← All retail signals
Consumer signal

The Deflection Trick

The support bot that keeps customers from a person is meeting resistance: a California law guarantees a route to a human, and complaints pushed some firms to rehire.

What is it?

Support bots are built to close a contact without a person, and when they fail the customer is left in a loop. A California law and a New York City proposal write a route to a human into the rules. Whether that route becomes the norm or stays a legal minimum is unsettled.

Why now?

Replacing support staff with bots was the obvious cost cut once language models got cheap enough to answer routine queries at scale. The cut outran the customer’s patience: complaints pushed some firms to rehire, and lawmakers moved to guarantee a human. How far the rules spread is open.

Who’s building it?

Vendors selling automation software; contact centres deploying it; lawmakers in California and New York City setting a floor for human access.

Support AI vendorsContact centresEnterprise buyersShoppersEscalation teams
The full read
Why now, who is building it, where the market goes, and what stays unsettled.
Get weekly →
01 · The market
Who’s in it
Builders — vendors selling automated resolution software to contact centres.
Buyers — enterprises that cut support staff for bots and then face the cost of rehiring.
Gatekeepers — shoppers whose complaints have pushed some firms back to human staff.
What’s at stake
Human access itself: where the law sets it, a route to a person stops being a company’s choice and becomes a customer’s right.
Which query tier stays automatable versus needing escalation.
Trust erodes fast once a shopper hits a closed loop.
What’s unsettled
Whether other states and cities follow California with a human-access rule.
How much of the harder dispute tier ever gets automated.
Whether a visible human path becomes standard, or stays a differentiator.
02 · How the demand matureda progression, not a timeline
Cost case
Bots replace support staff on the promise of handling routine contacts.
The reversal
Customer complaints have pushed some firms to rehire the staff they replaced.
Customer experience
Shoppers hit the gap directly — stuck in a loop the dashboard calls solved.
The rules
California guarantees a human within 15 minutes; New York City proposes a quick human off-ramp.
03 · Where the market goesthe forks, not a prediction
→Cut or rehire: firms could keep replacing support staff with bots, or follow the ones that brought people back after complaints.
→The harder dispute tier stays with humans permanently, or slowly gets automated too as models improve.
→A visible human escalation path becomes standard practice, not a premium feature vendors upsell.
→Firms keep pricing support on the bot’s savings, or start counting the cost of the customers it loses.
Everything here turns on access: whether reaching a person stays a customer’s right only where the law demands it, or becomes the standard everywhere.
04 · The Agentic Delegation deskadjacent markets, moving now
Checkout Moves to the Agent
Browser agents can complete checkout on Shopify merchants’ sites.
Shopify opens checkout to browser agents.
TechCrunch
Shopping Through the Agent
Shoppers take an agent’s pick seriously, yet few let it buy for them.
Few shoppers let agents make purchases.
Semafor
The Prompt Native
Chat interfaces are absorbing real shopping search volume.
ChatGPT processes about 50M shopping queries a day.
Digiday
05 · The Wire
01
New York City proposes a rule forcing a human off-ramp from chatbots
Semafor ↗
02
California requires large firms to offer a human agent option
CX Network ↗
03
Gartner projects half of firms that cut jobs for AI will rehire
The Washington Times ↗
The panel on this signal
They agreed on
–When a bot fails, the cost lands on the customer stuck in the loop, not on the dashboard.
–Complaints and rehiring show that cutting support staff for bots can be reversed.
–A visible human path is what actually preserves the shopper’s trust.
The behavioral economist disagreed on
–A rule that guarantees a human sets a floor, but the cost incentive to under-escalate stays; a firm can meet a 15-minute promise while still steering most contacts away from people, so the bias may be durable rather than a passing limit of the models.
Know more about the method →
✦How this signal was built
01 · Panel
A Delphi panel of 4 AI analysts, each reasoning independently.
→
02 · Debate
They challenge and revise each other’s reads over two rounds.
→
03 · Evidence
Every claim is tied to a cited, linked source on The Wire.
→
04 · Synthesis
A moderator writes it into one read, disagreement and all.