Who's Really at Your Front Door?
Plus: Medicare puts AI review on the clock, bots outnumber humans online, and Asmi hints at the next service curveball: the caller may be an assistant with a customer behind it.
Your daily signal on AI and CX — minus the hype.
DCX Stat of the day: Thales 2026 Bad Bot Report says bots made up 53% of observed web traffic in 2025. Bad bots alone accounted for 40%.
In this issue:
→ Medicare AI review gets an audit-trail test
→ Bot traffic outpaces human traffic
→ APIs become the new customer journey exposure
→ Customers start sending agents into voice channels
→ Brand trust moves from promise to proof
🔍 DEEP DIVE
What to Do When AI Touches Access
Medicare is about to run a real AI customer-experience test.
Not a chatbot. Not a shiny copilot. An AI-assisted review process sitting inside a journey where the customer consequence can be delay, denial, confusion, or care getting unstuck.
The Centers for Medicare & Medicaid Services’ WISeR (Wasteful and Inappropriate Service Reduction) model runs from through December 31, 2031 across New Jersey, Ohio, Oklahoma, Texas, Arizona, and Washington. The six participants are Cohere Health, Genzeon, Humata Health, Innovaccer, Virtix Health, and Zyter. CMS says they will use enhanced technology, including AI, to make medical-necessity recommendations, while licensed clinicians determine recommendations for non-payment.
That last sentence is doing a lot of work.
Because once AI enters an access decision, the experience includes the wait, the explanation, the appeal, the handoff, and the person who can fix the mess if the system gets it wrong.
So the operator question is not “Are we using AI?”
It is: can we show our work?
Any AI-assisted access journey needs four visible receipts:
Clock: When did the review start, pause, move, and finish?
Evidence: What information supported the recommendation?
Human check: Who reviews edge cases, errors, and disputes?
Recovery owner: Who fixes the customer impact if the process gets it wrong?
Without those receipts, the customer gets a black box with a longer hold time and better branding.
Bottom Line: If AI can affect access, eligibility, payment, service priority, or account action, the audit trail is part of the experience.
📬 Copy-Paste Take
Before we put AI into an access decision, we need to define the clock, evidence, human review step, and recovery owner. The customer should not have to investigate the system to understand what happened. If we cannot explain the decision path, we are not ready to automate the decision.
🧭 OPERATOR PLAYBOOK
Build the Decision Receipt
Pick one journey where AI can influence access, approval, eligibility, fraud review, payment, refund, escalation, or service priority.
Write the receipt the customer and the business would need if the decision is questioned:
What was decided?
What data was used?
What policy or rule applied?
What did AI recommend?
Who approved, reviewed, or overrode it?
What can the customer do next?
Then test whether a frontline employee can answer those questions without opening five systems or asking three teams.
Ask your team: Where do we have automated judgment without a customer-readable receipt?
Signal: The future of AI governance will feel a lot less like a policy deck and a lot more like service recovery.
📊 MARKET REALITY CHECK
Half the Web Is Already Automated
Thales’ 2026 Bad Bot Report says automated traffic accounted for 53% of observed web traffic in 2025. Bad bots made up 40%, up from 37% the prior year, while human traffic accounted for 47%.
The more useful finding is not volume. It is ambiguity.
Thales says AI agents are becoming a third category of automated traffic alongside good bots and bad bots. Some retrieve data or perform tasks for users. Others scrape, probe, bypass workflows, or attack APIs. The surface behavior can look similar because both legitimate and malicious automation may use normal-looking requests, browser environments, and application workflows.
For CX teams, this turns bot management into journey management. Search, pricing, login, checkout, booking, account recovery, and support APIs are no longer just technical endpoints. They are places where automated actors can distort availability, trigger fraud controls, inflate demand signals, create customer friction, or impersonate legitimate intent.
Why it matters: As customer journeys become more automated, “is this human?” becomes the wrong first question. The better question is: what is this actor trying to do, and should the business allow it?
Traffic identity + business intent = the new front-door control
🧰 TOOL WORTH KNOWING
Asmi
What it does: Asmi is an AI assistant that makes real phone calls for personal chores. It can call users, take tasks, use phone, WhatsApp, and iMessage context, place calls to businesses or service providers, and brief the user afterward.
CX use case: This is the customer-side version of voice automation. Instead of a brand deploying an AI agent to handle customers, customers can send an AI representative into the brand’s phone channel.
Worth watching because: Asmi puts an AI voice into ordinary service situations: restaurants, appointments, local services, scheduling, and follow-ups. The company’s own terms describe disclosure when Asmi calls on behalf of a user and rules around recording, consent, blocked call categories, and call timing.
That makes it useful for CX teams even if they never use Asmi directly. It is a preview of a near-term operating problem: how should frontline teams respond when the caller is an AI agent acting for a real customer?
Bottom line: Your phone experience needs an AI-caller policy before the first disclosed customer agent shows up in the queue.
Demos and Sign up for the Waitlist: Asmi
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
As AI becomes a shopping and discovery layer, marketers are making a practical point: brand trust, loyalty, first-party data, and consistent customer relationships become harder to fake.
The CX angle is the handoff. If an assistant mediates discovery, the brand still needs a relationship strong enough to survive the next step.
Why it matters: AI may change the interface, but customers still judge whether the brand knows them, helps them, and shows up consistently when the answer turns into action.
✅ YOUR MOVE
Today’s thread: customer intent is getting harder to read from the surface.
AI can review access. Bots can look like normal traffic. A customer’s assistant can call on their behalf. And the brand still gets judged for what happens next.
So run a 30-minute “automated customer” drill.
Pick one entry point: login, checkout, booking, claims, account recovery, appointment scheduling, or the phone queue.
Ask CX, operations, and security what happens when the actor is a good bot, a bad bot, or a customer-approved AI assistant.
Where does the system recognize the difference? Where does it guess? Where does the customer get stuck?
End with one decision: what should be allowed, limited, escalated, or logged differently this week.
The next customer may not arrive alone. Design the journey like you know that.
Until Monday,
👥 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.