Seel Resolves 99% of Ecommerce Support Cases With AI That Owns Outcomes
Seel CEO Zack Peng says AI now resolves 99%+ of support cases, accelerates 80% of ticket volume, and carries liability for outcomes — a new bar for AI-dependent ecommerce service.
By Nathan Brooks
5 min read

What's News
- Seel resolves 99%+ of support cases with AI vs an industry norm of 60-70%
- AI accelerates 80% of Seel's ticket volume; 69% of repeat buyers protected their first order
- Returns and refunds run 10-20% of GMV; 10-20% of shoppers paid to make final-sale items refundable
- 49% of Gen Z shoppers would rather resolve issues with a person (Seel 2026 report)
- LLM Router bills per model usage, not per completed customer outcome
Seel resolves more than 99% of its support cases with AI, against an industry norm of 60 to 70%, by letting agents complete resolutions instead of deflecting tickets. Co-Founder and CEO Zack Peng laid out the numbers in an interview that doubles as a blueprint for merchants weighing AI agencies: AI accelerates 80% of ticket volume, resolution rates run north of 90% on end-to-end post-purchase cases, and 69% of repeat buyers across Seel's network protected their first order.
Where the model came from
Seel started with a single marketplace founder. Chargeback prediction, Peng said, absorbed real industry effort at roughly 0.5% of GMV, while returns and refunds — running 10 to 20% — went unmodeled. Seel built a prediction model and let shoppers pay a small fee to make final-sale items refundable.
Peng expected 3 to 5% opt-in. He got 10 to 20%. "That told us shoppers will pay for confidence after checkout, the same reason people buy AppleCare or a car rental waiver," he said.
The AI workforce emerged from Seel's own inbox. More than 70% of its support tickets had nothing to do with returns — shoppers wanted order status or exchanges. Rather than send them back to merchants, Seel resolved them itself. Peng's takeaway: AI's value in commerce "isn't saving money. It's in doing things instantly at scale even during peak volume," including approving eligible refunds before return shipments arrive.
Chatbot, agent, agency — what actually changes?
"The difference is who owns the outcome," Peng said. A chatbot answers "where's my order?" An agent completes a single task, like filing a claim. An agency owns the result: a customer who got what they paid for.
Technically, an agency works across the merchant's store, carrier, and payments rather than inside one system. It tracks a case over days, not one conversation, and knows when the outcome is done. Operationally, Peng said, you stop staffing for ticket volume and start staffing for exceptions.
His example: a package marked delivered that never arrived. A chatbot tells the shopper to wait. An agent opens a claim. Seel checks the carrier scan and order history, decides eligibility, pays out or ships a replacement, updates the merchant's records, and tells the shopper it's finished.
Who pays when an AI shopping agent buys the wrong size?
Peng's answer is blunt: the shopper shouldn't have to referee it, and the fix is the same regardless of fault — a fast exchange or refund. Three things make that possible:
- The order records who placed it and on what instruction, so the agent's decision is traceable
- The merchant's return policy is machine-readable, so agents can act on it
- Someone stands behind the outcome before fault is sorted out
That's Seel's position. "We take on the liability," Peng said. If the order was protected, the shopper is made whole first and attribution comes after — provided the right protection was added at checkout and the issue falls within coverage.
How Seel measures real service
Closure rate alone can hide bad service, Peng argued, because a bot that deflects shoppers looks great on a dashboard. So Seel tracks reopen rates, escalation rates to humans, and customer satisfaction. A shipping-address change that usually means an email chain across separate systems gets done by its agents in minutes.
The measure Peng cares about most is whether the shopper buys again — hence the 69% repeat-buyer protection figure.
'Pro-agent, just not pro-autopilot'
"Autopilot means nobody's watching," Peng said. Seel reserves three decision types for people: anything that sets policy, cases where a wrong call is expensive to unwind, and cases unusual enough that the rule's answer may not be the fair one. The AI can't expand its own authority; unauthorized or special-handling cases escalate straight to a person.
The system reads context — risk, urgency, customer frustration, policy sensitivity — and routes high-risk cases to people. Thresholds are configured per merchant and workflow, not as one universal dollar amount. Elevated-payout requests can fall outside the AI's autonomous authority.
Permissions, logs, and resale
Every agent gets only the access its job needs, scoped to workflows the merchant agreed to, via dedicated API tokens the merchant can suspend or revoke anytime. Large refunds can trigger human review or amount-based thresholds. Every action is logged — what happened, when, what triggered it — so Seel and the merchant can investigate, roll back, and tighten rules.
Seel's 1-Click Resale agency reviews a shopper's photo to check condition, prices against comparable sold items, creates the listing, and delists after sale. Descriptions start from the original product record; the shopper confirms the quote before the team lists.
The LLM Gateway economics
Seel built its LLM Gateway for itself first, then opened it after merchants reported the same cost creep, model by model. An LLM Router assesses request complexity and routes within the models a customer allows — pulling an order number from an email doesn't need the model a multi-step problem does. Customers pick performance, savings, or a balance. Billing is per model usage, not per completed outcome, and Seel covers the cost of complexity assessment.
What should merchants demand before handing over a function?
"A demo shows the system working once. A service shows it working on the ten-thousandth case at 2 a.m., when the carrier's API is down and the shopper is angry," Peng said. Merchants should ask for end-to-end resolution share, resolution times, repeat-purchase rates, who pays when it's wrong, and how many people it takes to run.
On trust, his Gen Z research is telling: 49% of Gen Z shoppers in Seel's 2026 report would rather resolve an issue with a person, and 87% of shoppers expect resolution within two to three days. "Trust comes from the exit, not the bot," Peng said — shoppers accept AI when they know they can reach a human, and trust collapses when AI becomes a wall, even if the answer was right.
What does this change for practitioners? The bar for AI support just moved from deflection stats to financial accountability — and vendors who won't stand behind outcomes are about to look very dated in merchant procurement calls.
Source: unite.ai
More from Nathan Brooks
Show full bio
News editor covering regulation, policy and their impact on entrepreneurship, startups and venture funding at Business Bearings.
2,426 articles