Parcel Perform Rebrands as Perform.AI and Launches an AI Commerce Operating System: Are Delivery and Returns Records What Get Brands Recommended by AI Agents?
Delivery tracking specialist Parcel Perform has relaunched as Perform.AI with an AI commerce operating system. We explain its four modules and MCP access, the evidence and caveats behind the claim that AI agents recommend brands on delivery performance, and the delivery and returns data merchants should audit.
Key Takeaways
- On September 22, 2026, delivery tracking specialist Parcel Perform renamed itself Perform.AI and launched an "AI Commerce Operating System." Four modules (AI Commerce Visibility, Checkout, Post-Purchase and Returns) plus a Logistics layer are run together by AI Decision Intelligence, which optimizes for gross margin, and teams can query their own e-commerce operations data from the AI tools they already use via MCP
- The core claim is that "AI agents are not moved by advertising or loyalty, and recommend brands based on whether they delivered what they promised." However, the public documentation from ChatGPT and Google names price, availability, and merchant-declared delivery times and return terms as evaluation and ranking factors, and does not say that actual delivery performance is read directly
- Pricing, the number of companies on the new system and outcome metrics are undisclosed. For merchants, aligning the delivery and returns "promises" shown on product pages, feeds and checkout, and measuring in-house how well those promises are kept, is preparation that holds up under any AI's evaluation criteria
Parcel Perform Relaunches as Perform.AI
Perform.AI has announced the launch of an AI Commerce Operating System for end-to-end ecommerce.
thepaypers.com"An AI agent deciding which brand to recommend reads what you promised, and checks what you actually delivered." With those words from co-founder and CEO Dr. Arne Jeroschewski, delivery experience platform Parcel Perform announced in a press release on September 22, 2026 that it had renamed itself Perform.AI and launched the "AI Commerce Operating System."
The company was founded in 2016 as Parcel Perform and, from bases in Singapore, Berlin and Chicago, has provided delivery tracking and post-purchase experience software to e-commerce merchants. According to the announcement, its system runs on more than 100 billion parcel updates a year for thousands of brands, with more than 1,100 carrier integrations across over 160 countries.
Explaining the rebrand, co-founder and Chief Customer Officer Dana von der Heide said performance was always the point for customers, and that "parcel was just the unit of work we began with." Payments trade outlet The Paypers also covered the launch on September 25, relaying the company's view that brands must now design experiences and data for both the people who buy and the AI agents that assist them.
The Core Claim: AI Chooses on Performance, Not Advertising
The boldest part of the announcement is less the product itself than the market view underneath it. Perform.AI describes the AI agent as focused on finding the best product at the best price with the fastest delivery, and says it picks only the brands that keep their promises.
Jeroschewski put it even more directly: "Loyalty doesn't move an AI agent. Advertising doesn't move it either. But consistent performance does." In other words, the operational record itself, what delivery time a merchant promised and what it actually achieved, becomes a condition for being recommended. He added that merchants can see that record before the agent does, and fix what is in it.
The claim did not appear out of nowhere. In October 2025 the company announced the enterprise beta of AI Commerce Visibility. It covers eight AI platforms including ChatGPT, Perplexity and Gemini, and analyzes brand visibility, presence at the product category level and "AI Trust Signals," the information AI relies on as grounds for trust.
That announcement stated that when AI picks the "best" brand, it also analyzes real-world data such as delivery speed and return policies, and called this GEO (Generative Engine Optimization, optimizing to be chosen in generative AI answers). Unlike SEO, it argued, GEO is won by proving real-world performance. In February 2026 the company also released a free AI Visibility Index that ranks the retailers AI recommends each week. The new operating system connects this "get found by AI" capability with the company's core delivery tracking business in a single system.
How the AI Commerce Operating System Is Structured
The system splits the e-commerce customer journey into four stages and places a module at each.
| Stage | Module | Role described in announcements |
|---|---|---|
| Get found | AI Commerce Visibility | Shows how AI tools recommend the product lines a brand sells and what they say about its delivery and returns (beta announced October 2025) |
| Purchase | Checkout | The checkout experience. The February 2026 announcement cites showing an estimated delivery date (EDD) at checkout |
| After purchase | Post-Purchase | The post-purchase experience from order to delivery |
| Returns | Returns | The returns experience |
| Foundation | Logistics | The layer handling day-to-day work with shipping partners |
| Across everything | AI Decision Intelligence | Reads the data, decides the next action and executes it, optimizing for gross margin |
Beneath the four modules, the Logistics layer handles day-to-day work with shipping partners. Running across everything is AI Decision Intelligence, which is said to read the data, hold a view on what should happen next, and act on it. The optimization target is neither revenue nor shipping cost but gross margin. The announcement gives examples such as catching a delay before a customer emails to ask, and picking the cheapest carrier among those most likely to hit a promised date.
The other notable feature is support for MCP (Model Context Protocol). MCP is a standard for connecting AI applications to external data and tools, and once a business is live on Perform.AI, its marketing, logistics operations and customer service teams can ask questions about their e-commerce operations data from the AI tools they already use. Instead of clicking through a dedicated dashboard, teams ask a familiar AI assistant how operations are going.
Jeroschewski frames this as a break from the past decade of e-commerce software, in which merchants bought the tool and did the work themselves. The system takes on the work, and people focus on strategic decisions.
That is as far as the public information goes. Pricing, the number of companies using the new operating system and outcome metrics from deployments are all undisclosed. The "thousands of brands" figure refers to the existing delivery data platform, not adoption of the operating system. It is also unclear what data is accessible through MCP, whether it allows actions as well as queries, and how far AI Decision Intelligence executes without human approval. The announcement does not say whether AI Commerce Visibility has moved from its October 2025 beta to general availability.
Are AI Agents Really Choosing on Delivery Performance?
How well supported is the claim that "the agent checks what you actually delivered"? The public documentation for the major AI shopping experiences shows that, for now, what is explicitly documented is the promise rather than the track record.
OpenAI's help article "Shopping with ChatGPT Search" explains that product selection draws on structured metadata such as price and product description, third-party content, and reviews from public websites. When multiple merchants sell the same product, it lists factors like availability, price, quality, and whether the merchant is the maker or primary seller, but says nothing about delivery reliability. OpenAI's product feed specification for merchants includes fields for shipping price, whether returns are accepted, the return window, the return policy and store ratings, yet it states that a shipping amount does not guarantee a delivery date.
Google's Store Quality program evaluates delivery time and cost, return window and cost, browsing experience and purchase experience, and awards a "Top Quality Store" badge to merchants at the top level. Including delivery and returns in the criteria is close to Perform.AI's view, but delivery time is described as a value the merchant sets in Merchant Center or via the API.
The route by which actual delivery quality reaches AI, as far as public documentation shows, is indirect: through reviews and store ratings. Late deliveries can turn into negative reviews whose summaries surface in answers, but there is no documented statement that AI checks carrier tracking data directly. It is also worth keeping in mind that a company whose core business is delivery data has a commercial motive to argue that delivery performance decides recommendations.
Consumers are not uniformly on board either. In an ACI Worldwide and YouGov survey reported by The Paypers the same day, 53% of more than 3,300 fashion and sportswear shoppers in the UK and US were uncomfortable letting an AI assistant make purchases on their behalf, and only 7% would allow autonomous purchasing under predefined conditions. Interest in price comparison and price drop alerts was high, so AI's role for now centers on the "find and compare" stage. What recommendations are based on is being questioned even before delegated purchasing spreads.
Delivery and Returns Data Merchants Should Audit
The "promises" AI reads today and the "track record" that reaches it indirectly through reviews both come from the same delivery and returns operations. The audit works best in the following order.
The first thing to check is whether the promises are consistent. Compare four places: delivery estimates on product pages, delivery times and shipping costs registered in Merchant Center or feeds for ChatGPT, the delivery date shown at checkout, and the shipping policy page. If the numbers conflict across these four places, AI will build its answers from contradictory information. Watch too for feeds that update slowly and fail to reflect peak-season delays.
Next, hold a number for how well you keep those promises. Looking at the share of orders delivered by the promised date, broken down by carrier, region and product category, exposes weaknesses the overall average hides. The volume and share of "where is my order" (WISMO) inquiries also works as an indicator of how often customers doubted the promise. That Perform.AI's announcement uses catching a delay before the customer asks as a concrete example reads as a sign that these inquiries directly reflect a deteriorating record.
The same thinking applies to returns. Check whether you register whether returns are accepted, within how many days, and who pays return shipping as structured feed fields, not just on a written policy page. Both OpenAI and Google treat return terms as separate fields, and under OpenAI's specification anything left blank is treated as unspecified.
Finally, regularly check what AI actually says about your company. Ask ChatGPT, Gemini and Perplexity whether your brand delivers quickly and whether returns are easy, and look at the reviews and articles cited as grounds for the answers. This is exactly what Perform.AI's AI Commerce Visibility productizes, but with a small set of questions merchants can start on their own. If old reviews or articles about past delivery problems are being used as evidence, the fix is not advertising but improving operations and updating the information that shows the results.
Conclusion
Perform.AI's announcement is a delivery tracking company re-centering its business on "being chosen by AI." The claim that AI picks recommendations based on track record is not yet backed by the public documentation of the major AI shopping experiences, and what AI explicitly reads today are the delivery and returns terms merchants declare. Even so, aligning promises, measuring how often they are kept and checking how AI talks about you will not be wasted whichever way evaluation evolves. The next things to watch are whether ChatGPT or Google explicitly name delivery reliability or store ratings as recommendation factors, and whether Perform.AI discloses concrete results from its customers.
