How to add AI: build a team, buy a platform or hire experts

Route one: build the team

You hire engineers who build and own your AI capability: an AI platform lead, data engineers, MLOps engineers and LLM engineers. The work, the code and the know-how stay inside the company.

This route fits when AI is part of what you sell or how you compete, and when the problem is specific enough that no off-the-shelf platform solves it. It is the slowest route to a first result and usually the most durable, because the people who built the system can change it.

What it needs: a lead who has built a platform before, someone who owns data quality, and engineers who can put models into production and keep them running. The first hire matters most, because the lead shapes every hire after it.

Route two: buy the platform

You license a platform, such as Palantir Foundry and AIP or a cloud AI platform, and engineers configure it to your problem. Those engineers are usually forward deployed engineers: they work inside your program, close to your users and your data, and adapt the platform to what you actually need.

Demand for the role has grown fast. Bloomberry found that postings titled “forward deployed engineer” grew 1,165% in January to October 2025 against the same months of 2024 (Bloomberry, updated January 25, 2026).

What they cost, from employer ads and a published analysis of postings:

• Palantir advertises its Forward Deployed Software Engineer role in New York at $135,000 to $200,000 a year in salary, with stock and a sign-on bonus possible on top (read October 11, 2026).
• OpenAI advertises its Forward Deployed Engineer role in San Francisco at $185K to $300K, with equity offered (read October 11, 2026).
• Bloomberry found a median of $173,816 across postings that disclosed a range. It does not say whether that figure is base or total.

The forward deployed engineer role page lists more employer ads, each linked and dated.

This route is faster to a first result, and you can start with engineers on contract while you decide what to hire permanently. The trade-off is dependence: on the vendor for the platform, and on a small number of people for how it fits your business.

Route three: hire the experts

You bring in qualified professionals, such as lawyers, accountants, clinicians and finance specialists, to write, grade and rank the data that trains and tests a model in their field.

This route fits two kinds of company. Vertical-AI builders in legal, tax, health or finance need expert judgment to make a model good at professional work. Any company deploying a model into professional work needs someone qualified to say whether its output is good enough.

What it needs: a pool of vetted, conflict-checked professionals, a clear unit of work (a graded answer, a ranked set, a written rubric), and a structure that respects each profession’s conduct rules. Costs depend on the profession and the unit of work. The rates AI labs and expert platforms publish are set out, with sources, on our domain experts role page and in how AI labs source domain experts.

The common failure: a platform with nobody to run it

The most common way this goes wrong is buying a platform and assuming it will run itself. A platform is a set of capabilities. Turning them into a working system for your business takes engineers who understand both the platform and your data, and they take time to hire.

The symptom: the license is signed and the pilot scheduled, while the people who should configure it are still being interviewed. The fix is to plan the people before the purchase: who configures the platform, who owns the data, who judges the output and whether each is a hire, a contractor or an expert.

How to decide

Most companies combine routes: a platform to start, a small permanent team to own it and experts where the work is professional judgment.

In practice

Answer three questions before you buy anything: what you need to own in two years, how soon you need a first result and who would run a platform if you bought one tomorrow. If the third answer is “nobody yet”, the next step is a hire, not a purchase.

Where to start

Start with a 30-minute consultation. Bring your answers to the three questions above, and we will tell you which route fits, what the first hire should be and whether a fuller plan is worth doing. If it is not, we say so.

When a fuller plan is worth it, a Kernavique Capability Blueprint sets out the roles, the hiring sequence, the contract and permanent mix, and what those people cost. It ends in a build-versus-buy recommendation you own.