TikTok’s Agentic Commerce Push, Explained

Your Pre-BFCM Shopify Playbook: 16 Hacks in 90 Minutes

Black Friday is close, and the next few weeks decide your Q4.

On Oct 28 (2PM EST / 7PM CET), 16 Shopify operators each share one growth hack they’ve already tested on real stores. It runs for 90 minutes, it’s free, and nobody is selling anything. Every hack is fast enough to put live before BFCM.

You’ll learn how to:
→ Turn up to 30% of dead email leads into buyers
→ Run 4 copy-and-paste CRO tests that made real money
→ Make BFCM creatives without another photoshoot
→ Get your brand recommended by AI
→ Win at TikTok Shop with creators
→ Cut cross-border return costs

Speakers include Chase Dimond and Will Laurenson, and Nick Penev hosts.

If you only show up for one ecommerce webinar this season, make it this one. Save your free seat →

TikTok’s Agentic Commerce Push, Explained

TikTok used Advertising Week New York to unveil its biggest batch of AI features yet, timed (entirely coincidentally, I’m sure) a few weeks before Black Friday. Shoppers get a conversational Shopping Assistant and a one-click checkout called Buy Direct. Advertisers get, well, nearly everything else.

Most coverage has filed all of this under “agentic commerce.” TikTok uses the phrase too, because launching anything in 2026 without it is presumably a regulatory offence. But the term means something. An agentic system doesn’t just answer questions. It takes steps toward a goal on your behalf, like picking a product, setting a bid or completing a purchase, with limited human supervision.

Hold the announcements up to that definition and something interesting falls out. The tools that genuinely act on their own sit almost entirely on the advertiser’s side of the counter. The shopper tools help and shorten the route to checkout, but the human still makes the final tap.

That split is not an accident. And new consumer research suggests TikTok has read the room rather well.

The Shopping Assistant That Knows Its Place

According to TikTok, the Shopping Assistant understands context and remembers what you’re looking for (a higher bar than most of us clear in a group chat). It answers questions about product details, shipping, sizing and availability. It then carries the shopper through to purchase in a single conversation.

Buy Direct is the other half: one-click purchase directly from a brand, inside the For You feed. TikTok is building it with Salesforce, Shopify, Shoplazza and Stripe, which suggests the plan is to make brands’ own stores buyable inside TikTok. How this coexists with TikTok Shop’s existing checkout hasn’t been explained (two checkouts under one roof, what could possibly get confusing).

Notice what neither tool does: buy anything without you. The assistant guides. Buy Direct removes the steps between “ooh” and “paid.” This is a faster checkout, not handing over the decision. For contrast, Amazon’s Rufus can already be told to buy an item automatically once it drops to a set price. TikTok has stopped short of that.

Shoppers Will Outsource the Homework, Not the Decision

YouGov’s research on this year’s sales season explains the restraint. Its July–August survey of US adults found 34% of likely Black Friday and Cyber Monday shoppers expect to use AI to navigate the sales. Last year’s figure was 16%, though the question was worded differently. So read it as “clearly up” rather than “precisely doubled” (statisticians everywhere, you’re welcome).

The revealing bit is where the comfort stops:

• 43% are happy using AI to find the best price.

• 37% would use it to track price changes, and 34% to compare products.

• Only 9% would let AI add items to their basket.

• 5% would let it complete a purchase after they approve it.

• 4% would let it buy without asking.

So consumers are drawing a firm line between research and buying. The machine can fetch and sort. The human keeps the decision. Roughly how most of us treat a very well-read intern.

TikTok’s tools sit squarely in that accepted zone. The logic is tidy: there’s more to gain by removing friction around a human decision than by asking people to give it up before they’re ready.

Trust is thin even inside the comfort zone. 80% of likely shoppers trust checking retailer sites themselves to find the best deal, versus 48% for AI tools. Among people who’ve shopped with AI before, confidence in AI climbs to 83%, and the habit sticks: 92% of last year’s AI shoppers expect to use it as much or more. Still, 56% worry about inaccurate information and 51% about sponsored recommendations.

One detail matters more than it first appears. Asked which tools they’d use, shoppers named ChatGPT and Gemini (29% each), then Copilot (14%). All standalone assistants. It’s not clear anyone would describe asking TikTok’s assistant about sizing, or Rufus to compare two products, as “using AI to shop” at all. If they wouldn’t, AI’s real influence on holiday purchases may be higher than the surveys suggest. The assistants built into apps are the ones shoppers notice least. (The most influential AI is the one you don’t realise is in the room. Like a good butler. Or a tracking pixel.)

Meanwhile, on the Advertiser Side of the Counter

Here, TikTok is far more relaxed about letting software drive.

Smart+ now extends to search ads, automating keywords, creative, bidding and budget. It treats TikTok search as a buying-intent channel in its own right. That puts it after the “I am looking for this exact thing” moments Amazon and Google have historically had to themselves (and been paid handsomely for).

Smart+ Creative Selection pulls your own assets, AI-made creative from TikTok’s Symphony tools and creator content into one approval workflow. It then recommends which groups to run. You keep final approval. The shortlist, however, is the machine’s. (Rather like being free to order anything from a menu the restaurant wrote.)

GMV Max shifts TikTok Shop budget toward content that’s actually selling. It will now factor in affiliate commissions, coupons and referral fees, so it aims for return rather than revenue alone. That’s a genuine improvement. A campaign driving heavily discounted, high-commission orders can look magnificent on revenue while quietly setting fire to your P&L.

But the costs TikTok named are the ones it can see inside its own walls. As far as the announcement indicates, product cost, fulfilment and returns aren’t in the calculation. Neither is the cost of selling stock here instead of on a higher-margin channel. The algorithm’s definition of “true” return is narrower than yours. (Your accountant’s is still the only definition the taxman cares about.)

The Ad Network arrives in the US, extending campaigns to roughly 400,000 third-party apps worldwide. Integral Ad Science will independently measure viewability, invalid traffic and brand safety.

The most structurally significant piece is the least glamorous. TikTok for Business has a server built on the Model Context Protocol (MCP), an open standard that lets AI assistants take actions in other software. You ask an AI tool to build a campaign or analyse performance, and it does so directly in TikTok’s ad system. TikTok says MCP use is up more than 200% since launch (a figure verified by TikTok, which is a slightly different thing from verified). It’s now available on Claude, Perplexity, Manus and others.

So the asymmetry is plain. TikTok is perfectly comfortable letting machines decide where advertising money goes. It is not yet asking consumers to let machines decide where their money goes. For sellers, the job shifts from pulling levers to feeding the automation what it runs on: accurate cost data, a deep creative library, clean product information and clear limits on what the system may do.

Your Product Data Is Now the Sales Assistant

If shoppers mainly use AI for research, whatever the AI reads becomes the point where the sale is won or lost. The assistant can only answer sizing and shipping questions as well as the data behind it allows.

This is where YouGov’s accuracy worry becomes a commercial risk. Say an assistant gives a wrong delivery estimate or a dodgy size recommendation. The shopper experiences it as your failure, not the robot’s, and you may never know it happened. (The silent one-star review, now fully automated.)

Amazon sellers know this from Rufus. The difference now is that the same product may be described by assistants on Amazon, TikTok, ChatGPT and Gemini, each pulling from slightly different sources. List one delivery window on Amazon, another on Shopify and a third in a TikTok catalogue, and every assistant gets a fresh opportunity to confidently repeat the wrong one.

The most useful pre-Black Friday prep may be the least exciting. Make size charts, materials, compatibility details and shipping times complete and consistent everywhere. Not keynote material, but neither is losing a sale because a chatbot thinks your medium is a large.

The Bottom Line

TikTok has drawn its line: machines can spend advertisers’ money, but shoppers still spend their own.

If shoppers stay in research mode while platforms automate the selling, the fight moves to the data machines read rather than the ads people see. And if brands can’t see the conversations behind their sales, building their own AI tools will rest on thinner data than they assumed.

And that 4% willing to let AI buy without asking? Small today. Whether it stays small, or moves the way AI-assisted shopping has moved this past year, will decide where TikTok’s line ends up.

• Reddit’s real value isn’t AI citations, which hinge on a $60M Google deal now in tense renewal talks. It’s where buyers, especially developers, compare tools mid-funnel.

• Meta’s Business Assistant gains memory, brand-guideline imports and agentic powers to change creative, targeting and budgets by prompt. Add Shopify checkout inside Messenger chats, and Meta wants the whole funnel inside its AI.

• Amazon ad spend before Prime Big Deal Days is up 15-40% YoY, with budgets shifting up-funnel into DSP, Prime Video and CTV. But shrinking discount depth could cap how far that growth goes.

• Claude Dashboards connects to BigQuery, Snowflake or Salesforce and keeps charts live, each showing its query. Claude Motion turns reports into editable animated explainers. Docs, Slides and Design exit beta on every plan, Free included.

• Google launches the Gemini agent: one agent for chat, knowledge work and code, working inside Workspace and spinning up coworker agents with their own email accounts. It routes jobs across Gemini and Claude models to cut costs.

• SE Ranking’s 100,000-keyword study finds social links in 67.66% of Google AI Overviews but just 8.4% of ChatGPT answers. YouTube leads Google, Reddit and LinkedIn lead ChatGPT. AI visibility now needs a per-platform playbook.

📹 Captiwiz - Create videos with AI-powered captions.

🌪️ Sune - Designed to help your team collaborate and automate workflows.

📋 Deformity - AI-driven forms that break the mold.

🤖 Shot Solve - Instantly solve your problem with GPT-4 vision.

😍It was great, all I needed

🙂Good, not great

🙄It sucked