When AI Starts Doing the Work, Ownership Gets Real Fast
Plus: AI is starting to do more than answer questions, and that is exactly where messy ownership turns into customer pain.
Your daily signal on AI and CX — minus the hype.
DCX Stat of the day: Physicians and staff spend an average of 13 hours each week completing prior authorization requests. That is a CX problem because the customer feels that admin drag as delay, uncertainty, and one more moment where trust can slip. Cognizant
In this issue:
→ AI is moving from talking to doing
→ Ownership gets fuzzy right where risk gets real
→ Brands may lose ground before checkout even starts
→ Better testing is becoming a CX requirement
→ Travel keeps showing where this goes next
🔎 Deep dive
Cognizant just showed where this gets real
Cognizant’s latest move matters because it takes AI out of the safe demo zone and drops it into a workflow people already hate. Its first live use case is prior authorization, where AI can check what is needed, gather the right documentation, and submit the request. That is not a nicer chatbot. That is AI stepping into real work that already creates delay, cost, and frustration.
The interesting part is not speed. It is responsibility. Once AI starts doing work inside the flow, the old questions about handoffs, permissions, approvals, and cleanup are not back-office details anymore. The customer feels them. If the AI gets stuck, oversteps, or sends the work into a dead end, the customer does not care that the workflow looked elegant on a slide.
The useful part is bigger than healthcare. Every company has journeys where the customer is really waiting on internal coordination, not on another message. Claims. Onboarding. Returns. Service approvals. That is where the work gets stuck. If AI is going to touch those moments, the better question is not “Where can we automate?” It is “Who owns the decision when the AI goes one step too far?”
📬 Copy-Paste Take
The next CX AI mess will not come from a bad answer. It will come from letting AI act inside a workflow nobody really owns.
OPERATOR PLAYBOOK
Check where AI is being allowed to do too much
Pick one customer journey where the customer is really waiting on internal coordination, like onboarding approval, return handling, claim intake, order exceptions, or service eligibility.
Audit every action step in that flow for four things:
What the AI is actually allowed to trigger, submit, change, or approve.
Where the rules still need to stay obvious to the team.
Where a human should step in before the customer feels the impact.
How you will trace the decision when the outcome is wrong, delayed, or challenged.
Then test whether the AI is being treated like a supervised helper or a vague new owner nobody wants to name.
Ask your team: Which customer-facing workflow would create the biggest mess if an AI agent acted one step too far?
Signal: If ownership gets fuzzy the moment AI crosses from answering into doing, you are not ready.
📈 Market Reality Check
Luxury brands may lose the experience before the customer reaches the brand
McKinsey’s point is especially sharp in luxury. A lot of the value is not just the product. It is the feeling of guidance, discretion, reassurance, and being understood. If AI starts shaping demand earlier, inside general assistants or third-party shopping tools, then part of that high-touch experience may get decided before the customer ever reaches the brand’s own site, store, or adviser.
That is where the CX risk shows up. Luxury brands have spent years trying to control pacing, tone, service quality, and the overall feel of the relationship. If AI becomes the first layer interpreting what the customer wants, the brand risks losing control of the very part of the experience that makes luxury feel different. And if that AI gets the nuance wrong, the customer does not blame the system. They blame the brand for making an expensive experience feel generic.
AI-led discovery + weaker brand control = premium experiences that start to feel ordinary.
🧰 Tool Worth Knowing
Cresta Synthetic Customers
What it does: Cresta turns your conversation history into realistic customer personas so teams can test AI, pressure-test messaging, and train agents against situations that sound more like real life.
CX use case: Useful when you want to test AI agents, policy changes, or frontline messaging before real customers run into the rough edges, especially in service and retention journeys.
Worth watching because: This gets at the real problem behind a lot of shaky AI launches. Most teams test the happy path, or they test against the way they imagine customers behave. Real customers interrupt, repeat themselves, get impatient, lose trust, and change direction halfway through. That is the stuff that breaks the experience.
Bottom line: If AI is going to play a bigger role in the journey, better testing becomes part of the CX job. Cresta is worth watching because it treats customer behavior as something to learn from before launch, not something to clean up after launch. No public performance metrics were shared beyond its reference to a blind evaluation methodology.
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
Gainsight launches an agentic stack aimed at retention, not just support
This is a useful reminder that the AI conversation is moving deeper into post-sale experience. Once AI starts reaching into onboarding, education, communities, and product usage signals, retention stops being a reporting metric and starts looking more like a daily operating system.
AI and Content Reshape Attractions & Tours at Trip.com Group’s Global Partner Forum
Travel is a good early signal because discovery, comparison, and booking already get messy fast. Trip.com is betting AI can tighten that path, which is probably where a lot of other experience-heavy sectors are heading too.
🧭 Your Move
The thread running through this issue is pretty simple: AI is starting to do more of the work, not just talk about the work. That sounds efficient until ownership gets fuzzy, testing gets shallow, and the customer ends up feeling the cost of a workflow nobody really thought through.
Pick one workflow this week where AI is likely to move from answering questions to actually doing something. Then get very clear on checkpoints and ownership before anybody starts celebrating speed.
The teams that handle this well will not be the ones with the most AI in market. They will be the ones that know where the AI stops, where a human steps in, and how the customer stays protected when the flow gets messy.
If AI is going to do part of the job, ownership has to be designed into the experience.
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.