AI agents need humans in the loop

Companies like AT&T and Crocs are deploying AI agents, but they're keeping humans in control where it matters most.

By ·

Sep 18, 2026

11 cards · 2m · tap to begin

From · · · 2m

AI agents need humans in the loop

Companies like AT&T and Crocs are deploying AI agents, but they're keeping humans in control where it matters most.

In brief

Companies like AT&T and Crocs are deploying AI agents, but they're keeping humans in control where it matters most. AI automation works best when it handles routine tasks while humans stay involved in decisions that need empathy, judgment, or understanding customer intent. Originally reported by Adweek.

Salesforce's big AI bet

dreamforce entrance

Marc Benioff wants AI agents to replace human work, but only 30,000 of Salesforce's 150,000 customers use Agentforce, the company's agent feature launched in 2024.

The safety debate heats up

At Dreamforce, OpenAI's Sam Altman, Anthropic's Dario Amodei, and Nvidia's Jensen Huang debated whether AI development needs to slow down for safety research to catch up.

Microsoft challenges Anthropic

Microsoft AI chief Mustafa Suleyman criticized Anthropic's approach to training Claude on consciousness questions, arguing it could make AI systems harder for humans to control.

Amodei calls for caution

Amodei calls for caution

Anthropic CEO Dario Amodei published a 3,800-word essay arguing AI is advancing too fast for safety research, and called for independent evaluators to verify companies follow safety commitments.

Huang takes the opposite view

Nvidia's Jensen Huang argued AI safety is an engineering problem, not a legal one, and said companies should self-regulate rather than face new laws.

Trump rejects AI slowdown

President Donald Trump dismissed calls for slowing AI development, citing concerns about losing the race to China and arguing only a strong president can provide guardrails.

Consumers fear full automation

Consumers fear full automation

A survey of 2,606 consumers found 65% are comfortable with AI recommendations, but only 15% trust it with transactions and just 7.5% accept fully autonomous decisions.

AT&T speeds up phone upgrades

AT&T uses AI agents to cut phone upgrade time from 30 minutes to 10 minutes, but deliberately keeps humans in charge of cancellations to understand why customers are leaving.

“AI can make us much more efficient. But we really want to make sure that's still that human interaction.”

Southwest reserves empathy for humans

Southwest Airlines uses AI agents for routine tasks like flight changes but keeps human employees for what it calls right-here-right-now problems requiring empathy and assurance.

Crocs builds with guardrails

Crocs deployed AI agents for customer chat and internal tasks like product categorization, but maintains human oversight as it learns what automation can and cannot handle.

Data quality matters most

Crocs SVP Feliz Papich said the real challenge isn't building fancy AI interfaces, it's ensuring the underlying data is good enough to deliver what customers expect.

“Data is such an important conversation, respective to AI. Sometimes that gets lost because you're talking about the interface.”

The essence

AI automation works best when it handles routine tasks while humans stay involved in decisions that need empathy, judgment, or understanding customer intent.

Read the original on Adweek

React

Sign in to react and comment.

Comments (0)

Life is short. Keep it sweet. Respect others' opinions and be kind!

    Recommended next

    More decks on ai agents and related topics.