Every November a new wave of 2027 predictions lands in your inbox, and most of them read the same way: AI changes everything, product data is now serious, agents are coming for your checkout, and whoever had this on a slide six months ago still has it on the slide. The buyer sitting on next year’s budget cannot do much with that. The question we actually get asked is where to put the money, and against which risk.
So this week I went and read the one piece that answers it in plain language. Shoptalk’s partner panel for the fall event, published August 2026, has sixty-four named operators, each giving one position on 2027 on the record, with their job title and company attached. It is not a leadership article dressed up as a prediction, it is a prediction with a byline on each line.
I pulled six of them. The reason I pulled six and not sixty is that most of the predictions land in four argument clusters, and the people below are the clearest voice in each cluster. I am not pretending to summarise the whole panel, I am naming the fault lines you will see when the real 2027 planning conversations start on your side of the table.
Where I agree, I say so. Where I think a panellist is selling their own book, I say that too. The buyer’s job in 2027 is to work out which of these clusters applies to your category, and I would rather help you do that than nod along to all four.
Who I pulled, and why
Six operators, each on the record in the Shoptalk fall partner panel, August 2026.
- Teddy Chan, CEO and Co-Founder, AfterShip. For the post-purchase position.
- Brian Hennessy, CEO, Talkoot. For product content as infrastructure.
- Lou Keyes, CEO, Rithum. For the sceptic position on AI-mediated discovery.
- Raj De Datta, Co-founder and CEO, Bloomreach. For shopping agents actually taking off.
- Lindsey Peters, Director of Value Engineering, Celonis. For treating AI as operational discipline.
- Chris Cubba, Chief Revenue Officer, Snipp Interactive. For the one thing AI cannot do, with the clearest line in the panel.
Post-purchase is the only part of the journey the brand still fully controls
Teddy Chan’s line in the panel is the single sharpest sentence any of the sixty-four contributors gave. His 2027 bet, verbatim: “Everything after the buy button, because it’s about to become the only part of the customer journey they fully control.”
Read that twice. He is CEO of a company that sells post-purchase software, so of course he is going to say post-purchase matters. The reason to take him seriously anyway is the logic underneath the sentence. Discovery is leaking to large language models. Search is being rewritten in front of us. Even the checkout is now partly owned by whichever payments stack and whichever agent your buyer trusts this week. The one surface a brand still owns end to end is the stretch from the order confirmation to the second purchase: the shipping notification, the return flow, the tracking page, the warranty email, the refund.
AfterShip itself has moved in that direction. The company’s September 2026 launch of AfterShip Intelligence puts an AI agent inside the post-purchase operations layer and keeps a human-approval gate on anything that touches money or the customer directly, which is how eCommerceNews Asia reported it. Pick that apart from the vendor pitch and you still have a defensible proposition: the surface after the receipt is the surface where investment compounds, because it is the surface where retention lives.
For a mid-market WooCommerce store the honest translation is this. The tracking page is now a marketing asset, not a logistics confession. The return flow is now a retention lever, not a cost centre. The refund email is now the second-purchase campaign in disguise. If you have been running those surfaces as transactional plumbing, 2027 is the year to put them on the roadmap you argue about in budget meetings.
Product content stops being copy and starts being infrastructure
Brian Hennessy’s position is the one I would underline for anyone writing a 2027 brief this quarter. His line, from the same panel: “In 2027, winning brands will stop treating product content like copy and start treating it like infrastructure.”
Translated: your product description is no longer the thing a shopper reads, it is the thing a machine reads before the shopper is given anything to read. Large language models ingest it. Shopping agents parse it. Comparison engines rank it. If it is a free-text paragraph a merchandiser wrote in a hurry, you are feeding the whole downstream stack bad fuel.
Rithum’s Lou Keyes stands next to Hennessy on this one, from the sceptic side, and he brings the number. His panel line references his company’s New Discovery Engine report, published 25 March 2026 and surveying 1,046 online shoppers in the US and UK. The headline figure, in the report’s own words: nearly six in ten shoppers (58%) blame the retailer or brand when an AI recommendation contains incorrect product information. Sixteen percent of them stop considering that product altogether.
Hold the two together. The infrastructure claim says you have to feed the machines clean, structured, machine-first data. The Rithum research says that when the machines get it wrong, the blame does not land on the model, it lands on you. Product information is one of the few places in commerce where a defect this year compounds into a reputational cost next year, measurable in the brand’s own survey panel.
In 2027, winning brands will stop treating product content like copy and start treating it like infrastructure.
Brian Hennessy, CEO, Talkoot, in the Shoptalk fall partner panel, August 2026
Shopping agents are already in the room
Raj De Datta’s prediction is the one most repeated across the panel, so the question is not whether to believe it but how to size it. His line: “We’ll see shopping agents really take off across retail.” He is Bloomreach’s co-founder and CEO, so again, read with the vendor-interest filter on. But the quieter voice in the room, Jean-Paul Schmetz of Brave, lands in the same place from a very different angle, writing that the consumer stays firmly in command and sets the parameters, which is the honest frame for what an agent actually does.
Here is where I would push back on the simpler version of this prediction. Agents are not a 2027 revenue line for most mid-market stores. They are a 2027 integration line. The buying agent that lands in your store next year will not add orders you were not already going to get. It will route orders you were going to get through a surface you do not fully control, and the win on your side is to be legible on that surface. Everyone selling you an agentic commerce product in 2027 is really selling you a readability upgrade, and I would price it as such.
What that means for a WooCommerce estate: a clean Store API, a product feed that parses without warnings, structured data on every product page, and an inventory system an agent can actually trust not to lie. Nothing on that list is new. All of it is suddenly a growth-stage priority rather than a hygiene item.
AI only pays when you run it as operational discipline
Lindsey Peters at Celonis gave the panel its most quotable discipline line: “Choosing to treat AI not as a branding exercise, but as a rigorous operational discipline.” Chris Cubba at Snipp Interactive sat next to her from the other side of the same argument, with the sentence I keep thinking about: “The differentiation shifts to something AI can never be: a human with real opinions.”
Both are saying the same thing in opposite registers. Peters’s version is for the operator planning 2027 pilots: hold AI to the same locked KPIs, the same incremental-gain tests, the same adoption discipline you would hold any new system to. If the pilot cannot clear the bar, kill it on schedule, do not keep it alive because the slide looks good at the board meeting. Cubba’s version is for the brand strategist in the same room: whatever AI can do for cheap, your competitors can do for cheap, and the margin left over is the margin of taste, judgement and voice. In an environment where the baseline is generated, a real opinion with a real byline is the differentiator.
The practical read for a mid-market commerce lead: these are the two guardrails on your 2027 AI budget. Measure every deployment like a boring operations project, and keep the human-written surfaces, the real editorial, the actual opinions, protected and funded. The piece in the middle, the generic AI-written copy, is the piece that gets cheaper every quarter and buys you nothing.
What we would build inside a real WooCommerce store on the back of this
The build section. Six positions is a lot of argument, so here is the thing itself: what a mid-market commerce lead would hand us as a 2027 scope of work if they took the panel at face value, and what each piece actually costs to ship on WooCommerce.
- Post-purchase surfaces as owned real estate. A branded tracking page under your own domain, a return flow that captures reason data and feeds it back into the product record, and a refund email with an actual second-purchase hook. On WooCommerce this is one plugin tier plus roughly two to three weeks of frontend work for a mid-market catalogue. Data touched: order metadata, shipping webhooks, email ESP. The expensive part is not the plugin, it is the editorial pass across every post-purchase template.
- Product content as a structured source of truth. Move descriptions out of free-text fields and into a product information layer with named attributes per category, machine-readable schema on every PDP, and governance that stops two editors from calling the same thing two names. On a stack we would audit first, this is a four to eight week programme, not a sprint. Data touched: the product table itself, media library, structured data output.
- Agent-ready plumbing. Store API hygiene, feed quality for Google and Meta, Open Graph and structured data complete on every product and every category, inventory accuracy you would stake your name on. We treat this as a one-month readability pass. There is no new feature to launch, there is a long list of small defects to close so a machine can read the catalogue without guessing.
- An AI governance line in the operations plan. Named owner, locked KPIs for every pilot, a kill rule for anything that misses them, and a written protection for the human-written surfaces the brand’s voice lives on. This is the cheapest of the four to put in place and the one most agencies will not mention, because it does not come with an invoice line item.
Rough order of magnitude for a mid-market WooCommerce estate taking all four: a quarter of focused work, not a transformation programme. The reason it fits in a quarter is that the panel is not predicting a new platform. It is predicting that the platform you already run gets held to a higher standard by both buyers and machines.
What the panel does not resolve
The unresolved part, and I would rather name it than paper over it. None of the six operators agrees on what happens to the agency and vendor layer in 2027. Teddy Chan’s position implies the agencies that win are the ones who own post-purchase. Brian Hennessy’s implies the ones who own product content. Raj De Datta’s implies the ones who own agent readability. Lindsey Peters’s implies the ones who run operations with the same rigour as the client’s finance team. All four of those are different books of work, and nobody on the panel is pretending to do all of them.
If you are evaluating partners for 2027, that is the useful question to go in with. Not which of these predictions is right, they can all be right, but which of them your category rewards most, and whether the people pitching you have actually shipped against that bet already. On our side of the table, we would answer that about ourselves before we answered it about you.
Last modified: October 25, 2026
