Automated customer contact is software that handles customer questions without a person writing each reply. It ranges from auto-replies and help centre articles to chatbots, AI-drafted replies your team approves, and AI agents that resolve tickets end to end. Good automation answers from live order data, knows its limits and hands over to a person with full context.
This guide is for an e-commerce operator deciding how far to go. It sits under our guide to AI agents and covers the levels, the contacts worth automating, hand-offs, channels and risks.
What is automated customer contact?
Also called automated support or automated customer service, it covers any part of your conversations that software handles instead of a person, from a one-line acknowledgement to an AI agent that reads an order and issues a refund. For the wider topic, see AI support.
Salesforce frames it as technology doing routine service work with no human directly involved. For an online store we use a sharper test: automation only counts when it closes the question. That gives four useful verbs: it acknowledges, it informs, it drafts or it resolves.
Most store contact is order status, returns, refunds, address changes and product questions. That is why the gap between “answers” and “resolves” matters more in e-commerce than in a generic contact centre.
What are the levels of automated support?
There are five levels, from a message that only says “we got it” to an agent that reads the order, takes the action and closes the ticket. Each level up hands the customer more of an outcome, and needs more access to your data and more trust in your policies.
| Level | What it does | Who sends the reply | Reads live order data? | Can take an action? | Store example |
|---|---|---|---|---|---|
| 1. Auto-reply | Acknowledges and sets expectations, resolves nothing | The system | No | No | ”We received your message, we reply within 24 hours” |
| 2. Self-service and FAQ | The customer finds a static answer | The customer helps themselves | Usually no account data | No | A returns policy page |
| 3. Rule-based chatbot | Menus and keyword matching | The bot, with scripted text | Limited | Rarely | ”Pick 1 for returns” |
| 4. Agent Assist drafts | AI drafts the reply and suggests actions in the help desk | A person reviews and sends | Yes | Suggested only, a person confirms | A drafted WISMO reply with the live tracking status |
| 5. Autonomous agent | Reads the order, applies your policy, acts and resolves | The agent | Yes | Yes, within your policy | Cancels an unfulfilled order and confirms the refund |
Levels one to three deflect: they try to stop the question reaching your team. Levels four and five resolve. Level three is where a scripted chatbot stops: it sees little of the order and cannot act on it.
For bots and agents side by side, see the comparison.
Which customer contacts should you automate, and which should stay human?
Automate the contacts where the answer lives in your data and a clear policy covers the action. Keep a person on contacts where judgement, empathy or money beyond your policy is at stake. Decide per contact type, not for the whole inbox at once.
| Contact type | Automate? | Why | Where a person comes in |
|---|---|---|---|
| Where is my order (WISMO) | Yes, end to end | The answer is in order and carrier data | The parcel is lost, or the customer is upset |
| Returns within policy | Yes | The policy is clear | Refunds above a value you set, or outside the return window |
| Address change or cancellation before fulfilment | Yes | A single action in the store | After the order has shipped |
| Product advice and sizing before purchase | Yes | Answers come from catalogue, stock and storefront content | Complex custom or bulk orders |
| Damaged or missing items | Yes, with guardrails | Policy plus a photo settles most cases | Repeat claims or high-value items |
| Complaints, legal threats, chargebacks, suspected fraud | No, keep human | Judgement and liability | Always |
| VIP or emotionally charged conversations | Keep human or Agent Assist | Tone matters more than speed | The person sends, the AI drafts |
Results follow this pattern. Otrium’s team was busy with repetitive questions about returns, refunds and “Where is my order?”, and its Otrium case shows how that changed. HelloPrint automated 80% of its tickets. The benefits come from resolving contacts, not just deflecting them. Our automation guide walks through the steps in order.
How does automation hand over to a person without losing context?
A good hand-off passes the conversation, the order it concerns, what the automation already checked or did, and why it stopped. The customer never repeats themselves, and your agent starts at the answer instead of the question.
When the AI cannot solve the problem, the chatbot should say it is connecting the customer to a person, and the chat should continue from where the bot stopped. Pass on five things:
- the transcript so far
- the identified customer and order
- the data already looked up, such as tracking status or return eligibility
- the actions taken or proposed
- the reason for the hand-off
The MR MARVIS agent works this way: it handles routine questions and prepares more complex cases for human agents, with the relevant information already collected. At Otrium, when the agent cannot solve an issue it hands it to a person, so nothing falls through the cracks.
July 2026This has made me rely on ‘agent handover’ sections with agent instructions more than initially, i.e, focusing less on mapping all the troubleshooting steps, and more on ‘gathering valuable context’ the agent can leverage, while letting them handle the complexities based on their wisdom. · Zendesk CommunityWith Engaige, hand-offs go back to your team inside the help desk you already run, and our ticketing guide covers how routing and tagging work around them. Once autonomous, Engaige AI also learns from the tickets the agent hands back.
Which contact channels automate well?
Text channels automate best: email, live chat, WhatsApp, SMS and social DMs. Text gives an AI time to read the order, check the policy and act before it answers. A live phone call gives it seconds, so resolution quality on voice is lower.
| Channel | Automates well? | Note |
|---|---|---|
| Yes | Asynchronous, so there is time to check data before replying | |
| Live chat | Yes | Short wait is accepted, a lookup fits inside it |
| WhatsApp, SMS, social DMs | Yes | Same text logic as chat |
| Phone | Harder | Real time, with seconds to reason |
Engaige is text-first by design and has no voice AI. If phone is a big share of your mix, a voice specialist is the honest answer, and our guide to voice AI covers that side.
What are the risks of automating customer contact?
Three failures make automation worse than none: wrong answers, dead ends and no way out to a person. All three are design choices rather than fate, so you can prevent them with live order data, clear policies and a guaranteed hand-off to a person. Transparency is a separate, legal point, covered below.
Wrong answers. A tribunal has already ruled that a company owns what its chatbot says; our knowledge chatbots guide has the case. For a store, the lesson is simple: what your bot says, your brand says.
Dead ends. When an issue falls outside a chatbot’s limited capabilities, customers get stuck in repetitive answers with no way through to a person. Our store chatbots guide covers why scripted bots do this.
No way out. A common worry about AI in service is that it gets harder to reach a person. Our reading: customers reject dead-end bots, not resolution.
On transparency, Article 50(1) of the AI Act requires providers to design AI systems that interact with people so those people are informed they are talking to an AI, unless that is obvious. It applies from 2 August 2026. Whether the duty also reaches your store is a question for counsel. Not legal advice.
The wider questions are covered in team replacement, ROI, customer experience and, for choosing a vendor, the vendor guide.
Where does Engaige fit?
Engaige (that’s us) isn’t a chatbot or an auto-responder: it’s the e-commerce-native AI support agent that automates customer contact in text channels, working inside the help desk you already run. Keep your helpdesk. Replace the AI. Engaige AI learns from your conversations and inputs, configures your support agent and shows which policies it used. Otrium resolves 60% of its 120,000 annual tickets end to end with Engaige.
You choose the start: Agent Assist, where nothing closes without a human, or autonomous from day one. Ticket types then switch to autonomous one at a time, and brands that move quickly get there in about a week. Across ticket types, autonomous can cover up to 90% of supported tickets. The go-live path is in our AI agents guide.
Policies give you control over how the agent reasons, what process it follows and when it may act. What it does alone, and where a person checks first, is a decision you make per step.
Scenario
The customer asks about a delivery address.
Instructions
Read the order with Get order , then take the branch that matches.
Get order Scenario
The order has not been picked yet.
Instructions
Move it with Change picqer shipping address .
Change picqer shipping address Scenario
The parcel has already left the warehouse.
Instructions
Say it cannot be redirected, then offer Process refund .
Control is also what keeps the experience human. Otrium chose Engaige for exactly that.
Engaige offered control, flexibility, and the ability to really incorporate AI in a more human way.
In Agent Assist, Engaige AI learns from the difference between what your team sends and what the agent suggested, and you approve every improvement. The full loop and the Shopify depth are in our AI agents guide. To see it on your own tickets, book a demo.
Frequently asked questions
What is automated customer contact?
It is software that handles customer questions without a person writing each reply. It ranges from auto-replies and help centre articles to chatbots, AI drafts your team approves and agents that resolve tickets end to end.
What is the difference between automated support and an AI agent?
Automated support covers all five levels. Auto-replies, self-service and rule-based chatbots mostly deflect. An AI agent sits at the top: it reads the order, applies your policy, takes the action and resolves the ticket.
Which customer questions should an online store automate first?
Start with where is my order, returns within policy, address changes before fulfilment and product advice. The answer sits in your data and a clear policy covers the action.
Which customer contacts should stay with a person?
Complaints, legal threats, chargebacks and suspected fraud need judgement and carry liability. VIP and emotionally charged conversations are better kept human, or handled with AI drafts a person sends.
How do customers reach a human when contact is automated?
Through a guaranteed hand-off that carries the transcript, the order, what was checked and why the automation stopped. The conversation should continue from where the bot left off.
Can automated customer contact work on the phone?
Text automates best, because the AI has time to check data. Engaige is text-first, no voice AI.
Do I have to let AI answer customers on its own from day one?
No. With Engaige you can start in Agent Assist or go autonomous from day one, then switch ticket types over one at a time.
Do customers have to be told they are talking to AI?
Article 50(1) of the EU AI Act requires providers to design AI systems that interact with people so those people are informed they are talking to an AI, unless that is obvious. It applies from 2 August 2026. Whether that reaches your store is a question for counsel. Not legal advice.
