The Post-Sales Assistant That Sells
Subscription E-commerce · 2026 · Customer Support Automation
Subscription changes now happen inside the conversation, following the rules the CEO wrote — and an operator can take over at any point.
The challenge
A subscription e-commerce company ships consumer goods on a recurring plan. The sale is the easy part. What comes after it is a steady stream of small requests: move this delivery, ship it earlier, change my frequency, pause me for a month, start me again.
Each one meant a person opening the back office and editing a live subscription by hand. The requests are few in kind and many in number — the worst possible shape for manual work.
There was a second cost, less visible than the first. Every one of those messages is a customer explaining how they actually consume the product. "Skip my next box, I still have three" is not an admin request; it is a plan that no longer fits the customer. Handled by hand, under time pressure, it ends as a date change and nothing else.
What we built
An assistant on the company's post-sales channel, running in the client's own cloud account and connected to the subscription system. It takes five actions on live subscriptions:
- move the delivery date
- move the shipping date
- change the delivery frequency
- pause a subscription, under human supervision
- reactivate a paused subscription
It does not buy. New purchases are routed to the purchase page, where checkout belongs. The perimeter is deliberate: the assistant edits plans that already exist, it does not open new commercial commitments on a customer's behalf.
The commercial rules come from the CEO. When a customer asks to postpone a delivery, the assistant asks why. If the answer is "I still have stock", it proposes the plan that matches the real consumption rate — three boxes every three months instead of one box every month. Fewer shipments for the company, fewer interruptions for the customer, a larger order. That rule is the company's commercial policy, written by the person who owns it, executed the same way in every conversation. When the policy changes, the rule changes with it.
An operator can take over at any time. They step into a live conversation, read everything that was said, and continue it as themselves. And some conversations are sent to a person on purpose: logistics — a missing parcel, a courier dispute, a delivery that went wrong — is handed to a human by design, with the order and subscription context already attached. That work needs someone who can call a carrier and make a commitment; it is not a gap in the assistant.
The results
Around 70% of conversations are closed end-to-end, with no operator taking over. The remaining share is not failure: it is the routing the system was built to do — logistics to a person, judgment calls to a person, everything else handled in the chat.
The figure is an order-of-magnitude estimate from the team running the system, which is why it carries a tilde. The action count and the escalation rule are not estimates — they are how the system is built.
| Number | What it measures |
|---|---|
| ~70% | of conversations closed without an operator taking over (estimate) |
| 5 | actions taken on live subscriptions: delivery date, shipping date, frequency, pause, restart |
| By design | logistics conversations go to a person |
| 0 | new purchases executed by the assistant — orders go to the purchase page |
What you get
The assistant on your post-sales channel, wired to your subscription system, with the action set and the commercial rules you decide — plus the operator takeover console and the dashboard showing what was handled and what was escalated. Process audit first, scope in writing, and 30 days of support. It runs in your own cloud account. The wider package is Customer Support Automation.
Answering the same five subscription questions by hand? Start with the free AI Audit, or book a call.