There is a specific sentence that tells you everything about an AI agent's real capability. It begins: “To start a return, please visit our Returns Centre…”
Only 4 of 15 live AI shopping agents could complete a return, cancellation or address change. All 15 could explain how to do it. That distance, between describing a task and finishing it, is the Answer→Act gap, and it is the widest performance divide in agentic CX today.
What does answer-only fallback look like?
Here is a conversation that happened, in some form, in ten of the fifteen deployments Alhena stress-tested in 2026.
Alhena's researchers logged this as answer-only fallback, and it was the single most common failure in the study.
How big is the Answer→Act gap?
| Capability | Deployments | What it means |
|---|---|---|
| Answer a question accurately | 15 / 15 | Commoditised |
| Navigate the shopper to the right product | 6 / 15 | Most name a product but won't route to it |
| Complete a real action | 4 / 15 | Where the field collapses |
On the eight-dimension scorecard, Agentic Capabilities averaged 1.6 out of 3.0: the lowest of all eight dimensions, against Context at a perfect 3.0.
Why do AI agents fall back to explaining?
Four architectural reasons, none of them fixable with prompting.
- No write access. Reading order status is a query. Cancelling an order is a mutation with financial consequence and real permissions to negotiate.
- No identity model. To act on someone's behalf, the agent must know who they are and be authorised to do it.
- No UI control. Even when the action is possible, the agent cannot move the storefront to complete it.
- No safe failure path. Teams that could not guarantee a clean rollback reasonably chose to disallow actions entirely.
Why is deflection the wrong metric?
Deflection asks whether a conversation avoided becoming a ticket. Resolution asks whether the customer's problem was solved.
An answer-only agent scores beautifully on the first and zero on the second. The ticket arrives twenty minutes later through a different channel, attributed to a different intent.
Three symptoms suggest you have this problem even when the dashboard looks healthy:
- Containment is high and ticket volume is flat.
- Repeat contacts within 24 hours cluster on task-shaped intents.
- Your most-contained intents are also your most-ticketed intents.
What does real resolution sound like?
“I've found it: the merino crew in charcoal, delivered on the 3rd. I've started the return and emailed your prepaid label. Refund lands 3–5 days after we receive it. Want me to swap it for a different size instead?”
Four things happened there: the order was identified, the mutation performed, the consequence confirmed, and a retention path opened. That last move is the commercial argument for agentic capability.
An agent that can act can sell during a support interaction. An agent that can only explain cannot.
Key takeaways
- Answer-only fallback appeared in 10 of 15 live deployments: the most common AI CX failure of 2026.
- Only 4 of 15 agents could complete a real action. All 15 could describe one.
- Agentic Capabilities scored 1.6 / 3.0 field-wide, the lowest of eight dimensions.
- The cause is architecture, not language. Write access, identity, UI control and rollback paths cannot be prompt-engineered.
- Deflection hides the failure. Measure resolution and 24-hour repeat contacts instead.
Frequently asked questions
Because most are built with read-only access to the store's order system. Explaining a cancellation only needs retrieval, while performing one needs write permissions and identity checks. Alhena found 4 of 15 live shopping agents could complete an action like this in its 2026 stress test. The other 11 produced accurate instructions and left the task with the customer.
Four things: authenticated shopper identity, write access to the order record, an integration that can trigger a label from your returns provider, and a rollback path when an action fails partway. Read access to your catalogue and help centre is not enough, which is why so many deployments describe the return rather than starting it.
Place a real order, then ask the agent to cancel it, change the delivery address and start a return. Count completed actions rather than accurate descriptions. The 2026 field average across 15 live deployments was roughly one in four, so an agent that fails this test is not unusual, only unfinished.
Authenticated shopper identity, an app with write scope on orders, handling for orders already in fulfilment, and a defined behaviour when the cancellation cannot proceed. The platform supports it. Whether your agent was architected to use it is the real question.
It is a scoping question rather than a yes or no. Limit write scope to the task types you have approved, require authenticated identity before any mutation, log every action, and insist on a defined rollback when a step fails partway. When you evaluate a vendor, ask them to demonstrate a failed action rather than a successful one: how the agent recovers from a wrong cancellation tells you more than a clean demo does.
Four of 15 live deployments, including Alhena, which scored 3.00 on the eight-dimension scorecard across every vertical tested. Agentic Capabilities averaged 1.6 out of 3.0 field-wide, the lowest of the eight dimensions and the clearest measure of the Answer to Act gap.
See which actions the field could complete
Every agentic task attempted across 11 verticals, plus the seven failure modes and how often each appeared.
