magic Theatre alongside suitable restaurants in close proximity. From that screen, purchasing tickets requires only a single instruction: selecting two general admission tickets prompts Muse’s autonomous browser to initiate the booking flow. Beyond planning specific occasions, users can set up automated recurring routines like a "weekend scout," which regularly monitors community calendars. In Peter's case, this feature uncovered an upcoming Comedy Day event in Golden Gate Park that he otherwise would have missed.

---

[13:03](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=783s) Autonomous Customer Support Calls: Negotiating an Xfinity Cable Bill

The definitive "killer feature" of Muse—and one Peter predicts competitors will inevitably clone—is its capability to place outbound voice calls to negotiate bills with live customer service representatives. Peter tasks Muse with lowering his Xfinity Comcast internet bill. Before dialing, the software prompts the user for setup parameters, asking whether to use a male or female voice persona and requesting the user's full legal name and phone number for account verification.

While Peter is eating breakfast, Muse dials Xfinity autonomously. It successfully secures an offer dropping the bill down to a promotional rate of $60 per month, plus an additional $10 discount upon enrolling in autopay. Reviewing the call log, Muse handled the conversation transparently rather than spoofing identity:

> *"Hey, this is Haley calling on Peter's behalf. I'll transcribe his call for notes. I'm calling to see if Peter can keep the same internet plan but lower the $84 monthly bill. Can you help me with that?"*

```mermaid flowchart LR accTitle: Muse Autonomous Bill Negotiation Call Flow accDescr: Sequence diagram style flowchart showing Muse negotiating an Xfinity bill. subgraph Setup["Configuration"] A["Set Voice & Identity (Voice: Haley)"] --> B["Account Details (Peter Yang)"] end subgraph LiveCall["Live Interaction"] B --> C["Dials Xfinity Call Center"] C --> D["Identifies as Personal Agent (Discloses transcription)"] D --> E["Negotiates $84/mo rate down"] end subgraph Outcome["Resolution"] E --> F["Promotional Rate: $60/mo"] F --> G["Autopay Discount: -$10/mo"] end ```

The support agent asked to verify the account holder's name and ultimately indicated that the primary account owner would need to finalize the discount personally. Nevertheless, Peter highlights the massive value delivered: skipping the interactive voice response (IVR) tree, bypassing hold music, and discovering available promotional discounts with zero manual effort. Peter playfully adds that he originally gained early access to this feature by instructing Muse to file feedback directly to the Meta product development team on his behalf, which was approved that very morning.

---

[15:47](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=947s) Meta’s Distribution Playbook and Peter's Everyday AI Stack

Reflecting on all ten use cases, Peter argues that Meta has taken the lead in the personal AI agent race. Beyond product polish, Meta holds a structural moat through distribution. Just as the company leveraged its existing platforms to outmaneuver Snapchat with Instagram Stories and challenge X with Threads, it can effortlessly market and deploy Muse to billions of active users across Facebook, Instagram, and WhatsApp. Within just one week of launch, Muse surged into the top three apps on the App Store.

Peter explains how he has consolidated his personal workflow to avoid constant migration between AI tools:

Peter concludes that Muse represents the first consumer AI agent with a realistic path toward reaching a billion mainstream users, a development that makes him decidedly bullish on Meta stock. He encourages viewers to take advantage of the free tool to lower their monthly expenses, curate feeds, and automate life administration.

## Notable quotes > [00:00](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=0s) insane value — Peter Yang

> [02:45](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=165s) make it cuter — Peter Yang

> [06:44](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=404s) AI slop junk and ads — Peter Yang

> [08:01](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=481s) kind of like an old man — Peter Yang

> [10:05](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=605s) give me a side byside comparison of the best options and let's see what it comes up with. — Peter Yang

> [13:03](https://www.youtube.com/watch?v=eU1ICyI9bCs&t=783s) Hey, this is Haley calling on Peter's behalf. I'll transcribe his call for notes. I'm calling to see if Peter can keep the same internet plan but lower the $84 monthly bill. Can you help me with that? — Muse

## Mentioned AT&T, Bank of America, Best Buy, ChatGPT, Comedy Day, Facebook, Facebook Marketplace, Geico, Golden Gate Park, Grokbot, Instagram, Linear, Magic Theatre, Messenger, Meta, Muse, Panasonic, Peter Yang, Progressive, San Francisco Bay Area, State Farm, Tesla, Threads, WhatsApp, Wirecutter, X, Xfinity Comcast, iMessage

## Video description Copy my best AI workflows to save time and automate busywork: https://www.behindthecraft.com 🤝 Subscribe to my practical AI newsletter read by 150K+ professionals: https://creatoreconomy.so

Muse from Meta is the best personal agent I’ve used to date. It actually saved me over $800 a year in my cable and phone bills, which is just insane value for an AI tool that you can use for free. And the craziest part is that it did this by calling customer support to negotiate my bills for me.

In this video, I walk through 10 of my favorite Muse use cases, including checking for important messages, building a personalized morning news feed, tracking and fixing bad habits, booking restaurants and shows, and of course, negotiating all my bills.

What I cover: (00:00) Why Muse is my favorite personal agent (01:00) Lower my phone bill with Muse (03:00) Personalize Muse and connect my apps (06:00) Shop Facebook Marketplace with Muse (07:00) Build a personalized morning Feed (08:00) Track a phone-free bedroom goal (10:30) Use Ideas to shop and plan a night out (13:00) Have Muse call to negotiate my cable bill

Try Muse: [https://muse.ai](https://muse.ai)

Where to find me: X: [https://x.com/petergyang](https://x.com/petergyang) LinkedIn: [https://www.linkedin.com/in/petergyang/](https://www.linkedin.com/in/petergyang/)

Subscribe to this channel - more tutorials coming soon!

[Source](https://www.youtube.com/watch?v=eU1ICyI9bCs)

中文简报

Meta Muse智能体代打客服电话降低家庭账单 Meta触达数十亿社交媒体用户的现有优势，为Muse构筑了无可比拟的分发护城河，使其有望成为首个主流消费级AI智能体。在本次演示中，Peter Yang展示了Meta的个人AI智能体Muse作为简化日常生活事务的强大工具。他演示了Muse如何通过重新配置其AT&T账单设置，并自主致电Xfinity客服协商宽带折扣费率，为其实现每年超过800美元的经常性节省。Yang还介绍了Muse如何聚合社交消息、促进Facebook Marketplace上的买卖交易、定制无广告早间新闻流、督促个人习惯追踪以及发掘当地周末活动。最后他指出，Meta无可比拟的消费者分发能力使Muse在将自主AI智能体推向主流用户方面具备决定性优势。 Meta的免费AI智能体Muse展现出直接的实际应用价值，能够自主致电客服热线协商账单，成功将Xfinity宽带月费从84美元降至60美元。 (13:03) 通过应用内账单审查，在安全的手动凭证交接下将自动扣款更改为支票账户，从而将AT&T手机套餐月费从200美元降至160美元，每年节省480美元。 (0:46) 与Meta消费级生态系统的深度整合，使Muse能够自动化处理Facebook Marketplace的交易事务，包括上架商品、搜寻产品以及通过Messenger与卖家议价。 (5:48) 依托Meta庞大的用户分发护城河，Muse在面向消费者推出后的一周内便跃升至App Store排行榜前三名。 (15:47)

Reading summary

Peter Yang argues that Meta's free personal AI agent, Muse, distinguishes itself from competitors by taking autonomous real-world actions, such as negotiating bills and managing personal logistics across Meta's consumer ecosystem. Automated Bill Audits and Phone Negotiations (0:46) Muse delivers tangible financial savings by auditing recurring service accounts and directly negotiating rates. In an initial test on an AT&T account with four lines, Muse navigated the carrier's portal via a cloud browser to identify that switching from credit card autopay to direct bank account autopay yielded a $10 discount per line. To protect user privacy, the agent handed control over to the user to enter bank routing and account credentials manually, reducing the monthly bill from $200 to $160 and securing $480 in annual savings. Yang identifies Muse's outbound phone call capability as its primary breakthrough feature. When tasked with lowering an Xfinity internet bill, Muse autonomously dialed customer service, transparently identified itself as an AI agent transcribing on the user's behalf, and bypassed interactive voice menus and hold times while Yang ate breakfast. The agent successfully negotiated the $84 monthly bill down to a $60 promotional rate plus an additional $10 autopay discount, though the carrier still required the account holder to verify details to finalize the agreement. Everyday Commerce and Consumer App Integrations (4:11) Unlike enterprise-oriented AI platforms, Muse prioritizes native integrations with consumer applications, particularly Meta’s own properties such as Instagram, Messenger, Threads, and Facebook Marketplace. The agent acts as an autonomous commerce assistant on Marketplace by sourcing local listings, publishing sales listings—such as drafting an offer for an iPhone 13 Pro—and messaging sellers directly via Messenger to confirm item availability and negotiate prices. For general communications, Muse unifies inbound message monitoring across platforms like iMessage, Instagram, and Threads into a scheduled daily briefing that filters promotional spam and flags urgent conversations. Yang notes that while Muse failed to connect with workplace developer tools like Linear, its plug-and-play setup targets ordinary consumer workflows without requiring technical API configurations. Proactive Guidance and Daily Behavioral Tracking (8:01) Rather than relying on a blank prompt box, Muse leverages conversation history to surface proactive suggestions in an dedicated "Ideas" tab, recommending actions such as comparing consumer products, triaging inboxes, calculating meal calories, or reporting municipal road issues. When asked to evaluate electric shavers, the agent browsed retail listings to deliver a ranked comparison of five models within five minutes, though Yang observed that it relied on storefront listings rather than dedicated editorial review publications. In addition to generating ad-free, topic-specific morning news digests, Muse incorporates an interactive "Goals" interface to assist with personal accountability. To address late-night screen time, the agent instituted an automated routine: triggering a 10:00 p.m. reminder to leave the smartphone outside the bedroom, followed by a daily morning check-in that tracks consecutive streaks visually within the application workspace. Meta's Distribution Advantage and Market Position (15:47) Yang contends that Meta possesses a structural distribution moat that positions Muse to outcompete rival personal assistants. Much like Meta's previous deployment of Stories and Threads against established competitors, Muse can be distributed directly to billions of existing users across Facebook, Instagram, and WhatsApp. Yang notes that the application reached the top three rankings on the App Store within one week of release. Within his personal workflow, Yang divides responsibilities across three distinct systems: ChatGPT handles professional cognitive work, Grokbot runs background cloud computations, and Muse manages personal life logistics, local event discovery, and service calls. Based on Muse's rapid consumer onboarding and free availability, Yang expresses a strongly bullish outlook on Meta's long-term commercial trajectory. Yang concludes that Muse represents the first consumer AI agent with a viable trajectory toward hundreds of millions of mainstream users, though full autonomy remains constrained by service providers that still require human verification to execute contract changes.

Source quotes

[0:00] Peter Yang: insane value [2:45] Peter Yang: make it cuter [6:44] Peter Yang: AI slop junk and ads [8:01] Peter Yang: kind of like an old man [10:05] Peter Yang: give me a side byside comparison of the best options and let's see what it comes up with. [13:03] Muse: Hey, this is Haley calling on Peter's behalf. I'll transcribe his call for notes. I'm calling to see if Peter can keep the same internet plan but lower the $84 monthly bill. Can you help me with that?

Original transcript

[0:00] Muse from Meta is the best personal [0:03] agent that I've ever used to date. It [0:05] actually saved me over $800 a year in my [0:08] cable and phone bills, which is just [0:10] insane value for an AI tool that you can [0:12] use for free. And the craziest part is [0:15] that it did this by calling customer [0:17] support to negotiate my bills for me. [0:20] So, in this video, I'm going to walk you [0:21] through 10 of my favorite Muse use [0:23] cases, including checking for important [0:25] messages in Messenger and iMes, building [0:29] a personalized morning news feed, [0:31] tracking and fixing my bad habits, [0:34] booking restaurants and shows, and of [0:36] course, negotiating all my bills. All [0:39] right, so let me start by jumping right [0:41] in and showing you how to use Muse to [0:44] save money right now. Okay, so this is [0:46] the Muse desktop app. And here's my chat [0:49] with Muse to save money on my AT&T phone [0:53] bill. [0:54] It started by asking me to log to AT&T [0:57] on the website. And I basically gave it [0:59] my username and password in the secure [1:01] store and it uses browser to login for [1:05] me. And then I looked up my account and [1:07] what it found was that if I switched my [1:10] bill from my credit card to my bank [1:13] account, it can actually save me around [1:15] $10 per line per month. So in total, [1:18] that's $40 off per month. So I told it, [1:22] hey, give me the discounts, but don't [1:24] mess up my lines. And I started to [1:26] enroll in this stuff. And then [1:28] eventually it asked me to enter in my [1:32] bank account number and also routing [1:34] number in its cloud browser at the AT&T [1:37] website. You know, it didn't want to do [1:38] it for me because of privacy and other [1:40] issues. So I did that and then it said [1:44] done. The discount is live. Here's where [1:47] things stand. You got $40 a month off. [1:50] It's now pulling from your Bank of [1:52] America checking balance. And now we've [1:55] reduced the price of your bill from $200 [1:58] to around $160. [2:00] Right? And down here I asked it, hey, [2:02] how much money did this save me per [2:04] year? And it told me that I saved around [2:06] $480 per year, which is just incredible [2:09] from a 5 to10 minute conversation. [2:12] So you can see here that I basically set [2:14] up a finances thread with Muse and I'm [2:17] asking it you know for more ways to save [2:19] money and I've started working with it [2:21] to potentially save money on my Tesla [2:23] insurance too. So basically it asks me [2:25] for information about my car where it's [2:27] based and so on and it can look up a [2:30] bunch of other insurance providers Geico [2:32] Progressive and State Farm to try to [2:34] negotiate my insurance bills. And later [2:37] in this video, I'll also show you how I [2:38] called and negotiated my Comcast cable [2:40] bill with a real customer support [2:42] person. But let's take a step back real [2:45] quick and get oriented around the Muse [2:48] app. So, by default, Muse comes with [2:51] this pretty cute SAS avatar, but [2:54] personally, I'm a huge Red Panda fan. [2:57] So, the first thing I did when I [2:58] installed Muse is ask it to switch to a [3:01] red panda avatar. All right. So [3:03] basically I asked it, hey, can you [3:05] become a red panda with animations? And [3:07] it generated four images of red pandas. [3:10] And then uh I told it to make it cuter. [3:12] And it came with four more. And then I [3:14] picked this one. And uh what's [3:17] incredible is that it updated look and [3:19] it added animations by default, right? [3:22] So it actually animated avatar to look [3:24] around. So this is super cute, but I [3:26] wanted to push it a little bit further. [3:27] I told it to add more animations. You [3:30] know, by default, it has animation [3:31] facing me and also animation showing it [3:33] typing. And then I got it to generate a [3:36] bunch more red panda clips. So, here's a [3:38] clip of red panda waving hi. Uh here's a [3:41] clip of red panda eating some bamboo and [3:43] so on. Okay. So, then I asked it to add [3:45] the animations to be edible on my [3:47] profile. And unfortunately, I said that [3:49] it's not possible. Uh there's no path [3:51] for me to add more animations. So, we [3:53] kind of got stuck there. But the fact I [3:56] can customize the avatar so easily and [3:59] have it be animated in a few simple chat [4:01] requests is just a sign of how polished [4:03] the Muse experience is. You know, my red [4:06] panda just feels so much more personal [4:08] than a bunch of random chat messages. [4:11] All right, next I want to show you what [4:12] apps Muse can connect to. So, basically, [4:15] I asked it, you know, what apps can you [4:17] connect to? As you can see from Mus's [4:19] reply, there are quite a lot of apps and [4:22] beyond the standard email and calendar [4:24] stuff, right? I think one of Mus' [4:26] biggest differentiators is that it can [4:28] connect natively to all of Meta's apps [4:31] like Instagram threads, Facebook, and [4:34] Messenger. And real quick, I actually [4:37] set up a messages thread and basically [4:41] asked it to monitor my messaging apps, [4:43] not including email, and let me know if [4:45] there's something that I need to follow [4:46] up on. And then it checked iMes and it [4:49] asked me to connect to all these other [4:50] things. And it turns out that all I'm [4:52] getting through Instagram and threads [4:54] DMs is a bunch of spam. So nothing super [4:57] important, but you can see how useful [4:59] this is. You know, we all use multiple [5:00] messaging apps. Personally, I use [5:02] Messenger, WhatsApp, and iMes. And it's [5:05] just hard to keep track of everything. [5:06] So be able to set up, for example, a [5:08] scheduled task for my muse to check [5:11] unread messages every day and let me [5:13] know what's important is just incredibly [5:15] valuable. Now going back to this list of [5:17] apps that Muse can connect to. What I [5:19] found is that Muse can actually easy [5:22] connect to apps beyond this list. So for [5:25] example, I tried to get it to connect to [5:26] Linear and it couldn't figure out how to [5:29] do it. But I actually really like that [5:31] Muse is focused on personal use cases [5:33] and personal apps instead of being kind [5:35] of a jack of all trades. You know, for [5:37] normal people, they're not going to want [5:38] to deal with API keys and other things [5:41] like that. So just being able to connect [5:43] to these everyday apps in one shot is [5:45] very useful. All right, so since Muse [5:49] can connect to Facebook, it's also [5:51] really great for buying and listing [5:53] things on Facebook Marketplace. So, for [5:56] example, here I asked it to look up [5:58] single golf clubs near me to buy on [6:01] marketplace and it gave me a list of [6:03] these, you know, three different options [6:06] and I can click on each one to check out [6:08] the full listing. [6:10] And if I want to, I can also list my own [6:13] stuff. So, if I want to list my old [6:14] iPhone 13 Pro on Facebook Marketplace, [6:17] it just asks me for the photos, the [6:19] price, and everything else, and it will [6:21] list it for me. And because Muse is also [6:23] connected to Facebook Messenger, um it [6:26] can also message sellers and negotiate [6:28] on price. So for example, here I'm [6:30] asking it to send a message to the [6:32] person who listed this golf club to ask [6:34] about availability and you know after [6:37] confirming it can send a message and [6:39] negotiate with the seller to buy the [6:41] golf club at the right price. All right. [6:44] So now let's explore some of the tabs to [6:47] the left of the Muse app here. Starting [6:49] with the feed tab. You know, the funny [6:52] thing is these days Facebook's feed [6:55] itself is just full of like AI slop junk [6:58] and ads. And Muse's feed here is [7:00] actually much more valuable to me. So [7:02] all I did to generate this feed was I [7:05] asked Muse to send me important AI news [7:08] and updates on topics that I care about. [7:11] And every morning you can see here that [7:12] I can basically skim this feed to keep [7:15] track of all the latest AI stuff from [7:18] the major labs and news. And I can also [7:21] click on each thing to open the browser [7:24] to dig into more. Another thing you can [7:26] do with feed is you can customize it [7:28] here. And you can also ask it to add [7:31] personal information. So things tied to [7:33] my life and projects, you know, like my [7:35] newsletter. So basically, you can make a [7:38] personal feed with whatever you want. [7:41] And at least right now, this feed [7:42] doesn't have any ads at all, right? So [7:45] I'm I'm sure someday, you know, if Muse [7:47] becomes super popular, Meta is going to [7:49] add ads to this feed. But for now, I [7:52] think it's just incredibly valuable to [7:53] have a personalized feed that you can [7:55] look at every morning while drinking [7:56] your coffee on topics and stuff that you [7:58] actually care about. All right, next, [8:01] let's take a look at Muse's goals tab. [8:05] So, I really love how Muse also makes a [8:07] personalized to-do list here based on [8:10] all the conversations that I've had with [8:11] it. So, you can see here how Muse has [8:14] items for uh negotiating my cable bill, [8:17] for booking my Japan flight, for making [8:19] a YouTube video about Muse and so on. [8:22] And down here, you can see that there's [8:23] a goal section. So, let me show you the [8:26] chat that I did to set up this goal. You [8:29] know, overall, I think I have pretty [8:30] good habits. I work out every day and I [8:32] try to eat healthy. But one of my worst [8:34] habits is I like to bring my phone to [8:37] the bedroom and look on my phone before [8:39] bed. And often I will wake up at 5:00 [8:41] a.m., you know, decide to doom scroll my [8:43] phone and not be able to sleep again. So [8:45] I asked Muse, hey, I'm trying to break [8:46] this habit. I want to stop bringing my [8:48] phone to my bedroom, right? And it told [8:51] me, hey, what usually pulls your phone [8:53] to the bedroom? What do you actually do [8:54] with it? I say, I like to scroll. And [8:56] instead of bringing my phone to my [8:58] bedroom, I aspire to read books. And I [9:02] usually go to sleep around 10 to 10:30 [9:04] p.m. You know, I'm kind of like an old [9:06] man, so I go to sleep pretty early. All [9:08] right. So then decide to make a plan. [9:11] And there we go. And what's really cool [9:13] is that I set up this artifact that's [9:14] also visible on the left. Let's take a [9:17] look at this artifact right now. [9:19] So, what this artifact does is every [9:22] day, you know, around bedtime, it's [9:25] going to ask me and remind me, hey, [9:26] don't bring your phone to your bathroom. [9:28] And then every morning, it's going to [9:29] check in with me on whether I actually [9:31] did it or not. And then based on that, [9:33] it's going to show me this chart along [9:35] with streaks. So, um, [9:38] basically the record so far has not been [9:40] very good. Uh, pretty much I' I've still [9:43] brought my phone to my bedroom, uh, two [9:45] nights in a row. But, you know, having [9:48] this artifact and having me used to send [9:50] me reminders every day will hopefully [9:52] motivate me to fix this bad habit. And [9:55] as I mentioned, all of your goals also [9:57] show up here. And down here, you can set [10:00] up goals for, you know, your health, for [10:03] your relationships, and so on. And if [10:05] you say, I want to start a health goal, [10:07] it's going to walk you through what kind [10:08] of health goals you want to start. So, [10:11] yeah, using Muse to track your habits is [10:13] definitely a very interesting use case. [10:15] All right. All right. So, another tab [10:16] that I really like is Muse's [10:19] personalized suggestions for things that [10:22] it can do for me. You know, with chat [10:24] GPT and some of these other providers, [10:26] it's just a chat box and most people [10:28] have no idea what to put into the chat [10:30] box. But Muse actually based on my [10:33] conversations and based on what it knows [10:34] about me, proactively suggest stuff. So, [10:37] some ideas it has is, hey, it can stage [10:41] my Muse video shoot packet. It can try [10:43] to lock in a lower bill for me. It can [10:46] triage my inbox, right? Typical personal [10:49] agent use case. It can snap a pick of my [10:52] meal and track every calorie and other [10:55] things. So, let's actually try this [10:56] right now. Let's actually tell Muse to [10:59] help me figure out, you know, I want to [11:00] buy electronic razor. So, let's figure [11:02] out how to do a side-by-side comparison [11:05] of a razor. [11:07] All right. So, let's say I want to buy [11:10] an electronic razor. [11:12] give me a side byside comparison of the [11:16] best options and let's see what it comes [11:18] up with. Okay, so that took about five [11:20] minutes and what it did was it basically [11:22] browsed the uh product pages on Best Buy [11:25] for electronic razors and it gave me [11:27] this list. I kind of wish it actually [11:29] went to some review sites like wire [11:31] cutter or other things to look things [11:33] up, but this is a good start, right? And [11:36] among the five things, it looks like it [11:38] recommends Panasonic for the reasons [11:40] that it sites here. And I'm sure if I [11:43] wanted to buy it for me, it can buy on [11:45] the spot here. But the point is, you [11:47] know, I really recommend if you download [11:49] and install Muse to check out the ideas [11:51] tab. There's just so many really great [11:53] ideas for what it can do for you, [11:55] including some of the things that are [11:56] more out there. Like for example, if you [11:58] have a pothole or a busted street light, [12:00] you can actually maybe message the [12:02] government division to try to get it [12:04] fixed. So there's a lot of really great [12:06] ideas here that's personalized based on [12:08] your usage of Muse. All right, I want to [12:11] show you just a few more use cases. So [12:14] you can also ask Muse to look up [12:16] upcoming concerts and ticketing events [12:18] for me in the Bay Area and it came with [12:20] this list, right? Then I asked some [12:23] questions about, you know, look up stuff [12:24] that's highly reviewed. the Magic [12:26] Theater for September 30th, which is my [12:28] anniversary with my wife, and also look [12:30] for a nice dinner spot nearby, and it [12:34] came up with a list of these things. And [12:35] of course, if I want to actually book [12:37] the show, I can basically just hit, hey, [12:38] I want to get two general admission [12:40] tickets, and I'm sure it's going to use [12:42] this web browser to book things for me. [12:45] So, definitely use Muse to look up local [12:47] events. And you can also set up a [12:50] weekend scout to look up interesting [12:52] things to do in the area, right? For [12:55] example, Muse told me about this list [12:57] here and I found out there's actually a [12:59] comedy day in Golden Gate Park that I [13:01] should check out. Okay, so last but not [13:03] least, I want to show you what I think [13:04] is one of the real killer features of [13:06] Muse that I think the other AI systems [13:09] are going to copy, which is actually [13:11] making phone calls for you. So I told it [13:14] to basically help me negotiate my [13:16] Xfinity Comcast bill. And first it asked [13:19] me what kind of voice do I want? Do I [13:20] want a female or male voice? And then it [13:23] asked me for my full name and my phone [13:25] number so that it can make the call for [13:27] for me. Then the magical part is you [13:30] know while I was eating breakfast it [13:31] actually called Xfinity for me and found [13:34] that there's actually a way to reduce my [13:36] bill just like with AT&T earlier. So, [13:39] basically, I can get a promotional rate [13:41] of $60 per month. If I turn on autopay, [13:44] I can get another $10 off. And here's [13:47] the actual transcript of the [13:48] conversation, right? Which is quite [13:50] quite insane to see. All right, so here [13:52] is Muse talking to an actual human. So, [13:55] Muse asked to lower my phone bill and it [13:58] says, "Just a moment." And Muse said, [14:00] "Hey, this is Haley calling on Peter's [14:03] behalf. I'll transcribe his call for [14:04] notes. I'm calling to see if Peter can [14:07] keep the same internet plan but lower [14:08] the $84 monthly bill. Can you help me [14:11] with with that? Then the human custom [14:13] support asked me to confirm the first [14:15] and last name. I'm not really sure why [14:17] uh it said me in here. That seems kind [14:19] of weird, but I confirmed it and yes, it [14:23] went back and forth with the this [14:24] conversation and eventually it got the [14:27] human support agent to tell me that if I [14:30] turn on autoplay discount, I can get $60 [14:32] per month. Right. Okay. So, one thing to [14:35] call here is that Muse is representing [14:38] itself as my personal agent, not [14:42] pretending to be Peter. And based on [14:44] this transcript, you know, the Xfinity [14:46] agent actually wants me to call myself [14:49] to get the discount, right? They want to [14:51] talk to me. But the fact that this all [14:54] this research and found the discount [14:56] still saves me a lot of time. We've all [14:59] experienced how annoying it is to call [15:02] these companies and have to, you know, [15:04] dial one or say two to try to get to a [15:07] human to talk to and also wait on a line [15:09] for a human be put on hold and all that [15:11] kind of stuff. And now we have Muse to [15:14] basically do this stuff for us, which is [15:17] just a massive timesaver. So, in my [15:19] opinion, this is the killer feature of [15:21] Muse and I think everyone should give it [15:23] a try. And the funny thing is I got this [15:26] feature by asking Muse to basically [15:29] message the meta team and request access [15:32] to the feature for me, right? It's like, [15:34] hey, please give me access to the phone [15:35] thing. And uh, you know, Peter wants [15:38] Muse, please give him access. And I sent [15:40] the feedback for me. And this morning, I [15:42] actually got access. So maybe you can [15:44] also use Muse to kind of give feedback [15:45] to the product team behind it. All [15:47] right. So that's all 10 use cases that I [15:50] want to show today on how you can get [15:51] the most out of Muse. And overall, I'm [15:53] really blown away by the product. I [15:55] think Meta is now the leader in the [15:57] personal Asian race. You know, the whole [15:59] experience is just a joy to use. And [16:01] Meta also has a massive distribution [16:04] advantage. Meta can basically use the [16:06] same playbook that it did with stories [16:08] against Snapchat and threads against X, [16:11] which is basically promoting Muse across [16:13] Facebook, Instagram, WhatsApp, and all [16:16] of its properties. You know, Muse is [16:18] already a top three app in the app store [16:20] as I'm recording this after only a week. [16:24] And personally, I'm a little bit tired [16:26] of always migrating to a new AI tool or [16:29] a new AIA agent. So, I want to kind of [16:31] share my stack right now, which is I [16:34] still use Chat GPT for everyday work. I [16:36] use Grockbot for cloud-based work tasks [16:40] that can run without my computer being [16:42] on. And I've since switched to Muse for [16:45] basically all personal tasks. Honestly, [16:48] I can see Muse being the personal agent [16:50] that reaches a billion people and is [16:53] made me really bullish about Meta Stock, [16:55] too. So, go out there, give Muse a try. [16:59] It's free to use. uh have it negotiate [17:02] your bills, have it give you a [17:04] personalized feed, and all the other use [17:05] cases that I highlighted. And as always, [17:08] if you enjoy this video, please like and [17:09] subscribe for more practical AI [17:11] tutorials just like this. And I'll see [17:13] you soon.