Consumer AI Adoption: Who Wins, and What It Means for Stocks

Consumer AI adoption, through agents that act for people, could become the second way AI makes money after corporate productivity. If it takes hold, AI revenue comes from consumers and merchants as well as company budgets, computing demand shifts to agents that run all day, and some search and advertising profit moves to whoever owns the agent. Meta's Muse, with more than 10 million sign-ups in about six weeks, is the first large test. We think the winners will be the companies people already trust, which points more to Google and Apple than to Meta.

Below: why agents may succeed where voice assistants failed, who is best placed to win, what it means for the AI buildout after October's selloff, and four scenarios with likely winners and losers. The assessments, scenarios and stock groupings are our AI-assisted view. They are neither a forecast nor advice.

Phase one: AI paid for by companies

So far, the money in AI has come mostly from corporate budgets. Companies buy AI to make staff more productive, and the revenue shows up in cloud and enterprise software. In the September quarter, Microsoft's Azure grew 48%, Google Cloud more than doubled, and Palantir's US commercial revenue grew 148%.

Consumers have used AI heavily, but mostly through chatbots. ChatGPT has around 800 million weekly users and the Gemini app about 1 billion monthly users. A chatbot waits for a question, answers it and stops. Most of that use is free, paid for by the same capex that October's selloff called into question. Our post on the October 2026 AI selloff covers that in detail.

A reactive assistant helps only when you know what to ask and remember to ask it. A proactive agent does the remembering for you. That makes it something people might pay for, directly or through the merchants they buy from.

Why agents may succeed where voice assistants failed

This has been tried before. Around 2018, smart speakers were expected to make voice shopping a large market; one consultancy forecast US voice shopping sales rising from about $2 billion to more than $40 billion by 2022. The Information reported that only about 2% of Alexa owners had bought anything by voice that year, and that about 90% of those did not try again. Alexa reached about a billion interactions a week, but most were requests like the weather that earned nothing, and Echo devices were largely sold at cost. Insider reported in 2022 that Amazon's devices unit was on pace to lose about $10 billion that year.

Six gaps held voice assistants back. Agents close most of them.

Gap	Voice assistants	AI agents
Understanding	Rigid commands; anything off-script failed	Handle open-ended requests and ask follow-up questions
Reach	Each action needed a company to build a "skill"	Use websites and apps as a person would
Seeing what you buy	No pictures, comparisons or reviews	Show options, compare prices and ask for approval
Memory	Every request started from scratch	Learn preferences and past purchases
Who starts	You had to know what to ask	Suggest things first
Business model	Requests earned nothing; hardware sold at cost	Sit at the point of purchase and can earn fees

The new problems are cost per user, access to websites, and trust with money. The rest of this post is about those three.

Meta's Muse, launched on 8 September 2026, is the clearest example of the new model at scale. It keeps running after its app is closed, acts on schedules and events, and can message first. Each user gets a persistent, isolated computer with a browser that the agent uses to complete tasks, and a separate system called Sentinel decides whether each action is allowed, blocked or sent to the user for approval. There is a free tier, plus $20 and $100 monthly plans.

Muse metric (Meta Q3 2026 call, 21 Oct)	Figure
Sign-ups, US and Canada, first ~6 weeks	More than 10 million
Share of sign-ups using it on a given day	About 25%
Share of active users who shop with it	More than a quarter
Average purchase	Above $200 (includes travel)
Planned compute-cost efficiency gain	More than 10x within a year

Read these with care. Sign-ups overstate active use, and the purchase average includes travel. Meta has disclosed no transaction value and no commission rate. Muse is not yet global, partly because it is expensive to run, and since 20 September it can no longer place orders on Amazon.com. Only about 31% of shoppers in Shopify's holiday survey said they would let an AI make a purchase for them.

The voice era's real test was repeat use, and it still is. Muse's daily-use figure of about 25% of sign-ups is the number to watch as the novelty fades.

Meta is not alone. Google's Gemini Spark, shown at I/O in May, also works around the clock, though it launched only to top-tier subscribers. Amazon merged Rufus and Alexa+ into Alexa for Shopping in May and says 400 million customers used it in the past year. Apple's Gemini-powered Siri in iOS 27 is moving towards proactive suggestions.

Who is most likely to win mass adoption

Persuading hundreds of millions of people to let an agent spend their money is much harder than getting 10 million to try one. Four qualities are likely to decide who manages it.

1. Trust. People hand an agent their logins, calendar, messages and payment details, and overall trust is low: an April 2026 Cloaked survey found only 18% of Americans trust AI to keep their personal data secure. This is Meta's biggest handicap. Its record includes the $5 billion FTC privacy penalty in 2019, and by default Muse lets Meta use people's queries to improve its models, with an opt-out. A confidential version is due by the end of the year. Google ranked first in Cybernews' 2026 AI trustworthiness ranking, and Apple has built its brand on privacy. One result cuts the other way: in the Cloaked survey, 67% of Meta AI users were comfortable with AI handling their personal data, the highest of any platform. The open question is whether the next several hundred million users feel the same.

2. Deep pockets. Always-on agents are expensive, and most of this service will be given away free. The company that can fund it for years without relying on credit markets has an edge, especially with Treasury yields above 5%. Microsoft is still generating positive free cash flow. In the September quarter Meta's free cash flow was about minus $2.1 billion and Alphabet's about minus $9.1 billion, and Amazon's trailing 12-month free cash flow had turned negative by June. Apple spends a small fraction of what the hyperscalers do on capex and can rent Gemini models instead of building its own data centres.

3. Security that stays trusted. A single serious incident, such as an agent leaking account data or making unauthorised purchases at scale, could set a company back years. When Amazon blocked Muse, it said Muse did not identify itself and appeared to capture and store customer login credentials. Meta says Muse cannot see passwords or payment methods and that shared credentials go into secure storage. Neither claim has been independently tested, but the dispute alone made Muse less useful. Apple's approach of running as much as possible on the device is built to avoid this kind of dispute.

4. Being hard to block. An agent is only useful if it can reach the stores and services people use. Amazon can afford to block outsiders because it is the destination. Most merchants cannot. For them, blocking Google risks losing visibility in Search and Shopping, where much of their traffic comes from. Google also starts with data merchants already send it: retailers upload product feeds with prices and stock to Google Shopping, and airlines and hotels share fares and availability with Google Travel. An agent that reads data merchants chose to share has no need to scrape their sites, which removes the main reason to block it. Google's agents also run inside Chrome and Android, and Apple's inside iOS, where the agent works through the user's own device and is harder to tell apart from the person.

Still, several large companies are blocking outside agents, because they fear losing the customer relationship to them:

• Amazon. On 20 September 2026 it stopped Muse placing orders on Amazon.com, after Meta declined to leave Amazon out. It has also restricted shopping agents from Google and OpenAI over the past year, and wants people to shop through its own Alexa for Shopping.
• JPMorgan Chase. On the bank's Q3 call in October, Jamie Dimon said it is "not going to allow just any outside agent to come in," citing the risk of "rogue agents." The bank confirmed it blocks external agents, including Muse, from moving money or paying bills in retail accounts. Read-only access through approved data-sharing providers such as Plaid still works. JPMorgan is also piloting its own agent, Smart Cash, which moves customers' money between accounts.
• Charles Schwab, Kohl's and Neiman Marcus have also moved to restrict personal agents such as Muse.
• Most retailers, often by accident. An audit of 51 large US retailers by Billions Network and AmanTrue found that 5 explicitly restrict agents, 37 block them mostly through old bot rules written for scrapers, and none offers a way for a verified agent to complete a purchase.

Security is the stated reason, and it is a real one: Visa's Michael Jabbara called fraud from fake agents, and agents acting beyond what the user intended, "very real and very tangible." But the deeper motive is control. As Mike Danford of Adverio, which advises Amazon sellers, put it, the dispute is about "who holds the data." A bank or retailer that lets an outside agent sit between it and the customer risks becoming a supplier to that agent. The common response is to block outsiders and build an in-house agent, as Amazon has with Alexa for Shopping and JPMorgan with Smart Cash.

Not everyone is resisting. Walmart has opened its site to the agents customers prefer, at the same prices people see. Accenture's banking lead argues that "the institutions that win are usually the ones that find a way to participate safely rather than resist the change." For now, blocking favours agents that already own the device, the browser or the store. It also helps the card networks: if banks keep agents out of accounts, agent purchases are more likely to run through the card credentials Visa and Mastercard are building for agents.

The tipping point: small merchants first

Blocking is a choice that only large companies can afford. A small online shop cannot turn away customers, whether they arrive by search, social media or an agent. Small merchants are therefore likely to welcome agent traffic first, and the platforms they run on will show it before anyone else.

• Shopify opened checkout to browser-based agents in September. In the June quarter, it said AI-driven traffic and orders to its stores had tripled year on year, while stressing that the volume is still small next to its total sales of about $116 billion that quarter.
• GoDaddy, which serves millions of very small businesses, is earlier in the process. Its Airo agents help owners run their businesses, and its Agent Name Service is meant to let agents prove who they are. It has not reported figures on orders placed by agents.

If agent shoppers start buying more from small merchants that welcome them, large merchants that block agents will see sales leak to those rivals. That is the point at which blocking becomes too expensive, and the big merchants have to open up. Walmart has already decided not to wait. The share of orders arriving through agents on Shopify, and eventually on GoDaddy, is the best early sign of how close that point is.

	Trust	Ability to fund	Security record	Hard to block	Distribution
Meta (Muse)	Weaker	Mixed: strong ad cash flow, negative free cash flow last quarter	Mixed: approval system, but a credential dispute with Amazon	Mixed: blocked by Amazon; ad ties help with other merchants	Stronger: billions of users
Google (Gemini)	Stronger	Mixed: strong cash flow, rising capex, Search under pressure	Stronger	Stronger: merchant feeds, Search, Chrome and Android	Stronger: about 1 billion Gemini app users, plus Android
Apple (Siri)	Stronger	Stronger: light capex	Stronger: on-device first	Stronger: works at the phone level	Stronger among premium phone owners
Amazon (Alexa for Shopping)	Mixed	Mixed: trailing free cash flow turned negative	Mixed	Strongest inside its own store, weaker outside it	Stronger: 400 million users in the past year
OpenAI (ChatGPT), private	Mixed	Depends on raising capital	Mixed	Weaker: restricted by Amazon	Stronger: about 800 million weekly users

This is our qualitative assessment, made with AI assistance. Our read: Meta has the head start and the distribution, but the qualities that decide mass adoption favour Google and Apple. Apple's weakness is execution: its new Siri has slipped repeatedly and relies on Google's models. Google's weakness is that every agent purchase it handles may replace a Search ad it used to sell.

What it means for the next phase of the AI buildout

October's selloff came down to two questions: who pays for the buildout, and at what interest rate. Consumer agents change the first answer and do little for the second.

A new payer, arriving slowly. Meta plans to earn from Muse through commerce fees paid by businesses, subscriptions for advanced models or extra usage, and business subscriptions. That money comes from outside corporate IT budgets. But Meta's CFO said the revenue will build over multiple years, and Meta has given no date for when it will be material. Meta narrowed its 2026 capex range to $135–145 billion, the top end of its earlier guidance, in the same quarter its free cash flow turned negative. Agents make the spending easier to justify. They do nothing about the gap between spending now and revenue later, and higher rates make that gap more expensive.

A different kind of compute demand. A chatbot uses compute when someone types. A proactive agent uses it all day. Meta says it is building capacity to serve more than a billion Muse users over the next several years. Each user's persistent virtual computer is CPU, memory and storage work as well as GPU work, which helps explain why AMD's Lisa Su expects the CPU market to grow roughly tenfold by 2030. Efficiency pulls the other way: Meta wants Muse more than ten times cheaper to run within a year, partly by using off-peak capacity. Cheaper AI tends to increase total use, as the DeepSeek episode showed, but chip spending per user could fall faster than user numbers rise.

The profit pool moves. When an agent shops, the person may never see a search results page or an ad. Alphabet's Search revenue missed forecasts for a second straight quarter in October, and Google's Universal Cart across Search and Gemini is its attempt to own the agent before someone else does. Merchants are pushing back too: Patagonia has blocked Muse, and in one public test Muse recommended a competitor instead.

Four scenarios, with winners and losers, through end-2027

The odds are our rough judgement. "Gains" and "loses" describe how each group's business is likely to be affected relative to today's expectations, and say nothing about whether a stock is cheap or expensive.

Scenario	Odds	What happens	Likely winners, such as	Likely losers, such as
1. Helpful but not trusted with money	35%	People delegate cheap, routine purchases; revenue stays small through 2027	Alphabet, Apple, Amazon, Visa, Mastercard	Meta, relative to its spending
2. Phone and shopping platforms win	30%	Siri, Gemini and Alexa take most of the usage	Apple, Alphabet, Amazon, Qualcomm, Arm	Meta, OpenAI, Booking, Expedia
3. Muse breaks out	15%	Muse passes 100 million users and discloses commerce fees	Meta, Nvidia, AMD, Micron, Broadcom, Shopify	Alphabet's Search, Booking, Expedia
4. Trust stalls adoption	20%	A breach, fraud wave or regulation slows delegated buying	Apple; click-based businesses get a reprieve	Meta, data-centre chip suppliers

1. Helpful but not trusted with money: 35% (base case)

Many people use agents, but few let them spend much. Agents become normal for research, price comparisons, reminders and routine reorders, but anything over roughly $100 still gets a human click, and most banks and large retailers keep outside agents at arm's length. Surveys already point this way: comfort with delegated buying is around 52% for low-cost everyday items and 15% for high-value purchases. Repeat use settles at a useful but unspectacular level.

• For the buildout: inference demand grows steadily, and efficiency gains absorb much of it. Agents support the case for capex without rescuing 2027 free cash flow.
• Winners: companies with trust, distribution and cash flow that can afford to wait, such as Alphabet, Apple and Amazon. Payment networks such as Visa and Mastercard, which earn on each purchase whichever agent makes it.
• Losers: Meta, in relative terms, because it is spending most aggressively on a product that pays back slowly. Click-based businesses face slow erosion.
• How we'd know: Muse's daily use holds around a quarter of sign-ups; agent orders grow on Shopify but stay a small share of its sales; no large platform reports agent commerce fees as a material revenue line; and 2027 capex guidance is not raised because of consumer agents.

2. Phone and shopping platforms win: 30%

People prefer the agent already built into their phone or their usual store. Apple's Siri, Google's Gemini and Amazon's Alexa for Shopping take most of the usage, because they already hold the contacts, calendar, payment details and purchase history, and they score best on trust and blocking. Amazon's Andy Jassy has argued that outside agents lack that personal data, and the Muse block shows Amazon will defend its store. Banks and retailers that block outside agents push users towards the agents built into the device or the store.

• For the buildout: compute demand still rises, but more of it may run on the device, which favours phone chips over data-centre spending.
• Winners: Apple, which gets an agent without hyperscale capex. Alphabet, through Gemini inside Siri, Android, Chrome and Shopping. Amazon, inside its own store. Phone-chip designers such as Qualcomm and Arm. Shopify, whose merchants can plug into whichever platform agent wins.
• Losers: Meta's Muse economics, if it cannot become the default agent on someone else's phone. OpenAI, which lacks a phone or a store. Businesses that depend on search clicks, such as Booking and Expedia, as platforms answer travel and shopping questions directly.
• How we'd know: iOS 27 Siri and Gemini agent features report heavy shopping use; Alexa for Shopping keeps growing; more banks and retailers follow JPMorgan and Amazon in blocking app-based agents while working with Apple and Google.

3. Muse breaks out: 15%

Muse goes global, passes 100 million users, and Meta starts disclosing commerce fees. Meta reaches terms with Amazon or makes the block irrelevant, and its existing users turn out to trust it more than outsiders expected. The odds are lower because Meta has to close the trust gap and solve the blocking problem at the same time, at Amazon, at JPMorgan and across most large retailers.

• For the buildout: the strongest case for more capex. Always-on agents across hundreds of millions of users would lift demand for inference chips, CPUs and memory, likely by more than efficiency gains can offset.
• Winners: Meta first. Data-centre compute suppliers such as Nvidia, AMD, Micron and Broadcom, which designs custom chips for large AI buyers. Small-merchant platforms such as Shopify and GoDaddy, whose merchants welcome agent shoppers.
• Losers: Alphabet's Search business, as shopping intent moves to Meta. Amazon, which would face pressure to open its store on Meta's terms. Online travel agencies such as Booking and Expedia.
• How we'd know: Muse launches globally with daily use above a quarter of sign-ups; Meta discloses commerce fees; agent orders become a meaningful share of Shopify's sales; Amazon or a major bank reaches terms with Meta; and Meta raises capex citing Muse demand.

4. Trust stalls adoption: 20%

A serious incident, such as an agent leaking private data, a fraud wave using agent payments, or a regulator restricting automated purchases, slows delegated buying. Japan's Consumer Affairs Agency has already said existing mail-order rules apply to agent purchases. Consumer AI stays mostly chat and search, and companies that block agents feel vindicated.

• For the buildout: the consumer half of the monetisation case fades, leaving corporate productivity as the only proven payer. That makes October's financing worries harder to answer.
• Winners: Apple, whose privacy positioning becomes more valuable. Businesses built on clicks and search, such as Alphabet's Search and online travel agencies, get a reprieve as the old model lasts longer. Security and identity software would also benefit; we will cover that separately.
• Losers: Meta, which has the most riding on consumer agents and the weakest trust position. Data-centre chip suppliers priced for a large consumer agent market. Shopify's agent-checkout upside.
• How we'd know: a widely reported agent fraud or data incident; a regulator restricting automated purchases or requiring human approval for each one; Muse daily use falling well below a quarter of sign-ups, repeating the voice-shopping pattern.

How each group fares across the scenarios

Group, such as	1. Helpful but not trusted with money	2. Platforms win	3. Muse breaks out	4. Trust stalls
Meta	Loses ground	Loses	Gains most	Loses
Alphabet	Gains	Gains	Loses (Search)	Mixed: Search holds, agent plans stall
Apple	Gains	Gains most	Mixed	Gains
Amazon	Gains	Gains inside its store	Loses	Mixed
Visa, Mastercard	Gains	Gains	Gains	Mixed: fraud risk
Shopify, GoDaddy	Mixed	Gains	Gains	Loses agent upside
Nvidia, AMD, Micron, Broadcom	Mixed	Mixed	Gains most	Loses
Qualcomm, Arm	Mixed	Gains	Mixed	Mixed
Booking, Expedia	Slow erosion	Loses	Loses most	Gains a reprieve

Visa and Mastercard do reasonably well in three of the four scenarios, because they earn on the purchase whichever agent makes it. Meta has the widest range of outcomes: it gains most if Muse breaks out and loses in the other three.

What this could mean for selling puts

This section is an AI-assisted view and is not financial advice. For any cash-secured put, the maximum loss is the strike minus the premium received, per share, if the stock goes to zero, and the breakeven is the same number.

Agent news is likely to arrive with quarterly results, so puts that run past an earnings date carry the agent story's risk on top of the usual earnings risk. The widest range of outcomes, at Meta and the chip suppliers, is also where premium tends to be richest. Payment networks usually offer thinner premium for steadier exposure. The levelbox screener updates each name's trend and premium daily.

What to watch next, in order

1. Shopify's agent orders. Its Q3 report, due shortly, and Q4 report should show whether AI-driven orders keep tripling. This is the best early read on adoption, because small merchants cannot afford to block agents.
2. Black Friday and Cyber Monday. AI-referred traffic and agent checkouts. Adobe expects AI-referred traffic to US retail sites to rise 130% year on year.
3. Muse on Meta's glasses, due in November, and any global rollout or deal with Amazon.
4. Apple's iOS 27 Siri as it moves from beta to general release.
5. Q4 earnings (late January to February). Muse active users, any commerce-fee disclosure, Alexa for Shopping figures, and 2027 capex guidance.
6. Agent-payment volumes from Visa and Mastercard, and any agent-order figures from GoDaddy.
7. A big merchant opening up. A large retailer or bank that has blocked agents changing course would suggest the tipping point is near.
8. A trust event. Any large agent privacy or fraud incident, or a regulator restricting automated purchases.

Method and sources

• Muse features and pricing: Meta's Muse page and Axios (8 September 2026); TestingCatalog.
• Muse usage, monetisation and compute plans: Meta's Q3 2026 earnings call (21 October 2026), as summarised by Yahoo Finance and Quartz.
• Other agents: Google's blog and I/O coverage (May 2026) for Gemini Spark; Amazon's May 2026 announcement for Alexa for Shopping and The Next Web for its user figure; WWDC and MacRumors coverage for iOS 27 Siri.
• Voice shopping: The Information (August 2018) via Retail Dive and Voicebot.ai, including the OC&C forecast; Insider (November 2022) via Yahoo and Popular Science for Alexa's losses.
• Trust: Cloaked's April 2026 survey via PPC Land; Cybernews' 2026 AI trustworthiness ranking; the FTC's 2019 Facebook order.
• Agent blocking: GeekWire via PPC Land on Amazon; CNBC (13 October and 1 November 2026) and Fortune (15 October 2026) on JPMorgan, Schwab and the Billions Network and AmanTrue retailer audit.
• Small merchants: Shopify's Q2 2026 results and call (5 August 2026) via Seeking Alpha and Longbridge; GoDaddy's Q4 2025 earnings materials and Airo announcements.
• Shopper surveys and holiday forecasts: Shopify, Adobe and Amex figures as collected by Retailgentic (October 2026).
• Enterprise and cash flow figures: company Q3 2026 results as reported by CNBC and Yahoo Finance; Palantir's Q3 2026 press release filed with the SEC.
• Assessments, scenario odds and winners and losers: our judgement, made with AI assistance. They are not model outputs.

Links:

• Meta: Muse
• Axios: Meta debuts Muse, its long-planned personal AI agent
• TestingCatalog: Meta introduces Muse as a proactive personal agent
• Yahoo Finance: Meta Platforms Q3 earnings call highlights
• Quartz: Meta Q3 2026 earnings beat as Muse tops 10M users
• Google: The Gemini app becomes more agentic
• ppc.land: Amazon merges Rufus and Alexa into one shopping assistant
• The Next Web: Alexa for Shopping hits 400m users after Amazon blocked Muse
• MacRumors: Google confirms Gemini-powered Siri coming later this year
• Retail Dive: Only 2% of Alexa users reportedly purchase via the devices (2018)
• Voicebot.ai: The Information on Alexa voice commerce (2018)
• Yahoo: Amazon on pace to lose $10 billion from Alexa and other devices (2022)
• Popular Science: Alexa, why are you losing so much money?
• PPC Land: Only 18% of Americans trust AI with their data
• Cybernews: AI Trustworthiness Ranking 2026
• FTC: $5 billion penalty and new privacy restrictions on Facebook (2019)
• CNBC: JPMorgan Chase blocks outside AI agents like Muse and Instinct from accounts (13 Oct 2026)
• Fortune: Jamie Dimon bars outside AI agents like Meta's Muse from JPMorgan accounts
• CNBC: For JPMorgan and Amazon, fighting AI bots may not be a winning long-term strategy (1 Nov 2026)
• PPC Land: Amazon blocks Meta's Muse AI agent from shopping on its site
• Retailgentic: October agentic commerce roundup
• Modern Retail: Why the AI shopping agent wars will heat up in 2026
• TweakTown: Lisa Su on inference and the CPU market
• Yahoo Finance: Alphabet, Meta, Microsoft and Amazon results
• CNBC: Alphabet faces record post-earnings loss (29 Oct 2026)
• Seeking Alpha: Shopify expects Q3 revenue growth in the low 30s as it expands agentic commerce
• Longbridge: Shopify Q2 2026 call, GMV +32% and the agentic AI bet
• TechCrunch: Shopify opens checkout to browser-based AI agents
• GoDaddy: Q4 2025 earnings call prepared remarks
• GoDaddy accelerates Airo.ai momentum with new AI agents
• Payments Dive: Visa, Mastercard race to agentic AI commerce

This article is analytical and educational and is not financial advice. It names companies as worked examples only. Figures are a point-in-time snapshot as of 3 November 2026 and can contain errors. This article was generated with AI assistance. If you spot a mistake, tell us.