The Honest Answer

AI will replace some support work. It already has. But the "AI is coming for all support jobs" take is lazy and wrong.

Here's what's actually happening.

The Work That's Already Gone

If your job is answering "how do I reset my password" eight times a day, that work is going away. It should have gone away years ago, honestly. It didn't need a human in the first place — it needed a self-service flow or an automated response.

The categories of support work that AI handles well right now:

• Password resets and account access. Fully automatable today.
• Order status inquiries. "Where's my package?" can be answered by pulling tracking data. No human judgment needed.
• Pricing and plan questions. Send the pricing page or a brief summary. Done.
• Basic how-to questions. If the answer is in your docs, AI can find it or classify the question and respond with the right link.
• Refund requests for simple cases. If the policy is clear (under $50, within 30 days), a rule can handle it.

This is roughly 40 to 60% of support volume for most companies. And it's the least interesting 40 to 60%.

The Work That's Not Going Anywhere

Now here's the stuff AI can't do well:

Angry customers who need to feel heard. A customer who's been overcharged three months in a row doesn't want a form response. They want someone who acknowledges the screw-up, apologizes sincerely, and fixes it. AI-generated empathy reads as hollow because it is hollow.

Complex multi-issue tickets. "I upgraded my plan last week, but my invoice still shows the old amount, AND I can't access the new features, AND my team member got locked out when I changed plans." This takes investigation, not classification.

Sales-adjacent conversations. When a support ticket is really a buying signal ("Can your tool do X?"), a good support rep turns that into a sale. AI can't read the room like that.

Product feedback that actually matters. When a customer describes a problem in a way that reveals a deeper product issue, a skilled support rep flags it and explains the context to the product team. AI classifies it as "feature_request" and moves on.

Anything that requires judgment calls. Should we make an exception to our refund policy for a long-time customer? Should we escalate this to engineering or is it a known issue? Does this customer's frustration suggest they're about to churn? These are human decisions.

What Actually Happens to Support Teams

Here's the pattern I've seen play out at companies that adopt AI support:

Phase 1: Volume drops. AI handles the routine stuff. The support queue gets shorter. Team members have more time per ticket.

Phase 2: Quality goes up. With fewer tickets, support reps spend more time on each one. Responses get more thoughtful. Customer satisfaction scores improve.

Phase 3: The team shrinks (slightly) or refocuses. Some companies reduce headcount by 1 to 2 people through attrition, not layoffs. Others keep the same team but shift them toward proactive work: writing better docs, improving onboarding, doing customer success outreach.

Phase 4: The remaining team is more skilled. The repetitive work is gone. What's left requires problem-solving, empathy, and product knowledge. Support becomes a harder, more valuable job.

This isn't a future prediction. It's happening now at companies that have adopted automation.

The Numbers

Right now, a classification-based system can auto-resolve 60 to 70% of support volume for a typical SaaS or e-commerce company. The remaining 30 to 40% needs human attention.

As AI gets better at handling multi-turn conversations and accessing business data, that 60 to 70% might climb to 80%. But the last 20% — the conversations that need real human judgment — will be the hardest to automate and the most valuable to get right.

What This Means for You

If you're a founder doing your own support: automate the repetitive stuff today. You shouldn't be answering password reset questions at 11 PM. Your time is worth more than that. Let AI handle the 60% and spend your attention on the 40% where you can actually make a difference.

If you're a support rep: the reps who understand how to work alongside AI — setting up routing rules, improving automated responses, handling the escalations that AI can't — will be more valuable, not less. The ones at risk are the ones doing work that a $0.20 API call can do. Learn to be the person who configures the automation, not the person the automation replaces.

The Real Risk

The biggest risk isn't that AI replaces support teams. It's that companies use AI as an excuse to make support worse. Cut the team, automate everything, and let customers scream into a void when the automation breaks.

Bad AI support is worse than no AI support. Customers who get stuck in an automation loop with no way to reach a human will leave faster than customers who wait 2 hours for a human response.

The companies that win will use AI to handle the simple stuff fast and free up humans to handle the rest well. The companies that lose will use AI to avoid talking to customers entirely.