One Prompt Can Become a Standing Order
Meta AI keeps working, Michaels turns project questions into carts, and Experian checks the agent at checkout.
Your daily signal on AI and CX — minus the hype.
DCX Stat of the day: Michaels says shoppers who use its Ask Mike assistant convert at more than twice the rate of shoppers using traditional search. Since May, 27% of Ask Mike interactions have led to a product click or add-to-cart. Modern Retail
In this issue:
→ One prompt can keep working across days
→ Project questions are turning into fuller carts
→ Good chats can still build a bad relationship
→ Experian gives shopping agents an identity check
→ Messy order requests become drafts for approval
🔍 DEEP DIVE
The Assistant Is Still Working After You Leave
Meta’s new AI model is built to do more than answer. It can connect to email and calendars, keep recurring tasks, prepare daily briefings, watch Marketplace for a requested item, and maintain plans that stretch across days.
That changes the customer journey. A prompt no longer has to end with the session. It can become an assignment that keeps gathering context, making recommendations, and preparing the next action.
For customers, that continuity is useful. For brands, it creates a harder question: what exactly did the customer authorize? Preferences change. Prices move. Inventory disappears. A request that made sense Monday may be wrong by Friday.
The operating unit is no longer the conversation. It is the living assignment: what the customer asked, what context the assistant used, what it may change, when it must check back, and how the customer can stop or correct the work.
Meta has not published adoption or outcome data for these new capabilities. The signal is the product direction. Customer-side AI is gaining memory, connections, and time.
Bottom Line: When an assistant keeps the thread, brands need to keep the authority record.
📬 Copy-Paste Take
A persistent assistant needs a visible job record: what the customer asked, what context it used, what it may change, when it must check back, and how the customer stops or corrects the work. Convenience without that record turns memory into ambiguity.
🧭 OPERATOR PLAYBOOK
Audit the Standing Order
Take one journey where an assistant remembers or acts after the customer leaves:
Assignment: Can the customer see the current goal in plain language?
Authority: Which actions may happen automatically, and which require approval?
Memory: Can the customer inspect and correct the preferences being used?
Interrupt: Can the customer pause, revoke, or recover the work without starting over?
Then change one important preference midstream. If the assistant cannot surface the conflict and ask again, it is carrying old intent into a new decision.
Ask your team: Where does remembered intent expire?
Signal: Persistence creates value only when correction stays easy.
📊 MARKET REALITY CHECK
A Good Chat Can Still Build a Bad Relationship
Stanford researchers count more than 140 state bills addressing AI in mental health contexts. Their larger warning applies to any company putting an AI assistant into an ongoing customer relationship: we know far more about how these systems perform in one session than what happens after repeated use.
Most CX dashboards measure one interaction at a time: resolution, containment, accuracy, or satisfaction. An assistant can score well on every chat while an incorrect memory shapes later recommendations, customer trust grows beyond the system’s competence, or human escalation arrives too late. A model update can also change the experience midway through the relationship.
Why it matters: If the AI remembers customers across visits, CX teams must measure the relationship across visits. Track whether remembered context stays accurate, customers can correct or revoke it, trust matches performance, and escalation works when the stakes rise.
Good session scores - unmeasured relationship effects = an incomplete CX picture
🧰 TOOL WORTH KNOWING
Experian Gives the Shopping Agent an ID Check
What it does: Experian Agent Trust connects a verified person, device, and AI agent. It checks identity, payment, intent, consent, risk, and behavior before returning a transaction decision.
CX use case: A shopping agent finds a product and tries to buy it. The merchant can distinguish an authorized customer agent from unknown automated traffic, verify the request, and step up authentication only when the risk calls for it.
Worth watching because: Agentic commerce breaks the familiar assumption that the person, browser, and buyer are the same actor. Fastly’s new participation brings those trust decisions closer to the edge, where APIs, authentication, payments, and security controls meet.
Bottom line: The smoother agentic checkout becomes, the more precisely brands must verify who authorized what.
The DCX AI Today - AI Tool Directory - If you lead a CX team and want a curated shortlist of tools worth evaluating, this is your starting point.
📡 90-SECOND CX RADAR
Proton turns messy requests into order drafts
Proton.ai’s new system converts customer emails, spreadsheets, PDFs, screenshots, and spoken requests into draft quotes or orders. It checks products, availability, and substitutions, then leaves the final send to a sales rep.
Why it matters: The useful automation is not just reading an inbound request. It is preserving customer intent while making the approval step faster and clearer.
✅ YOUR MOVE
Find one AI-driven journey where the customer’s intent persists after the session.
Write down what the assistant may remember, change, or buy. Then show the customer that same record. If your team cannot make the assignment visible, it is not ready to keep working in the background.
Memory without a correction path is just stale intent with permission.
Until tomorrow,
👥 Share This Issue
Think of one person who’s wrestling with AI in CX right now
and forward this to them.
I’m obsessed with Wispr Flow Pro! Get a Free Month on me.
If someone forwarded this to you, they thought you needed to see it before your next AI planning meeting. Get your own copy.