AI Agents Need Adult Supervision
Plus: Companies are sending AI into customer journeys faster than they can explain what happens next. Customers may have other plans.
Your daily signal on AI and CX — minus the hype.
DCX Stat of the day: Only 11% of technology leaders in a new IBM study said they are completely prepared for the expected scale of AI agent deployment. IBM
In this issue:
→ AI control is now a CX problem
→ Customer trust is lagging the rollout
→ Onboarding gets an AI nudge
→ Retention models get morally weird
→ Leaders need receipts, not optimism
🔎 Deep dive
AI control is where the customer story gets messy
The agents have left the lab.
They are not just writing emails, summarizing calls, or making meeting notes look a little less tragic. IBM says AI agents are moving into live enterprise work while the control model is still catching up. Two-thirds of surveyed CIOs and CTOs said they are accountable for AI systems they do not fully control. Seventy-seven percent said AI adoption is already moving faster than governance.
That is an operating problem with customer consequences.
For CX leaders, the danger shows up when AI starts touching things customers actually feel: refunds, eligibility, routing, identity checks, complaints, personalization, scheduling, and recovery. The customer does not care which team owns the model. They care that the decision made sense, someone can explain it, and the business can fix it before trust takes the hit.
Bottom Line: If AI can change a customer outcome, CX needs proof of what happened, who can stop it, and how the customer gets made whole when the system gets it wrong.
→ Get the 2026 Tech Leader Study
📬 Copy-Paste Take
AI agents are not a CX win just because they move faster. They are a CX win when the business can still explain, correct, and own the customer impact.
OPERATOR PLAYBOOK
Put a leash on the customer-visible decisions
Pick one journey where AI can affect time, money, access, trust, or momentum.
Audit every AI-assisted control point for four things:
What customer outcome the agent can change.
What policy, data, model, and permission it depends on.
What signal shows the agent is drifting or overreaching.
Who can pause, correct, explain, or reverse the action.
Then run the courtroom test. If a customer challenged the decision tomorrow, could your team explain what happened without pulling four departments into a forensic group chat?
Ask your team: Where are we letting AI move faster than our ability to explain the customer impact?
Signal: Control is real when it helps the business recover the customer while recovery still matters.
📈 Market Reality Check
The trust gap is already on the scoreboard
Diginomica covered a useful trust split from MIT’s Renee Gosline at Salesforce Connections: 81% of professionals trust agentic AI with customer data, while only 36% of customers agree.
That is a big gap. Big enough that the keynote version of the story and the customer version may not be living in the same zip code.
The macro issue is simple. Internal confidence does not automatically become customer trust. Marketers, service teams, product leaders, and executives may be ready to let AI recommend, rank, personalize, and trigger more of the journey. Customers still want control, consent, and a way to challenge the machine when it reaches too far.
Why it matters: If companies scale agentic AI faster than customers can understand it, the efficiency gain may come with trust debt. And trust debt eventually shows up somewhere in the journey.
Internal confidence - customer confidence = trust debt.
🧰 Tool Worth Knowing
OnRamp Aero
What it does: OnRamp Aero brings agentic AI into customer onboarding. It monitors active projects, spots stalled accounts, drafts context-aware follow-up, guides customers in the portal, and helps ops teams turn playbooks into live workflows.
CX use case: This is for the awkward post-sale zone where the customer said yes, everyone celebrated, and then the actual setup starts moving slower than the sales deck promised.
Worth watching because: Onboarding is where trust quietly gets won or wasted. Aero treats follow-through as a managed journey, not a checklist that depends on someone remembering to chase people before the customer goes quiet.
Bottom Line: The useful test is whether Aero helps teams catch stalled momentum, assign the right owner, and keep customers moving toward value without making the experience feel abandoned to automation.
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
AI retention systems are starting to test the ethics of customer value
CX Today takes the idea into near-future territory, but the operating issue is real. Churn models, lifetime value scoring, and retention prioritization already exist. AI just makes them faster, colder, and easier to scale.
Why it matters: Leaders need a policy for how customers are treated when the model says they are no longer worth saving. That is not just analytics. That is brand behavior.
PegaWorld puts human-guided AI into the customer engagement debate
The useful signal is the frame. Customer engagement teams are now debating where human judgment belongs once automation can select, rank, and trigger the next customer action.
Why it matters: The next CX advantage may come from knowing which decisions should stay human, which can be automated, and which need both in the room.
🧭 Your Move
Today’s issue is about AI ambition getting ahead of customer confidence.
The pattern is showing up in governance, marketing trust, onboarding, and retention decisions.
Take one journey where AI is being added and mark the customer-visible decision points.
Approve. Deny. Recommend. Rank. Route. Refund. Retain. Escalate. Explain.
Then ask the plain version of the governance question: if the customer challenged that decision tomorrow, could your team show what happened without sounding like everyone is hiding behind the model?
AI does not earn trust by acting faster. It earns trust when the business can still own what happened.
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.