What OpenAI’s Support Openings Say About Getting Hired

Somewhere in San Francisco, a hiring manager at OpenAI wrote a job description that you’ll probably be measured against, wherever you apply next. To give you a sense of where this is going, here’s how a few of those postings describe the work.

You can be annoyed by that, and plenty of us are, but this is how our market works. When companies like OpenAI and Anthropic decide what a role should look like, other companies use their job descriptions as a template, and within a year or two the same words appear in postings everywhere. Reading those openings early gives us time to prepare, and time is what keeps us in demand when the market changes direction.

So I went through OpenAI’s support openings across San Francisco, Washington and Dublin, and the most encouraging thing I found is that experienced people in our field already have most of what these roles ask for. What usually needs work is the way we describe it, and I know that part is uncomfortable. We got into this work to put customers first and to build teams we’re proud of, and talking about ourselves feels wrong. So I wrote this piece to help you find the words for what you already do, whether you’re an employee or a consultant.

Support has a new name, and a much bigger job

What we used to call Support is becoming User Operations, and the new name carries a new way of thinking about the job.

Support used to begin when something went wrong. A customer got stuck, wrote in, and someone helped. User Operations looks at the whole path a person takes with a product, from the day they sign up, through onboarding and the moments they get stuck, to the point where they use it well enough to stay and expand. One of the postings describes the team’s job as helping customers adopt AI and get real value from it, and fixing problems is only one part of that.

The rest of the market is moving the same way. A year ago Intercom’s CEO predicted that the ideal service agent would cover the whole lifecycle, from pre-sales and onboarding to support and expansion. This October Fin (ex-Intercom) launched an agent for onboarding, adoption and upsell. Support, success and onboarding are folding into one journey, and the people who run it are being renamed accordingly.

In fashion terms, “customer support” is last season’s label. The work is still recognisable, but it’s cut wider now, and the companies hiring for it search for different words. That’s why it’s worth updating your headline and your CV to match.

Entry level now means experienced

The most junior opening in this set asks for five years of experience. Most ask for eight or more, and the senior roles ask for ten or twelve.

To see why, follow the simple work, the same work most of us learned the job on. AI and outsourced partners handle it, and one posting expects those partners to “expand automation and deliver measurable capacity gains,” which means they’re paid to automate their own work. What remains for the in-house team is judgment, and judgment comes from years of seeing how customers, products and teams behave.

A 2025 Stanford study led by Erik Brynjolfsson found the same pattern across the economy. In the jobs most exposed to AI, customer service among them, employment fell for people in their early twenties and held steady or grew for experienced workers.

The title on these roles may say entry, but the work behind it calls for someone who has seen enough to make good decisions alone. The market has restructured around judgment, and judgment is what your years in this field have given you. So if a role sounds junior but describes work you know well, apply. It isn’t a step back.

Each role now runs like a small startup

In a traditional support role, you receive a process and your job is to run it well. These roles begin with a problem and expect you to work out the rest, which is exactly how a small startup operates.

The safety analyst posting describes it clearly. “The work starts with the problem, not a prescribed process.” The person in that role decides “what needs to be built, what can stay lightweight, and when the work is ready to scale, transfer, or stop.” In plain words, you look at a problem, build the smallest thing that solves it, check whether it works, and then decide whether to grow it, give it to another team, or drop it.

You’re also expected to build the tools yourself. The same posting asks for someone who can “build or prototype practical solutions without waiting for a dedicated technical team,” and the AI support engineer writes Python with Codex to automate repetitive work. Today that means using AI to write the scripts, workflows and small internal tools you once had to request from engineering and wait months for. As in a startup, you own the outcome from start to finish, and the tools you build become part of how the whole team works.

If you’re wondering what you’d build, Plain recently reviewed more than 1,500 support workflows its customers had built, with close to two million runs between them, and published the ten most common.

What I built as a product support lead+

I worked this way myself as a product support lead. The product shipped updates faster than any team could keep up with by hand, so I decided which tools would give us back the most hours, and built them. One was an AI agent that reviewed what our answer bot told customers and which actions it took, then sent engineers and knowledge managers a clear list of what to fix. Another was a prioritization model for the tickets the bot couldn’t handle. It read every incoming case and ordered the queue objectively, so the most urgent and most valuable conversations reached a person first. I didn’t have a process for either of them, so I started with the problem.

You can build workflows like these inside a helpdesk that supports them, or on your own with tools like Codex or Claude Code.

But isn’t my experience with SLAs and vendors outdated now?

Your experience still counts, and these postings prove it. SLAs, quarterly business reviews, workforce forecasting, Salesforce Service Cloud and outsourcing contracts all appear in them.

What changed is what you’re expected to do with them. In the past, the job was to run these well. You met the SLA, ran the review, and kept the forecast accurate. Now you’re also expected to change them when the situation changes. When an AI agent starts handling half of the conversations, the old SLA may measure the wrong thing, so you rewrite it. When automation changes how many contacts reach people, last year’s forecast stops working, so you rebuild it. When an outsourcing partner is expected to automate, the contract needs new targets, so you renegotiate it.

That’s why these companies look for both kinds of experience in one person. Someone who knows only the old tools can keep them running but can’t adapt them, and someone excited about AI without that operational background doesn’t know what to adapt.

Automation comes first on the scoreboard

The support operations manager is measured on “automation rate, cost to serve, first-response time, productivity, and quality,” in that order. Automation leads the list and quality closes it. Another posting sets the goal of scaling “without linear headcount growth,” which means the team should be able to handle far more customers without hiring at the same pace.

The pay is the encouraging part. Where salaries are listed, they range from $216K to $310K plus equity. For people whose experience fits this new shape, the price tag went up.

Your experience needs re-tailoring, and the fabric is still good

April Dunford, who wrote the book on positioning, says the category you place yourself in decides who you’re compared with. Present yourself as a support manager in the old sense, and you’re compared with an AI agent plus an outsourced team, a comparison that comes down to cost. Present the systems you built, and you’re compared with the small group of people who can build them.

You don’t need a new career for this. You need new wording for the career you already have. Here’s the same work in the old language and in the new one. The numbers are illustrations, so swap in your own.

Each new line describes something that kept working after you, which is exactly what these postings are buying. Consultants can use the same approach, since their strongest offer is the system that keeps running after the engagement ends.

But I’m not good at selling myself!

Most of us aren’t, and that’s fine. You don’t have to sound like a salesperson. You only have to describe what changed because you were there, in numbers someone can check.

New doors are opening

Some paths barely existed a year ago. OpenAI is building a government support team “from the ground up” for agencies working under FedRAMP rules. Several postings bring support into launch readiness, before a product ships, and one mentions safety for connected devices. Regulated, pre-launch and hardware support are turning into specialties of their own, and they need people who already know how support works.

Where to go from here

If this piece was useful, subscribe and you’ll get the next forecast as soon as it’s out. It’s free.

If you’re an experienced practitioner, I’m also building a directory where companies can find and hire people like you directly. Twelve founding seats are open, listing is free, and I review every application myself.