What “AI-First Support” Actually Means for Teams with Physical Products

The layoff announcements are piling up. Salesforce quietly cut thousands of customer support roles, laying off workers in two rounds and crediting AI efficiency gains. Zendesk is repositioning. Intercom is repositioning. Everybody’s repositioning. The message from the industry is clear: AI-first support is the future, and the future arrived faster than expected.

Fifty-six percent of major 2026 tech layoffs explicitly cite AI. That’s not a footnote — that’s a structural shift.

And yet. If you run support for a company that makes a physical product (hardware, appliances, industrial equipment, consumer electronics, medical devices), you’re watching this wave crash over your industry with a growing sense of unease. Because the AI-first support narrative? It was written by software companies, for software companies. And it has a blind spot the size of a shipping container.

What “AI-First Support” Actually Means in Practice

Let’s be precise about what’s actually happening at companies like Salesforce. They’re using AI to handle high-volume, text-based, structured requests. Password resets. Subscription changes. Billing disputes. Order status. Policy questions. These are tickets with clear intent, abundant training data, and outcomes that can be measured in closed-ticket counts.

AI is genuinely excellent at this. A well-trained agent can handle thousands of these interactions simultaneously, 24/7, with consistent quality. The ROI math is obvious. The headcount reduction is real.

When the industry says “AI-first support,” this is what they mean. It’s a reasonable strategy if your support queue looks like that.

Why the Definition Breaks for Physical Products

Here’s the problem: physical product support doesn’t look like that.

When a customer calls because their industrial router keeps dropping connections after a firmware update, they can’t fully articulate what’s happening. Or, when someone’s HVAC system throws an error code that doesn’t match anything in the manual, they describe it in five different ways, none of which match your knowledge base. Or, when a field technician is staring at a failed component and needs to determine whether it should be replaced or just reseated, the bottleneck isn’t language. It’s information.

Physical product support is fundamentally an information problem. And the information that matters: the device’s physical state, the environment it’s installed in, and the actual error on the panel. That information doesn’t exist in any ticketing system. It lives in the real world. In someone’s hands.

Text-based AI support agents are reading from a script in a language that nobody at the job site speaks. The agent doesn’t know what it doesn’t know, and neither does the customer, which is why they called.

What “Seeing” Enables That AI Text Agents Can’t

There’s a reason experienced support engineers ask to see photos. There’s a reason field service teams video-call before rolling a truck. Seeing changes everything.

Visual information collapses the diagnostic loop. Instead of five rounds of clarifying questions, each introducing new misunderstandings. You get a shared frame of reference. The customer shows the blinking light pattern. The support rep sees the cable orientation. The error state is visible. You’re now solving the actual problem instead of negotiating a description of it.

This isn’t a workflow optimization. It’s a fundamentally different kind of support interaction. Real-time visual support lets support teams see what customers see — eliminating the translation problem entirely.

AI can analyze visual data once it exists. Computer vision models are getting genuinely good at diagnosing hardware faults from images. But AI can’t ask a customer to show it something useful, walk them through capturing the right angle, or make a judgment call about what context matters when the customer doesn’t know what to show. That still requires human guidance, at least for now.

The companies that figure this out will have a real advantage. They’ll use AI-first support tools to handle everything that can be handled textually, and they’ll use visual collaboration to handle everything that can’t. That’s not a hybrid strategy. That’s what AI-first actually means for physical product teams.

The Headcount Math Doesn’t Work the Same Way

Here’s the uncomfortable truth nobody in the AI-first support conversation wants to say: the headcount-reduction math that works at Salesforce doesn’t translate directly to physical-product companies.

When you remove support reps who handle password resets and billing disputes, the remaining tickets are still solvable. When you reduce headcount at a company whose hardest tickets require visual diagnosis, you’re not eliminating work. You’re eliminating the people who could do it. The tickets pile up. Truck rolls increase. Customer satisfaction craters.

The pressure to go AI-first is real, and the efficiency argument is valid. But blindly applying a software company’s playbook to a hardware company’s support org will create a quality collapse that shows up in churn, not ticket volume. And by the time the churn is visible, you’ve already done the damage.

What a Real AI-First Strategy Looks Like for Physical Product Teams

If you lead support for a physical product company and you’re getting pressure to “go AI-first,” here’s a sharper frame:

Separate your ticket types ruthlessly. Which tickets are text-resolvable? Go hard on AI there: deflection, automation, tiered triage. Which tickets require visual diagnosis? Don’t confuse automation tools with solutions for those.

Invest in the visual layer. Your AI-first strategy for complex tickets isn’t fewer agents. It’s faster agents. Visual tools that let one experienced engineer handle the diagnostic workload of three, because they can see the problem in real time instead of spending thirty minutes negotiating a text description of it.

Measure what matters. Ticket deflection rate is the wrong primary metric for physical product support teams. Time-to-resolution, first-contact resolution, truck roll rate: these tell you whether your strategy is working.

Don’t let the AI-first narrative borrow your credibility. The companies cutting support headcount and calling it AI-first are mostly operating in environments where that’s viable. Yours may not be. That’s not a failure of ambition. It’s about being honest about what your customers actually need.

The AI-first support wave is real. For physical product teams, riding it well means understanding where it applies and building deliberately where it doesn’t.