■ HIGH RISK ■ Customer Service
Tier-1 support — password resets, known issues, scripted troubleshooting — is being handed to AI right now, and it isn't coming back. The humans who remain will handle escalations, angry enterprise accounts, and the problems nobody has documented yet.
“Have you tried turning it off and back on? AI asks the same question, faster.”
Our AI replacement risk score — how we score jobs
Most support work is pattern matching against a knowledge base. A specialist's queue is dominated by tickets that have been solved ten thousand times before: locked accounts, printer drivers, VPN configs, the app that broke after an update. The job's actual skill mix is one part technical triage, one part translation (turning 'it's doing the thing again' into a reproducible bug), and one part emotional labor with users who are having the worst hour of their week.
Large language models are close to a perfect fit for the first category, which is why support was among the first departments where companies publicly cut headcount and pointed at AI. Modern support bots don't just answer FAQs — they read the user's account state, run diagnostics, push fixes, and write the ticket summary. Deflection rates that would have sounded like vendor fiction in 2022 are now routine, and every deflected ticket is a shift's worth of tier-1 work that no human sees. The economics are brutal: support is a cost center, and AI is the cheapest agent that never sleeps or quits.
The residue is genuinely hard, though. Escalations where the knowledge base is wrong, intermittent bugs that need real investigation, enterprise clients whose contracts require a named human, and the furious customer who needs de-escalation before they'll accept any fix — these keep humans in the loop. The role is bifurcating: fewer, more senior people doing complex troubleshooting and bot supervision, with the classic entry-level help desk shrinking underneath them. That lost bottom rung matters, because it's how most sysadmins and engineers got their start.
Automatability: our editorial assessment of current and near-term AI capability
This one isn't a forecast — it's a status report. AI deflection of tier-1 tickets is standard practice today and several major companies have already restructured support around it. Expect entry-level help desk roles to contract sharply through 2028, with surviving positions concentrated in escalation engineering, enterprise support, and bot supervision by around 2030.
The entry level is genuinely shrinking — companies now deflect a large share of tier-1 tickets to AI, and some have said so openly in layoff announcements. But complex troubleshooting, enterprise support, and escalation work still need humans. Treat tier-1 as a fast-moving on-ramp to something more senior, not a destination.
For scripted, known-issue tickets: it substantially already has. For novel bugs, angry enterprise clients, and problems that require investigation across systems, there's no full replacement on the visible horizon. The realistic picture through 2030 is fewer humans handling a harder, weirder residue of tickets while AI absorbs the volume.
Two directions pay off. Deep technical: networking, cloud platforms, security, or scripting, which convert into sysadmin and SRE-adjacent roles. Or AI operations: maintaining knowledge bases, evaluating bot answers, and handling what the bot escalates. Either way, document your de-escalation wins — soft skills on hard accounts are durable currency.
Yes — anything undocumented. Intermittent faults, misleading error messages, problems spanning multiple vendors' systems, and situations where the official fix is wrong all trip up AI, which is only as good as the knowledge it was trained or grounded on. It's also poor at earning back trust after a bad outage, which is half of enterprise support.