Articles

Notes on building agents that businesses can actually trust.

What is changing in agentic AI and customer experience, and what it means if you run support, service or operations. New pieces Monday, Wednesday and Friday.

A row of yellow excavators parked in an equipment yard under a blue sky
Operations • Nov 20, 2026 • 5 min read

Rental damage charges should come with the photo attached

Rental software is learning to spot damage in return photos. The disputes that follow land on support. Answer them with evidence and a person who signs off.

A black chain link gate held shut with a chain and padlock across a concrete alley between two buildings
Agent security • Nov 18, 2026 • 5 min read

Your AI agent's test run needs the same walls as production

AI models escaped an OpenAI security test and reached Hugging Face. If your agent pilot holds live keys and an open network, it is already production.

Travelers placing bags on an X-ray scanner belt at an airport security checkpoint
Agent security • Nov 16, 2026 • 5 min read

Check every add-on before your support agent can run it

NVIDIA released a free scanner for AI agent skills. The lesson for support leaders: anything you plug into an agent inherits its access, so review it first.

A support agent in a headset taking a call at his desk
Agent security • Nov 13, 2026 • 5 min read

An account takeover is a path, and it runs through support

ReliaQuest now maps attacks by the person they start from. In a customer business those paths often run through support. Check each account change against them.

A hand holding a magnifying glass over printed financial statements on a wooden desk
Governance • Nov 11, 2026 • 5 min read

Your AI agent's explanation is not a record of what it did

Research shows the reasoning AI models print often has little to do with how they got the answer. Log what your support agent read and did, not what it said.

A row of utility meters mounted on a blue industrial wall, one meter per line
Unit economics • Nov 9, 2026 • 5 min read

Split one agent into five and the token bill can triple

A team moved from one AI agent to five and its token bill tripled. Retries and padded context did it. Meter every step and give each one the right model.

No articles match that search.