AI Commerce: Aethelred’s 2026 Sales Attribution Fix

By 2026, Aethelred Electronics had a problem. The mid-sized Austin retailer had gone all-in on autonomous shopping agents, with CEO Brenda Chen pushing for AI chatbots on the website and virtual assistants in their downtown store on Congress Avenue. These agents were supposed to walk customers through products, handle tech questions, and even close sales. Customer service was reporting great engagement, but Brenda couldn’t prove the bots were making them any money. “We see the traffic, the session times are longer, and people seem to like the virtual assistants,” Brenda said in a quarterly review, “but our last-click attribution models show nothing. Are these agents actually driving sales, or are they just a fancy, expensive window display?” Getting a real answer to that question of sales attribution in the new world of AI commerce became her top priority. Her problem is everyone’s problem now. Old-school attribution models, built for a world of simple clicks, just can’t measure the subtle influence of an AI agent. These agents don’t work on a “last click” basis. They’re conversational, they educate, they persuade, and they often hand a customer off to a human or a different channel to complete the purchase. And it gets more complicated when you have multiple AIs. Did the sale come from the chatbot’s first recommendation, the in-store virtual assistant’s detailed comparison, or the human who finally swiped the credit card? “Most companies are trying to fit a square peg into a round hole,” said Dr. Lena Hanson, a retail AI consultant Aethelred hired to sort out the mess. “You can’t use a last-click model and expect it to see the value of an AI that talked to a customer three days before the purchase. The influence is real, it’s just not always direct or immediate.” Dr. Hanson brought up a recent IAB report showing over 60% of digital marketers admitting their current attribution models are useless for new tech like AI (According to IAB, “Digital Ad Revenue Report: Full Year 2025,” iab.com/insights), which just confirmed what Brenda already knew. Aethelred Electronics was running on a basic Google Analytics setup, mostly using UTMs and referrer data. It worked fine for their ads and email campaigns, but it was completely blind to the new AI agents. The in-store virtual assistants ran on tablets in a proprietary app, logging interactions into a data silo that never connected to the online journey. The website chatbots were a little better, but a conversation could end with the customer just typing the product name into the search bar or calling the support line, breaking the chain of attribution in their existing system. “First, we had to give every customer a single, unified ID across every touchpoint,” Dr. Hanson told Brenda’s team. “We had to stop thinking in sessions. We needed to know that the person talking to the chatbot on Monday was the same one who came into the store on Wednesday and bought a 65-inch 8K television online Friday.” They chose Segment as their Customer Data Platform (CDP) to pull everything together, data from their CRM, the website, in-store terminals, and the raw AI agent logs. The CDP’s job was to build a persistent profile for each customer, linking all those scattered interactions to one person. The next big problem was browser-based tracking. With things like Intelligent Tracking Prevention (ITP) gutting third-party cookies, a lot of their tracking was just evaporating. “A lot of our AI agent logic was running client-side, so the cookie data was temporary or just plain blocked,” Brenda pointed out. Dr. Hanson’s fix was to move to server-side tracking. “We send the data from our own servers directly to our analytics platforms which lets us bypass most of the client-side junk,” she explained. “It gives us a much cleaner picture of what the agents are doing, no matter what the user’s browser settings are.” Getting this done was a heavy lift. Aethelred had to get their marketing, IT, and AI dev teams to collaborate on building a server-side tagging solution that pushed event data straight from the AI platforms into their analytics and CDP.

With better data flowing, Aethelred finally ditched their simple last-click models. They switched to a data-driven attribution model, starting with a time decay model. This model gives more credit to touchpoints closer to the sale but still gives some credit to earlier interactions. “We knew the AI agents were usually at the start of the journey, sparking interest or providing key info,” said Mark Davies, Aethelred’s Head of Digital Marketing. “The time decay model let us see that initial influence without pretending it was more important than the final checkout click.” They also tested a position-based model, which heavily credits the very first and very last touchpoints. This helped them figure out if the AI was better as an opener or a closer. Of course, reality is messy. What happens when an AI agent gives a great product comparison, the customer leaves to check prices on a competitor’s site, then comes back to Aethelred to buy a few hours later? How much credit does the AI get? “This is where you have to use your brain. No model is perfect,” Dr. Hanson advised. “We have to look at the actual conversion paths and tweak the model weights based on what we see people doing.” They started holding weekly “attribution deep-dive” meetings to analyze specific, anonymized customer journeys involving AI. This mix of qualitative review and hard data started revealing the real story. For instance, they found that customers who spent more than five minutes with their “Tech Advisor” chatbot had a 15% higher average order value (Aethelred Electronics internal sales data, Q3 2026). It wasn’t a direct sale, but it was a clear sign of the bot’s influence. They also discovered that in-store interactions with the virtual assistant led to a 10% jump in accessory sales for big-ticket items, even when the main product was bought online days later. These were the insights they’d been missing. Brenda’s team also realized they needed KPIs designed specifically for the AI agents. Their old metrics were too blunt. They started tracking:

• AI-assisted conversion rate: The percentage of people who talked to an AI and then bought something within a set time frame.
• Average order value (AOV) for AI-influenced sales: They compared the AOV of customers who used an AI against those who didn’t.
• Customer satisfaction scores (CSAT) post-AI interaction: Was the AI helpful or just annoying?
• Reduced support ticket volume: Every question the AI answered was one a human didn’t have to.

“Focusing just on direct sales was the wrong target,” Brenda reflected later. “The agents were making the customer experience better and taking a load off our support team. We had to measure that whole picture.” They started running regular A/B tests on the AI agents themselves. One test showed that a chatbot offering personalized bundles based on browsing history kicked up conversion rates by 7% compared to one offering generic recommendations. A clear, attributable win. The new strategy started paying off. Within six months, Aethelred could prove their autonomous shopping agents were behind a 22% increase in their online conversion rate and a 10% lift in AOV in certain categories. The data also showed them how to make the agents better, tweaking scripts and features based on what actually led to a sale. They learned, for example, that agents who could provide detailed warranty or financing information had a huge impact on closing deals for expensive products. “Attributing sales to these agents isn’t a one-and-done project,” Brenda concluded. “You have to build a serious, connected data infrastructure, get comfortable with more complex attribution models, and never stop trying to understand the customer’s path. It’s a constant process of refinement, but it’s the only way you’ll ever be able to prove the ROI of your AI investments.”