Your AI engineer starts Monday. When do they actually ship?

The offer is signed and the start date is set. The question the spreadsheet never asks is the one that decides whether your AI feature ships this quarter: how many weeks pass between the new hire's first day and the day a real feature reaches production?

Most hiring math stops at the salary. The ramp is where the calendar actually goes.

• A senior or technical hire takes 6 to 12 months to reach full productivity, per Gallup and standard onboarding benchmarks.
• Engineers using AI tooling daily still hit their 10th merged PR around day 49; peers without it take roughly 91 days.
• Onboarding alone runs $3,000 to $7,000 per hire once you count access, training, and manager time.
• The first shipped AI feature usually trails the start date by a full quarter, not a sprint.

The gap nobody budgets for: hire date vs first shipped feature

Time-to-hire gets tracked because it is painful and visible. The req sits open, the pipeline stalls, and everyone feels it. We wrote about that funnel separately in how long it really takes to hire a senior AI engineer. But time-to-hire ends on day one. The clock that matters for your roadmap keeps running.

Call it time-to-first-feature: the span from start date to a production AI capability your users touch. For a senior AI engineer joining a team without an existing AI stack, that span is rarely under 90 days. The work before the first feature is real, and skipping it produces the broken demos that never survive contact with production traffic.

What the ramp actually looks like for a senior AI engineer

Weeks 1 to 4: access, context, and the codebase

Laptop, SSO, cloud roles, secrets, and model-provider keys. Then the harder part: reading the codebase, mapping the data, and learning why the last three architecture decisions were made the way they were. A strong hire is reviewing PRs and filing small fixes by week three. They are not yet designing your retrieval pipeline.

Months 2 to 3: the first real pull requests

This is where the 49-day-to-10th-PR figure lands. The engineer is now shipping, but the early PRs are scoped small on purpose. A senior person spends this window pressure-testing your assumptions: is the data clean enough for retrieval, do you have evals, what does latency look like under load. These questions are the value, and they take weeks to answer honestly.

Months 4 to 12: full productivity

By month four a good hire is owning features end to end. Full productivity, the point where output matches a tenured peer, lands somewhere between months 6 and 12. That is not a knock on the engineer. It is what ramping into an unfamiliar product and stack costs.

The ramp has a dollar figure

Base salary for a senior AI engineer in the US sits around $220,000 to $300,000, and the fully loaded year-one cost lands closer to $290,000 to $480,000 once you add payroll tax, benefits, recruiting fees, and compute. We broke that number down in the true cost of a senior AI engineer in 2026.

Now apply the ramp. If full productivity arrives at month six, you have paid roughly half a year of fully loaded cost, call it $145,000 to $240,000, before the hire is operating at the level you were buying. The feature that justified the headcount may not be live until the calendar is already deep into the year.

The risk tail is worse than the ramp itself. SHRM puts the cost of replacing an employee at 50 to 200 percent of salary, and for a senior engineer a mis-hire discovered at month five can run $150,000 to $340,000 once you count the months you knew it was not working but had not acted. The longer the ramp, the later you learn whether the bet paid off.

What compresses the ramp

Three things move the first-feature date earlier, and none of them are about working the person harder.

Write the context down before they arrive. A short architecture note, a data dictionary, and a list of the decisions you do not want relitigated cut the week-one-to-four block roughly in half. The teams shipping fastest treat onboarding docs as a deliverable, not an afterthought.

Give them a real first task, not a tour. A scoped, shippable AI feature with a clear definition of done teaches the codebase faster than any reading list. Watching how someone scopes and ships that first task also tells you in week two what you might otherwise learn in month five.

Do not make the first hire build the platform and the feature at once. If the retrieval layer, the eval harness, and the first user-facing capability all land on one new person, the ramp stretches because every task is also a yak shave.

When a subscription skips the ramp entirely

The ramp exists because a new employee has to learn your product, your stack, and your history before they are useful. A senior engineering team that has shipped this pattern many times starts further along the curve. They still need your context, but they are not learning RAG, evals, or production LLM cost control on your dime.

This is the model behind a Boundev first task: you describe the AI feature in plain English, and senior engineers ship a working version in 5 to 7 days instead of you waiting a quarter for a new hire to ramp. There is no onboarding cost, no recruiting fee, and no six-month wait to find out if the fit was right. You can see how the day-to-day works in how it works, and the monthly commitment is laid out on the pricing page.

The point is not that hiring is wrong. A permanent AI engineer is the right call when the work is continuous and core. The point is that the ramp is a real, expensive part of the hiring decision, and it deserves a line in the same spreadsheet as the salary.

Frequently asked questions

How long until a new AI engineer ships their first feature?

For a senior hire joining a team without an existing AI stack, plan on roughly 90 days to a first production feature and 6 to 12 months to full productivity. The first month goes to access and context, months two and three to small scoped PRs, and month four onward to owning features end to end.

Does AI tooling make new engineers ramp faster?

Somewhat. Engineers using AI tools daily reach their 10th merged PR around day 49 versus 91 days without, and onboarding documentation tools can cut early ramp meaningfully. But tooling shortens the coding ramp, not the time it takes to learn your product, data, and architecture decisions.

How do I reduce the time to my first AI feature?

Write the architecture and data context down before the hire starts, give them one scoped shippable feature instead of a tour, and do not ask the first hire to build the platform and the feature simultaneously. If you need a feature live in days rather than a quarter, a senior engineering subscription skips the ramp because the team has shipped the pattern before.

Is a long ramp a sign of a bad hire?

No. A 6-to-12-month ramp to full productivity is normal for a senior engineer learning a new product and stack. It becomes a problem only when you have not budgeted for it, or when the first hire is loaded with platform work and feature work at the same time.