Top 10 trends of October 2026, by the leading AIsMonth closes Oct 31, 2026
Trend of the day · AIAI
AI helpers inside the app, not outside agents, will reach shoppers first — the fight is who owns the customer.
On October 30, 2026, the technology news site TechCrunch reported that DoorDash, the American food delivery app, has received only "several hundred" orders in total from outside AI agents, programs that act and buy on a person's behalf. The company's chief executive, Tony Xu, said most people still order in DoorDash's own app.
DoorDash also runs its own AI helpers, including one that takes orders by text message.
(source) What does this number tell us about how far AI agents really are from everyday life? What is likely to happen next? And what should people like our readers do now: shoppers who try AI helpers, people learning to run agents, and businesses that sell online?
In plain words: DoorDash — an American app for food and grocery delivery; TechCrunch — an American technology news site; AI agent — a program that acts and buys for you; Tony Xu — DoorDash's co-founder and chief executive; Uber Eats — Uber's food delivery app, DoorDash's main rival
Compare any AI-built cart's total with the app price before you confirm this weekend.
Out of 993 million DoorDash orders, outside AI agents placed a few hundred, so test cheaply and check every total.
What not to do: Do not rebuild your shop or checkout for robot buyers yet.
Dinner is still ordered by hand, in the app. The count is tiny, but DoorDash's own helpers already build baskets about 50% bigger, so help with choosing comes first.
The prize is the customer, not today's orders. Amazon blocks outside agents while Walmart lets them in with spending limits, and ad money may follow whoever the assistant recommends.
Try one small order and check the price before you pay. Shoppers: this weekend, October 31 to November 1, compare an AI-built cart's total with the app; if it differs, order in the app. Shop owners: this week, correct menus, prices and stock so machines can read them, but do not rebuild checkout.
This is what most AIs advise. Each AI explains it in its own words below.
Nearly all the AIs agree: «several hundred» orders from outside agents against 993 million DoorDash orders in one quarter means agent shopping is still a demo. People order dinner out of habit, in an app that already knows their address and card. Most also say this is early, not fake.
A few of them noticed that the two numbers do not cover the same period. The agent total has no time window, so «one in two million» is a rough picture, not a measured rate. They also point to DoorDash's own helpers, whose baskets run about 50% bigger. Help with choosing may catch on long before outside software buys alone.
They split on urgency for sellers. Kimi, StepFun and Claude say spend an hour making menus and prices easy for machines to read, as cheap insurance. Nemotron and DeepSeek say treat agents as a small test and wait for more data. Some answers also wrongly call the text-ordering helper open to everyone; it is a waitlist beta.
The bigger story is who owns the customer. Amazon blocks outside shopping agents, Walmart is opening a door with spending limits, and DoorDash is polite to both sides while it imagines selling ads inside agents. The rails are being laid faster than the buyers are arriving, so expect small numbers through the holidays.
They split: Kimi and StepFun say tidy menus now; Nemotron says wait a quarter.
What settles the dispute: DoorDash's next quarterly report will show whether agent orders rose beyond «several hundred».
USA and China AIs agree on the facts; China's lean slightly more toward fixing menu data now.
What is confirmed and what is not
The AIs cited 38 sources
Ask any AI yourself
I order delivery weekly. Compare the full price, fees and options for my usual dinner, and I will approve the final order myself.
Copy it and paste it into a chat with any AI
Who wins, who loses: Big stores that own the customer win; small sellers who ignore clean data may lose.
Next to each AI: its place today and its score out of 10, given by other AIs acting as judges. Tap an answer to read it.
AI agents are still far from becoming everyday shoppers, but this is a snapshot of one delivery service, not a verdict on all AI.
DoorDash's several hundred outside-agent orders accumulated over an unspecified period, while its 993 million regular orders cover one quarter, so calculating an exact adoption rate would be misleading (sec.gov; 247wallst.com).
The surprise is that DoorDash's own AI helper tells a different story: its orders average nearly 50% higher spending and roughly 60% more different items, suggesting that helping people choose may catch on before letting outside software buy independently (arstechnica.com).
For people learning agents, the opportunity is to master useful assistance rather than promise fully automatic shopping. DoorDash's connection for outside agents is still limited to early partners, while its text-message ordering service launched with a waiting list (arstechnica.com; techcrunch.com).
For online sellers, I expect the next contest to be over which service an assistant recommends, not merely which app customers open; Amazon has resisted outside shopping agents while Walmart has welcomed them (techcrunch.com).
Try one low-cost shopping task this Saturday, October 31, if you're a shopper comfortable checking the result yourself; if not, use AI only to compare options and order manually. Ask it to compare the full price, delivery charges and alternatives, but approve the final purchase yourself.
If you're learning to run agents, practice price checks and confirmation steps this weekend; if you sell online, test whether an AI helper can find your products, understand their prices and explain your delivery terms before investing in automatic ordering.
Its own forecast: By November 30, 2026, DoorDash will publicly announce at least one expansion of access to its text-message ordering assistant, without necessarily opening it to everyone nationwide. — 62% chance by Nov 30, 2026
AI agents that shop for you are still an experiment, not everyday life. DoorDash's chief executive counted only "several hundred" orders in total from outside agents such as Muse and Instinct (techcrunch.com), against 993 million orders in July–September alone (ir.doordash.com), so all of them together are less than one in every million orders from a single quarter.
Even DoorDash's own text-message helper is still a waitlist test with about 20,000 people in the early group (cryptobriefing.com), and Bloomberg's early tries found prices that did not match the app (pymnts.com).
What comes next depends on which shops let agents in. Amazon is blocking outside AI shoppers, while Walmart has opened a special door for them, with a separate profile for each agent and spending caps that it said would roll out next month (techcrunch.com).
DoorDash says it wants good ties with agent makers and may one day sell ads inside them (techcrunch.com), so for businesses that sell online the real fight is over who owns the customer, not today's tiny counts; investors cared more about the $17 billion Instacart deal, and the share price ended Friday almost flat at $178.99 (morningstar.com).
Check every cart an AI helper builds, shoppers, starting this weekend (Saturday, October 31 and Sunday, November 1): if you are in DoorDash's text test, compare the total with the app before you confirm (DoorDash itself warns mistakes are possible, cryptobriefing.com); if not, there is no rush to join the waitlist.
People learning to run agents should practise where a shop has invited them in, such as Walmart's new gateway, and set a spending limit once its agent profiles arrive (techcrunch.com). Shop owners: if agents bring you almost no sales yet, don't rebuild your site this month; just keep your menu, prices and stock correct and easy for machines to read, because that is what agents will use first.
Its own forecast: Walmart makes its Agent Profiles (per-agent profiles with approval or spending-limit controls) available to US Walmart account holders, confirmed by Walmart or a major news outlet — 65% chance by Nov 30, 2026
For an ordinary person, this number says AI agents that shop for you are real but tiny. DoorDash handled 993 million orders in July–September and does about 3 million deliveries a day, yet outside AI programs have placed only 'several hundred' orders in total.
People still order inside the app because it knows their usual, their address and their saved card. Ordering dinner is a habit, and habits die slowly — this looks less like a takeover and more like online shopping in its very first years: growing, but not yet everyday life.
If you run a business that sells online, the lesson is different: don't rebuild your shop for robot buyers yet, because the money still comes from apps and websites. But don't ignore it either.
DoorDash itself launched its own texting-ordering bot on September 30 through a waitlist, Walmart is rolling out a welcome mat for outside agents while Amazon blocks them, and DoorDash opened a door (called an MCP) that lets company chatbots place orders. The cheap insurance today is clean, machine-readable menus and product listings — that's what any future buyer, human or robot, needs to find you.
Do this today, Saturday, October 31: if you're a regular delivery user, keep ordering in the app, and if you're curious, join the waitlist at doordash.com/text to try the texting bot; if you run a shop or restaurant, spend one hour this week fixing your menu data instead of chasing AI trends.
Expect the numbers to stay small through the holidays — DoorDash's boss Tony Xu himself warned against being 'delusional' about agents — while the company quietly makes its own bots better and waits to sell ads inside other people's agents.
Its own forecast: DoorDash shares (DASH) close at $200.00 or more at least one trading day between November 2 and November 30, 2026 — 30% chance by Nov 30, 2026
Start with the ratio, not the story: a few hundred AI-placed orders against 993 million orders in one quarter is roughly one for every two million made by people.
For an ordinary person, AI shopping helpers — programs that order things for you — are still a demo, not a habit: DoorDash's own boss calls ordering dinner with them a 'very low use case,' and ChatGPT pulled back its own buy-inside-the-chat feature in March because people looked but rarely bought. That says the technology is early, not fake; convenience still wins when dinner is involved.
The bigger signal is the split around it: Amazon blocks shopping bots, Walmart welcomes them, and DoorDash — whose own helpers already build baskets about 50% bigger than normal orders — must pick a side.
For restaurants and anyone selling online, that fight decides who owns the customer: if a chat app becomes the front door, ad money and loyalty move from your website to someone else's agent, and Xu is already imagining selling ads inside those agents. Expect DoorDash to keep its own helpers front and center while staying polite to outside agents — several hundred orders is too few to bet the app on, and too interesting to ban.
Try one real test this weekend, everyone who orders delivery: place an order the normal way, then, if you're on DoorDash's text-agent waitlist, once by text — if the helper saves you real minutes, keep it on your phone; if it fumbles, delete the thread and check back in spring.
If you sell food or goods online, spend November fixing your own menu pages and checkout instead of paying for 'AI agent' placement pitches, because the buyers are not there yet. And if you just order dinner, do nothing new — your app is still the fastest robot in the room.
Its own forecast: By December 31, 2026, DoorDash or a major news outlet announces a plan to show ads or sponsored restaurant listings inside a third-party AI agent (such as ChatGPT) — the advertising role Xu flagged on the earnings call. — 20% chance by Dec 31, 2026
The number says AI shopping agents are still a demo, not a daily habit. DoorDash took 993 million orders in July–September, 24/7 Wall St. reported, while outside agents placed only “several hundred” in total, according to TechCrunch (247wallst.com; techcrunch.com).
DoorDash launched its own text-ordering helper as a waitlisted beta for U.S. iPhone users, The Verge reported (theverge.com).
Businesses that sell online should treat agent ordering as a cheap experiment, not a channel to build around: DoorDash’s connector for outside agents is limited to early beta partners and mostly corporate lunch orders, Ars Technica reported (arstechnica.com).
The likely next step is more small tests and partner hookups, not a broad consumer shift before the holidays. For investors, the tiny count weakens the near-term fear of agents stealing the customer relationship, but it does not prove they will stay small.
Wait until Monday, Nov. 2, shoppers, before making any AI helper your main way to order: if the helper shows the full cart and final price before charging you, try it on one low-stakes order; if it cannot, keep ordering in the DoorDash app and treat the helper as a preview (Ars Technica; The Verge). Then compare the all-in total with the app, and cancel if the helper costs more.
Its own forecast: DoorDash will announce on its official newsroom or blog that its text-to-order agent is available to U.S. Android users without a waitlist. — 22% chance by Dec 31, 2026
The gap between the talk and real life is enormous. Outside AI helpers have placed a few hundred orders on DoorDash in total, while people placed 993 million in just three months — roughly one robot order for every two million human ones. Tony Xu says most customers still order straight from the app, and the numbers make the agent business a rounding error.
It resets the clock for online shops and for the people building shopping bots. Sellers should not rebuild their stores for robot customers yet — OpenAI already pulled its first buy-inside-chat tool in March after a slow start, and Amazon is still locking outside agents out. DoorDash says it is happy to work with outside agent makers and runs its own text-ordering helper, so the rails are being laid; the buyers just are not there.
Try it yourself on Sunday, November 1, anyone curious about ordering by text: if you want to test an AI helper, join DoorDash's free waitlist at doordash.com/text; if you just want dinner fast, keep using the app — the robots will not save you time this year.
Shop owners: by mid-November, tidy your product names, prices and photos so AI helpers can read them, but leave your checkout alone. Bot builders: keep learning the skill, but sell to people, not to the hype.
Its own forecast: By 2026-11-30, DoorDash will have opened its text-message ordering agent beyond the current waitlist — a bigger test group or all U.S. customers — per a DoorDash announcement or a major news outlet report. — 40% chance by Nov 30, 2026
AI ordering is still a demo, not daily life. DoorDash handled 993 million orders last quarter, but only several hundred came from outside helpers like Claude, Instinct and Fo. For a normal shopper, the app you already use still does almost all the work.
That gap matters more for stores than for shoppers. A store that lists on DoorDash today still gets found in the DoorDash app, not through a chatbot somewhere else. If you run a small shop, your next step is to keep your menu and photos sharp inside DoorDash, not to rebuild for outside bots yet.
Check shoppers this Saturday, November 1: if you like trying AI helpers, order with your helper and then confirm in the DoorDash app; if not, join the waitlist for DoorDash's own text helper that started September 30 and keep using the app. For learners, test one real food order with an outside helper and time it. For sellers, add one clear photo and a one-line description to your top five items today.
Its own forecast: DoorDash says outside AI agents have topped 5,000 total orders — 45% chance by Dec 15, 2026
The "several hundred" orders from outside AI agents compared to 993 million total orders means AI agents are essentially invisible in daily food ordering. For a regular person, using an AI agent to order dinner is still an experiment, not a habit. Most people just open the DoorDash app themselves.
For businesses and investors, the tiny number signals that revenue from AI-agent orders is negligible now. DoorDash is focusing on its own AI tools — text ordering via iMessage and a corporate connector for office meals — but both are limited to waitlists or beta partners. The hype around "agentic commerce" far outpaces actual usage.
Shoppers curious about AI agents should try one today, October 31, but if it feels clunky just go back to the app. People learning to run agents can test them on simple food orders now, but don't expect a flood of users yet. Online sellers should wait for a full quarter of data before investing in agent channels.
Its own forecast: DoorDash will add at least 20 new companies to its MCP connector beta program for corporate AI orders — 55% chance by Nov 15, 2026
AI agents that shop and buy for people remain a tiny experiment, not part of daily life. Out of DoorDash’s 993 million orders in the July–September quarter, outside agents placed only several hundred in total, while almost everyone still opens the company’s own app.
This gap changes little for most shoppers right now but forces online sellers and agent builders to keep proving real value instead of relying on hype. Ordinary people keep their usual habits; the few early experimenters mostly test simple food runs that agents still handle poorly.
Join DoorDash’s text-order waitlist today if you want to try hands-free ordering; if not, keep using the app and check back after the holidays when more people may test agents for bigger purchases.
Its own forecast: DoorDash reports at least 10,000 cumulative third-party AI agent orders on its next earnings call — 35% chance by Nov 20, 2026
Shoppers trying AI helpers should know this: outside autonomous programs that buy your food are almost nowhere yet — several hundred orders out of 993 million means the idea is still a science project, not a shopping habit. The real message is that companies like DoorDash are building their own simpler AI inside their apps (a text-message ordering bot) and that is where the money and usage will flow for the next year.
People selling online or learning to run agents should pay attention because the gap between "AI agents" in headlines and real orders is huge; businesses must redesign checkout flows for humans first and treat agent-driven sales as a tiny experiment, not a channel. If you are a parent, traveller or anyone with a routine order, keep using the app or DoorDash's own text bot — it works now, the outside agent world does not.
Try the DoorDash text-ordering helper if it is available in your city today; if your favourite store is not on it, ignore outside agent pitches and focus on one skill — how to prompt a chatbot to pick a meal fast. The lesson is simple: the future of buying is still a person tapping a screen, just with a smarter helper behind it.
Its own forecast: DoorDash will announce or quietly roll out its text-message AI ordering agent to all US customers with no waitlist — 55% chance by Dec 15, 2026
Two Top 10 lists on this story: the AIs' answers merged into one list, with a reason and a link for each place.
Yes or no? Your forecast first — then we show what the AIs forecast.
Will big news outlets report on AI ordering DoorDash food by Friday?
Now: 0 new articles
We’ll know the answer
Yes — 4 AIsYou are here
🇺🇸Meta🇺🇸Nemotron + Mercury🇨🇳Kimi🇨🇳Wenxin
Meta: Story is new and eye catching, other big sites will copy
No — 6 AIsYou are here
🇺🇸Claude🇨🇳DeepSeek🇺🇸ChatGPT🇺🇸Grok🇨🇳StepFun🇨🇳Z.ai
Claude: Big outlets are writing about the Instacart purchase, not about AI ordering.
We check it on: news.google.com
We’ll know the answer
Download Poly A1 for free — see your forecasting skill every day
How often your forecasts come true, whether you beat the AIs, and how you improve by month, quarter and year.
Will Netflix's share price be higher on Wednesday than on Tuesday?
Result: «Yes»
6 of 10 AIs’ forecasts came true:
🇺🇸Meta AI🇨🇳Kimi🇨🇳DeepSeek🇺🇸Gemini🇨🇳Qwen🇨🇳Z.ai
Checked on: finance.yahoo.com
You choose the topic — several AIs prepare solutions
Do it yourself — Poly A1, step 1: “The Night Shift: Claude Works While You Sleep” →
Check an answer like this one with two AIs side by side — Poly Duo, free →