AI customer service: answering well is not resolving

What AI customer service means now

AI customer service refers to software that uses machine learning and large language models to handle support interactions. In e-commerce it currently shows up in three forms: conversational agents that answer customer questions directly, copilots that draft replies for human agents, and agents that take actions in connected systems rather than just producing text.

The first two are widely deployed. The third is where the operational value is, and it is a different problem from making a chatbot sound helpful.

Every order is a promise. Keeyu keeps it. An AI that apologizes fluently for a broken promise has not kept it.

The limit of answering well

A conversational agent handling "where is my order" can produce an excellent reply in two seconds. It cannot make the order arrive. If the order never synced to the warehouse, the best possible answer is still a well-worded description of a problem the customer now has.

Brands notice the difference. In a LinkedIn post in February I wrote: "Brands are tired of reactive. Tired of tickets piling up. Tired of AI chatbots apologising for problems that should never have happened."

A helpdesk is designed to wait for a complaint, and the AI bolted onto one has the same architectural problem: it is not connected to the tools that manage an order after the buy button. I called a company recently and the AI that answered said it could not fix my problem and to wait a minute. The human never came and I had to leave a voicemail. On The 9-5 Exit Plan I called this AI reactive: it is still waiting for the customer to have a bad experience.

The agent-side version of that speed exists too. Our global search finds any order by order number, ID, tracking number, customer name, or email, and shows the status instantly, like 'awaiting collection'. Useful, and still downstream of the real fix.

At Papinelle help desk tickets ran at 110% of order volume, and roughly half of them were where is my order. Answer automation applied to the where is my order half would have made the failure cheaper to service and left the failure in place.

Average reply time and tickets closed per hour keep a team running faster on the treadmill instead of asking why it is running. The question worth asking of any AI support tool is simple: does it reduce the number of customers who have a bad experience, or only the cost of talking to them afterward?

The category line I use is read versus write. Answering tools read your knowledge base and your order status and compose a reply, so their ceiling is the quality of the reply. On The Ecommerce Edge I drew the same line for tools that only notify: they have read capability only. A system of action connects through deep APIs and can write back, so when the shopper replies we can cancel the order or send the replacement. An answer is a description of a problem. An action is the end of one.

What an operational AI agent does differently

An agent that works on operations rather than conversations watches the systems that hold the truth about an order, the store, the warehouse, the carrier, and acts when something breaks. Detect the stalled order, decide the remedy, execute it.

Concretely: an item still in transit well past its delivery date is functionally lost. The workflow cancels the order, raises a new one, allocates express shipping, pushes it to the warehouse and notifies the customer. No ticket is created because there is nothing left to ask about.

Out-of-stock runs the same way. The workflow triggers when a warehouse types 'out of stock' into NetSuite, searches other stores for inventory, finds product alternatives, checks the ERP for next shipments, and if nothing can replace the item, cancels the order and processes the refund.

We can identify 45 lost-in-transit orders in real time, select them, and resolve them in parallel against a predefined workflow. At EHP Labs an out-of-stock workflow that took 45 minutes now runs in five, reactive tickets fell 55%, and the CX team went from 18 people to 8.

Here is a real sequence rather than a capability list. An order stops moving between scans for long enough that it is almost certainly lost in transit, so the agent finds alternative stock, places the replacement, upgrades it to express, sends the shopper the new tracking number with an apology that arrives before their complaint does, then opens the claim with the carrier. I walked through it on Add To Cart. The shopper's part of that is reading one email.

Why fully autonomous is the wrong target

I have seen both extremes fail: teams trying to scale with people alone, and teams chasing "set and forget" automation. Both break trust fast. These workflows cancel orders, issue refunds and raise replacement shipments against real customers and real money. Agentic AI at this point cannot think autonomously and just run free. It requires a train track: a workflow with a defined start and finish.

What works is the loop: agents for foresight, workflows for execution, humans for judgment. A brand can set a workflow to trigger automatically, or keep a user in the loop where specific decisions have to be made. That person can see every system the agent went to, and when it cannot solve a problem it loops them in and says it is stuck.

Unsupervised agents need minding in my own work too. In a LinkedIn post in April I wrote: "My Claude Code and Moltbot Agents are keeping me so busy, it's starting to feel like I should hire a team of five to manage them ... The agents were supposed to save me time. Instead, I'm babysitting them."

There is also an integration reality. We lost two customers because we could not integrate with their systems, so our CTO Tahir built browser operator functionality that lets Keeyu work through the browser and read data from systems with no usable API. Real operations run on software that was never designed to be automated. We scored the main ecommerce customer service platforms, Keeyu included, against that reality in our ranking of ecommerce customer service software.

Where a carrier or 3PL offers no usable API, our AI runs a browser operator: it opens the browser, reviews the information, takes photographs, and converts them into usable data. We use it to constantly scan carrier portals like Royal Mail, and for a 3PL in Sydney notorious for giving nobody API access to their platform.

How to evaluate an AI customer service tool

• Can it take actions in your store, warehouse and carrier systems, or only produce text?
• Does it detect problems on its own, or only respond to a customer message?
• Where is the boundary between automatic execution and human approval?
• What happens with systems that have no API?
• Does it report tickets deflected, or operational breaks prevented?

I would add a sixth test as buying moves into AI shopping assistants. When the transaction happens inside an AI agent, your brand becomes invisible, and anything that goes wrong lands inside the shopper's personal assistant, not on your storefront. Can the tool signal that buying agent when a delay occurs?

Where this sits

This is proactive e-commerce operations, a system of action rather than another place to manage conversations. Detect. Decide. Act. The customer gets what they want, on time, as promised.

See how the workflows run or book a demo.

It sits above the answering layer, not inside it. We have watched a team of eighteen firefighting all day become eight running operations and ten reaching out to customers, which I described on The Ecommerce Edge. The AI did not replace the team. It replaced the exports, the system checks and the chasing that filled their day.

Related reading

For the automation distinction, read customer service automation. For the tooling layer, see helpdesk. For the posture shift, read proactive versus reactive customer service.

• AI customer service agent: what it is allowed to do
• AI in customer service: why inbox gains plateau
• Post-purchase operations, explained
• WISMO: the ticket that tells you a promise broke
• Ecommerce fulfillment: in-house, 3PL, and the gap nobody prices in

• Gartner, 2024: 14% of customer service issues are fully resolved in self-service.
• Gartner, 2025: more than half of service journeys now start on third-party platforms.
• Harvard Business Review, 2010: reducing customer effort predicts loyalty better than delight.
• The Breakout CEO #86, The Pivot This Founder Made After an Investor Called It Impossible - one in five US shoppers not getting orders on time, and the ticket ratios behind it.
• Marketing for SMEs, Jevon Le Roux on the E-Commerce Mistake Costing Millions - the orchestration layer, and the 55% ticket cut at EHP Labs.
• The 9-5 Exit Plan, How AI Is Making Customer Support Invisible - proactive AI and support that disappears.