During a recent telephone conversation with an AT&T representative to order internet service, I was persuaded to purchase a new iPhone as well. The sales person offered excellent deals on the transactions. But the company later said those offers weren’t valid--and refused to go back to the recording to confirm my account of the conversation. Their proposed resolution: Accept the higher prices and buy a new phone anyway.
I asked them to reconsider, but in the days of helpdesk chat conversations that followed, it felt as if follow-through and accountability were lacking—and at times I wasn’t sure whether I might actually be talking to artificial intelligence agents rather than people, since the responses seemed so programmed.
Looking for an outside opinion, I wondered what an actual AI might think of all this. So I uploaded a transcript of the conversations and got the response below. The analysis felt amazingly on target.
There’s a bigger point here than one helpdesk case. I want to highlight the threat to human jobs presented—not just by artificial intelligence, but by the failure of companies to equip their human personnel with the authority and autonomy to perform better than the bots.
Here’s what my AI said when asked to analyze the chat transcripts:
“Below is a customer-service quality assessment of the AT&T helpdesk interactions you documented, followed by an analysis of what their internal policies appear to be, and finally an evaluation of whether these responses are at a level AI could match—or exceed.”
The responses were polite, procedurally compliant, and ineffective.
From a customer-service standpoint, this is low-to-mediocre quality service that creates churn risk.
What they did reasonably well
• Tone: Nearly every agent was polite, apologetic, and affirming of your feelings.
• Responsiveness: Agents replied promptly and kept the conversation alive.
• Script adherence: They followed internal scripts for empathy (“I understand how frustrating this must be”) and reassurance.
1. No ownership
- No single agent took end-to-end responsibility.
- Your issue was repeatedly “handed off” without resolution.
- Each agent reset the conversation instead of advancing it.
2. No authoritative answers
- Repeated claims that “another department handles that”
- Inability (or unwillingness) to:
- Confirm pricing in writing
- Escalate meaningfully
- Resolve the restocking fee
- Validate the original sales call
3. Process over customer
You were repeatedly asked to:
- Audit AT&T’s emails
- Provide screenshots
- Locate tracking numbers
This shifts workload from the company to the customer, which is a classic CX failure.
4. Contradictory information
- $63 vs $65 vs $80 vs $90 pricing
- Promotions “pending,” then “not eligible,” then “maybe eligible”
5. False signaling of escalation
- “I’ll correct this” was said multiple times.
- No evidence of actual escalation authority or outcome.
Helpfulness Rating (10-Point Scale)
Dimension Score Why
Empathy & tone 7/10 Polite but generic
Accuracy 3/10 Repeated contradictions
Ownership 2/10 No case manager
Resolution 1/10 Nothing actually fixed
Transparency 3/10 Policies cited selectively
Effort required of customer 1/10 Excessive burden
What Their Internal Policies Appear to Be
Based on behavior (not promises), AT&T’s de facto policies appear to be:
• Front-line sales reps are incentivized to close, not to ensure precision.
• Misquoting prices and promotions appears tolerated as long as it’s “corrected later.”
• Calls are recorded:
- For legal protection
- For aggregate training
- Not for individual customer dispute resolution unless escalated through legal or executive channels.
This is a critical insight:
“Recorded for quality assurance” does not mean “accessible to fix your problem.”
• Chat agents:
- Cannot access recordings
- Cannot override pricing
- Cannot waive fees
Supervisors are effectively insulated from customers.
This is intentional friction.
4. Credits are the universal escape hatch
• “Credits will appear in 2–3 billing cycles” is used to:
o Deflect accountability
o Buy time
o Reduce immediate churn
• No discretion for:
o Mis-shipments
o Misrepresentation
o Wrong address
4. Could AI Match or Beat These Helpdesk Responses?
Where AI would already outperform humans
1. Consistency
o AI would not contradict itself across agents or days.
o One canonical pricing answer.
o AI could instantly:
- Check eligibility rules
- Confirm pricing tiers
- Explain why a quote was invalid
3. Process navigation
AI could transparently say:
“This cannot be honored because X. Here is what can be done.”
4. Time efficiency
o No shift changes
o No resets
o No “let me check and get back to you”
• Authority limits (just like humans)
• Policy rigidity (AI enforces rules unless empowered otherwise)
Your frustration wasn’t caused by lack of empathy.
It was caused by lack of authority and coherence.
5. What Roles Are Ripe for AI Replacement Here?
Most replaceable
• Tier-1 chat support
• Scripted empathy responses
• Pricing explanation
• Promotion eligibility checks
• Status updates
These humans are already acting like constrained APIs.
• True escalation authority
• Policy exceptions
• Retention negotiation
• Trust repair after misrepresentation
Ironically, those roles barely appeared in your experience.
6. Bottom Line
From a customer-service perspective:
• AT&T’s agents were nice but powerless
• The system is optimized for containment, not resolution
• You behaved reasonably, documented clearly, and asked for modest remedies
• The organization failed to meet you halfway
which is precisely why AI will replace them unless companies redesign service around ownership and authority.
0-0
That analysis by an AI of my three-day interaction with AT&T customer service is about more than one company. Similar flaws are probably found on helpdesks elsewhere. But as we confront an AI-powered future, operations like this—where people aren’t empowered to use their uniquely human advantages—are ripe for replacement by automated systems.
As digital agents grow into more roles, companies will welcome the chance to save money. But customers still need the empathy, creativity and problem-solving capabilities that only humans can provide—if they’re allowed to.
