AI Account Executive: Career and Portfolio Guide

Picture a normal software sale. You give a demo, the buyer likes it, you send a contract.

Now picture the same call at an AI company. The person across the table is an ML engineer, and they don't care much about the demo. They want to see the model do their own work, on their own data, and score the result before anyone talks about price.

That's the job of an account executive (AE) at an AI company. You still close revenue, like any AE. What changes is who you're selling to and what they need to see before they sign.

This guide reads seven live AE postings, opened on September 26, 2026, and pulls out what the job actually involves. It ends with what a first year tends to look like and two pieces of portfolio work you can build before you apply.

What the job looks like in the postings

Every posting below says the AE owns the whole deal, from first call to close. The differences are who the buyer is and what proof gets them to sign.

Baseten's Account Executive - AI Native sells inference infrastructure (the servers that run AI models) to early-stage AI companies. The AE will "Run fast, technical sales cycles with founders and early engineering teams, from first call to first production workload." The deal isn't done at the signature. It's done when the customer is actually running on it.

Fireworks' Account Executive, AI Native Team is similar. The AE will "Lead technical and product conversations with founders, ML engineers, and infra leads" and "Work closely with internal teams to shape tailored solutions, POCs, and fine-tuning approaches." A POC is a proof of concept, a small trial on the customer's real work. After the sale, the AE will "Track account health and usage data to identify expansion opportunities."

Harvey's Account Executive, Mid-Market sells to law firms and legal teams. It asks the AE to "Conduct high velocity, tailored client evaluations, including product demonstrations and presentations," and to manage "the full customer lifecycle from prospecting to contracting, onboarding, growing, and renewals."

Decagon's Enterprise Account Executive sells AI customer-support agents to big companies. It asks for "Experience in complex solution sales and consistently closing 7 figure deals." WRITER's Enterprise account executive (West) asks the AE to "Become a trusted product expert, educating users on how to set up WRITER to unlock value, going hands-on where needed."

Clay's Account Executive (GTME - New Business) sells a sales and marketing data platform. It wants AEs who can "Build business cases that win budget and internal champions" and mentions "new-logo deals in the $50K-$100K range with multiple stakeholders." Perplexity's Commercial Account Executive turns "existing self-serve customers into Enterprise relationships" and runs the deal "from initial presentation, technical evaluation, negotiation, to closing and onboarding."

Put together, the seats fall into three types.

Before you prepare for an interview, read the posting and put it in one row. An infrastructure AE who can't talk with an engineer about speed and cost per request will struggle. An application AE who can't build a business case for a company's top lawyer will struggle in a different way.

Why these deals are harder to close

Three things come up again and again in these seven postings. None of them is unique to AI. Together they explain why an AI product is hard for a buyer to judge.

The test is the sale

Harvey, Clay, Fireworks and Perplexity all list evaluations, POCs or technical evaluation as part of the AE's job.

With an AI product, the buyer can't judge the quality of the output from a demo. A demo shows the model on examples the vendor picked. The buyer needs to see it do their work, on their data, and score the result. The AE designs that test with the customer.

The AI sales cycle playbook walks through how to set one up.

Signing is often just the start

Baseten's deal runs to "first production workload." Fireworks tracks "usage data to identify expansion opportunities."

Model and inference companies charge by how much the customer uses, as the public OpenAI API pricing and Claude pricing pages show. So getting the customer into real daily use may matter more to the account's value than the signature does.

How that affects your commission depends on the pay plan. Ask what earns credit and when. How OTE and sales pay plans work lists the questions to ask.

Engineers are part of the deal

Baseten tells the AE to "Pull in applied research and engineering when a deal needs depth."

At some companies that engineer is a forward deployed engineer, someone who builds with the customer. The AE's job is to hand them a deal with a clear scope. The FDE playbook shows what a good handoff gives them.

The explainer on what is different about selling AI covers these changes from the buyer's side.

Posted pay, and what each number covers

Opened live on September 26, 2026. These are employer-posted figures quoted word for word, not a market survey.

Read the last column before you compare two rows. Some ranges are on-target earnings, or OTE: base pay plus the commission you'd earn hitting 100 percent of your quota. One is a single figure, and one doesn't say what it is.

A 50/50 split on $240,000 OTE means a $120,000 base. That base is what you get paid if you close nothing.

When a posting doesn't give the split, ask for it on the first recruiter call. The guide to OTE and sales pay plans explains quotas, accelerators and clawbacks. The AI sales pay calculator lets you see what a plan pays at different levels of quota.

The experience the postings ask for varies more than the pay. Baseten asks for "2+ years of closing sales experience in a SaaS or infrastructure business." Clay asks for "4+ years in a closing role." WRITER asks for "5 years of quota-carrying enterprise SaaS sales experience."

If you haven't closed deals before, the usual way in is the SDR route in the SDR and BDR career guide.

What a first year tends to look like

No posting publishes a ramp schedule. This is the order the duty lines suggest, not a timeline any employer promised.

1. Learn the product until you can run the test yourself. WRITER wants AEs "going hands-on where needed." Baseten wants AEs who "follow model releases and can hold a real conversation with engineers." Expect to spend the first weeks building demos and reading how customers actually set it up.
2. Take over the deals you're handed, and find your own. Baseten says "You own pipeline generation, not just what's handed to you." Decagon lists "Consistent performance meeting pipeline generation targets for net new business." Plan on finding a large share of your own deals, often with an SDR.
3. Run your first tests with an engineer beside you. The first two or three POCs teach you which success measures customers accept and which tests drag on. Write down what worked.
4. Close, then get the customer live. When the product is priced by usage, the first invoice may be small. The account grows once the customer is using it for real, so stay close during onboarding.
5. Forecast accurately. Harvey asks AEs to "accurately forecast key sales performance metrics." By the end of the year your manager will judge you on whether the deals you said would close actually closed, as well as on the total.

What to build before you apply

Sales hiring runs on track record, and a portfolio doesn't replace a history of hitting quota. It does give an interviewer something concrete to talk about, and for an AI seat it's the strongest proof you can make yourself.

A test plan for a real public company

Pick a company that could plausibly buy the product you want to sell. Use only public information: its annual report, job postings, engineering blog and product pages. Write a two-page evaluation plan that covers:

• the workflow the product would change, and the person who owns it
• three success criteria that are specific and measurable, in the sense Anthropic's evaluation guide uses those words
• the test data the customer would need to provide, and who has to approve sharing it
• what happens if the test passes: who signs, and what the first real use is

Mark everything you guessed. An interviewer will trust a plan that says "unknown: current ticket volume" more than one that makes up a number.

A mock mutual action plan

A mutual action plan is the list of steps both sides agree to between "the test passed" and "the contract is signed." Write it as a table of dated steps, each with an owner on the buyer's side. Include security review, legal and procurement.

The open-source GTM skills include a create-mutual-action-plan skill and a build-business-case skill that show what information these documents need. Use them to check your draft, not to write it for you.

Bring both to the interview. The AE interview guide shows where they fit in a mock discovery call or pipeline review.

Next steps

• Prepare for the interviews with the AI account executive interview guide.
• Learn the steps of a deal in the AI sales cycle playbook.
• If you're earlier in your career, start with the AI SDR and BDR career guide and how to break into AI sales.
• Find current openings in the directory of companies hiring for AI sales roles.
• Compare the AE seat with technical sales and marketing work in the GTM engineer career guide.