Yesterday, Gladly declared retail’s two-bot era dead. The customer service platform launched agentic commerce, a single conversational AI that sells and serves inside the same conversation, built on top of its existing agentic service. One bot answers the sizing question, checks your order history, processes the exchange, and recommends the thing you actually wanted. No handoff, no “let me transfer you,” no second chatbot with amnesia.
The real story is not another AI launch. It is that the split between selling and serving was always an org-chart fiction, and the shopper was the only one paying for it.
What it is
Every retail site worth its pixels currently runs two (or more) AI agents that have never met each other. One is the shopping assistant: great at recommending products, clueless about your order history, physically incapable of answering whether your return qualifies for a free exchange. The other is the support bot: fluent in your tickets and orders, but built to close cases, not open wallets. As Gladly CEO Charlie Besecker put it, shoppers are the only people touching both ends of the transaction, and they end up frustrated, stuck in loops, and passed from one unhelpful chatbot to the next.
Gladly’s move collapses both into one conversational AI running on one customer record. The architecture matters more than the marketing. Gladly built its platform around the customer instead of the session, so the AI that answers the product question and the AI that handles the return are literally the same AI with the same memory. The company claims agentic commerce is the first product that actually needed that whole architecture at once.
The commercial terms are just as aggressive as the product claim. Gladly sells it as a subscription priced on outcomes, not seats, guarantees it will beat the brand’s current AI on conversion rate, resolution rate, and customer satisfaction within 30 days or the brand pays nothing, charges no implementation fees, and says a typical launch takes about a week. That is a vendor putting its money where its press release is, and it tells you how confident they are that the two-bot setup is leaving money on the floor.
What changed
The context is a traffic wave nobody in ecommerce can keep ignoring. Adobe Analytics data covering more than a trillion visits shows AI-driven traffic to US retail sites rose 393% year over year in the first quarter of 2026. In March 2026 that traffic converted 42% better than non-AI traffic and generated 37% more revenue per visit. A separate Adobe survey of more than 5,000 people found 39% of US consumers now shop with AI, and 85% say it improved their experience.
Read that again: AI-assisted shoppers are your best shoppers, and they arrive with exactly one question left. The purchase is pre-loaded by ChatGPT or an AI overview or a comparison agent, and all that stands between you and the sale is whether your site can answer the final question. If the shopping bot can’t see the order history and the service bot can’t sell, you have built a wall at the exact moment the shopper is ready to climb through it.
That is the funnel everyone is missing. The industry obsesses over getting AI traffic to the site and then greets it with two bots that refuse to talk to each other.
What works
The early proof is Balsam Hill, and the numbers are genuinely strong. During the tool’s first 30 days at the holiday decor brand, one in five shoppers who viewed a product in chat added it to their cart. Resolution rates ran from 80% to 91%. First-month add-to-cart value cleared six figures. The deployment now spans the US, UK, and Australia.
The quote that sells the whole concept comes from Joe Balczo, senior strategist for AI and CX technology at Balsam Brands: customers would ask the shopping bot something only service could answer, or ask service something only shopping could answer, and either way they were the ones figuring out which box to talk to. Now it’s one conversation. That is the entire pitch in two sentences, and it came from a customer, not a marketer.
Why it works is structural, not cosmetic. Support conversations are purchase conversations. A sizing question is a buying question. An exchange question is a second purchase waiting to happen. Every time your service bot closes a case without selling, it walks a warm buyer out of the store. And every time your shopping assistant shrugs at an order question, it leaks a customer who was already sold. The unification is not a feature; it is the removal of a tax the shopper was paying on your behalf.
What breaks or stays fenced
Now the honest part. This only works if the customer record is real. Gladly’s architecture is customer-centric, which sounds obvious and is anything but: most brands’ order history, browsing data, loyalty status, and support tickets live in four different systems that barely wave at each other across the API gap. A single conversation needs a single source of truth. Brands without unified customer data are buying the car without the engine.
The 30-day outperform guarantee is also narrower than it sounds. It promises to beat your current AI on conversion, resolution, and satisfaction, not to beat your humans, and “outperform” is measured against whatever bar your existing bots set, which for most retailers is a bar lying on the floor. Read the terms before you read the headline.
And expect noise. Agentic commerce is now the industry’s favorite buzzword, with everyone from InMobi to the NBCUniversal/Carat upfronts machine shipping “agentic” everything this week. Most of it is a chatbot with a press release. The differentiator to interrogate in every demo is the same one: does the AI see the full customer, or does it see a session? If it can’t see the order history while it sells, it’s two bots wearing one trench coat.
Who it is for
Ecommerce brands with meaningful chat volume and a visible handoff problem. If your chat logs show shopping questions landing in service (or vice versa), you are already paying the tax. Holiday-season retailers especially: Balsam Hill is a decor brand, which means the playbook is proven in considered, seasonal, high-question purchases, exactly where chat matters most. Brands on Shopify and the major commerce platforms where the integrations already exist get the one-week launch story; custom-stack enterprises should budget for the data plumbing, not the AI.
Also for the CX leaders who have been quietly arguing that service drives revenue and losing to the “cost center” framing. This launch is your ammunition. The support conversation is a sales conversation now, with a vendor willing to price on outcomes.
What to do this week
- Audit your own two-bot tax. Pull a week of chat transcripts and count how many shopping questions hit your service bot and how many service questions hit your shopping assistant. That is your leaked revenue, in raw text.
- Measure the handoff, not the chatbot. Your bot vendors report resolution rates per bot. Almost nobody measures the drop-off between bots. Build that metric this week; it is the business case for unification.
- Ask one question in every agentic demo. Does the AI operate on the full customer record or the session? If the answer is the session, it is the same two-bot tax with better copywriting.
- Unify the customer record before you unify the bot. Order history, browsing, loyalty, tickets: one profile. That is the real project. The AI is the easy part.
Gladly did not invent the idea that shoppers hate being bounced between bots. It just built the first platform that acts like it, and put a 30-day guarantee on the bet. The line between selling and serving was never real to the shopper. It only ever existed in the org chart. Now it is a line item, and vendors are competing to erase it. Make sure your bots get the memo before your shoppers give up on both of them.
Sources: MarTech Cube, Oct 6, 2026; Daily AI Brief, Oct 6, 2026; Gladly press release and Balsam Hill case study, Sept 2026; Adobe Analytics via MarTech Cube.
