How to Hire AI Engineers in 2026: A Guide for US and UK Companies

A lot of advice on how to hire AI engineers stops at "find someone who knows Python and LLMs". That is not enough in 2026. The job title now covers people who build retrieval systems, AI agents, test pipelines and model infrastructure. Some have shipped systems that real customers use. Others have only built demos.

This guide is for US and UK companies hiring their first or next AI engineer. It covers both markets where they differ, gives costs in dollars and pounds, and links every figure to its source.

What does an AI engineer do?

An AI engineer builds software products on top of AI models, usually large language models (LLMs) in 2026. The work includes connecting models to company data and building retrieval and agent workflows. It also covers testing output quality, controlling cost and speed, and keeping systems safe in production. It is closer to software engineering than research.

LinkedIn ranked AI engineer as the fastest-growing job in the US in its 2026 Jobs on the Rise list. It names LangChain, retrieval-augmented generation (RAG) and PyTorch as the most common skills. The most common previous roles were software engineer, data scientist and full-stack engineer. The median person moving into the role had 3.7 years of prior experience.

In practice, an AI engineer's work falls into five areas:

• Retrieval: RAG systems that answer from your own documents and cite them.
• Agents: AI agents that use tools such as a CRM, within set limits.
• Evaluation: writing test sets and measures that show whether a change made answers better or worse.
• Integration: connecting models to your product through APIs, queues and databases.
• Operations: watching cost, speed, errors and misuse once real users arrive.

Engineers who focus on language-model products are often called LLM engineers. The titles overlap; what matters is what the person has shipped.

AI engineer vs data scientist

A data scientist mainly analyses data and builds models to answer business questions (BLS). An AI engineer mainly builds and runs software that uses models.

The BLS put the median US data scientist wage at $120,230 in May 2025 (BLS). If you need custom models trained on your own data, you may want a machine learning engineer instead.

Do you need an AI engineer at all?

You need an AI engineer when AI is part of your product or a core workflow, and ready-made tools cannot do the job. Typical cases are answering customers from your own documents, automating multi-step work with agents, or processing documents at volume. If a paid tool or a single API call solves the problem, you probably don't.

Signs that you need one:

• A clear use case: you can name the task and a measurable result, such as fewer support tickets.
• Your own data: the system must use your documents or records.
• Real actions: the system will change records or contact customers, so mistakes cost money.

Signs that you may not:

• No defined problem: the brief is to "do something with AI".
• A product already exists: a mature tool does the job for less than building.
• Data is not ready: the information is scattered, missing or locked away.
• You only need a prototype: a strong full-stack developer can often build one.

How hard is it to hire AI engineers in 2026?

It is hard in both markets. ManpowerGroup's 2026 survey found AI skills are now the hardest skills for employers to find worldwide (ManpowerGroup). That is the AI talent shortage in practice. Jobs needing AI skills are growing much faster than the job market as a whole, and they pay a premium.

ManpowerGroup surveyed 39,063 employers in 41 countries in October 2025. Overall, 72% reported hiring difficulty, including 73% in the UK and 69% in the US. AI model and application development was the single hardest skill to find, cited by 20% (ManpowerGroup).

Job ads show the same shift. UK permanent job ads citing AI rose from 2,961 to 10,100 in two years, comparing the six months to mid-October (IT Jobs Watch).

Pay follows demand. PwC's 2026 AI Jobs Barometer found jobs needing AI skills grew 69%, against 9% for all jobs, with a 62% average wage premium (PwC). Indeed data shows senior roles made up 71% of the US rise in software job postings from May 2025 to May 2026 (CIO.com).

What are the ways to hire an AI engineer?

There are four main routes: hire an employee, engage a contractor, use a talent platform, or use staff augmentation. Staff augmentation means engineers from an outside firm, often offshore. The routes differ in cost, speed, control and risk. Many companies combine them, for example a contractor to start while a permanent search runs.

An hourly rate is only part of the cost. An employee brings benefits, tax and recruiting costs. A contractor or vendor brings a markup; KORE1 says IT staffing markups are "normally 35% to 50%" (KORE1).

Talent platforms pre-screen engineers, but conversion costs money. Arc charges 20% of annual salary if you hire someone full-time (Arc).

Staff augmentation gives you named engineers who work inside your team, under your management. Our guide to AI staff augmentation covers the model in detail. Check who exactly will do the work, which hours overlap with yours, and how the firm handles your data.

How do you hire an AI engineer, step by step?

Define the problem, write a specific job description, set a budget, and source candidates. Then screen them, run a structured technical interview and test real work with a paid trial or work sample. Finally, plan the first 30 days. The first step matters most, because it decides what good looks like at every later step.

1. Define the problem

Write one page before you write a job ad. Describe the task, the users, the data the system needs and what success looks like in numbers. "Answer 60% of routine support emails with a correct, sourced reply" is a problem. "Add AI to the product" is not.

Then decide what kind of engineer the problem needs. Answering from documents points to RAG experience. Taking actions in other systems points to agent experience. Custom models trained on your own data point to machine learning engineering.

2. Write the AI engineer job description

Keep it short and specific. Here is a template you can adapt:

• Title: AI Engineer (LLM applications).
• About the work: You will build [system] so that [users] can [outcome]. Success after six months means [measurable result].
• What you will do: Design retrieval and agent workflows, build test sets, ship features to production, and monitor cost, speed and quality.
• Must have: Production experience in Python or TypeScript, at least one LLM feature used by real people, and experience testing model output.
• Nice to have: Vector databases, agent frameworks, cloud deployment, or experience in your industry.
• How we work: Required overlap hours, team size, reporting line, and remote, hybrid or office.
• Pay and process: The salary or rate range, the interview stages and the expected timeline.

3. Set a budget in dollars or pounds

Platform and offshore rates are in the table above. For local hires, these are the 2026 reference points:

• US employees: BLS has no AI engineer category. In May 2025, median pay was $135,980 for software developers (BLS). For computer and information research scientists it was $140,300 (BLS).
• US contractors: Agency bill rates run $140-240 an hour for AI/ML engineers, of which the contractor takes home $95-160 (KORE1).
• UK employees: The median salary for permanent jobs citing machine learning is £75,000, and the 90th percentile is £115,000 (IT Jobs Watch).
• UK contractors: Median day rates are £600 for AI engineer contracts (IT Jobs Watch) and £550 for machine learning engineers (IT Jobs Watch). The software developer median is £525 (IT Jobs Watch).

4. Source candidates

Strong AI engineers rarely answer generic ads. Look where their work is visible: open-source projects, technical write-ups and talks. Ask your own engineers for referrals. In Ashby's data, 40% of referred candidates reached an interview, against about 3% of inbound applicants (Ashby).

Recruitment agencies usually charge 15% to 30% of first-year salary, according to the hiring marketplace Underdog.io.

How do you vet AI engineers?

Vet AI engineers in layers. Confirm their identity, check that their past work is real and was used in production, and test how they think in a structured interview. Then watch them do real work in a paid trial or work sample. Each layer catches problems the others miss, and none is enough alone.

Screen for shipped work

In a 30-minute call, ask the candidate to walk through one AI system they shipped. Ask who used it, what broke and how they measured quality. People who have only built demos tend to struggle with the last two questions.

Check identity

Remote hiring has a fraud problem. In a Gartner survey of 3,000 job candidates, 6% said they had taken part in interview fraud. Gartner predicts one in four candidate profiles worldwide could be fake by 2028 (HR Dive).

The FBI has warned that North Korean IT workers used "artificial intelligence and face-swapping technology during video job interviews" (FBI). It advises identity checks during interviewing, onboarding and employment. Keep cameras on, verify documents, and call referees through a company's main number rather than one the candidate supplies.

Use a structured interview

Ask every candidate the same questions and score the answers against an agreed scale. A 2022 re-analysis of hiring research found structured interviews had the highest average validity of the methods studied. They scored 0.42, against 0.33 for work samples and 0.31 for cognitive ability tests (HumRRO).

Budget the time. Ashby found technical hires go through five interview events on average, with 23.3 hours of total interview time (Ashby).

Test real work with a work sample or paid trial

A short, paid work sample on a realistic problem shows more than a puzzle on a whiteboard. Give the candidate a few documents or a small dataset. Ask for a working prototype and a short note on the trade-offs. Pay for the time, because strong candidates have other options.

For contractors and vendors, ask for a trial on real work. Toptal and Arc both advertise trials of up to two weeks (Toptal; Arc). Agree the deliverables and reviewer before day one.

Ask how they use AI coding tools

In Stack Overflow's 2025 survey, 84% used or planned to use them, but 46% distrusted their accuracy (Stack Overflow). The top frustration, cited by 66%, was answers that are "almost right, but not quite" (Stack Overflow survey).

Veracode tested more than 100 models on 80 coding tasks. AI-generated code introduced security flaws in 45% of cases (Veracode). That is why the interview questions below ask how they review it.

What should you ask in an AI engineer interview?

Ask about systems the candidate has built, how they measured quality, and how they handle failure, cost and security. Good AI engineer interview questions have no single right answer. Listen for trade-offs, numbers from real projects and honesty about what went wrong. Here are 11 questions and what a strong answer shows.

What should the first 30 days look like?

The first 30 days should produce a quality baseline and one small, shipped improvement, not a finished product. Give the engineer access, data and a clear owner on day one. Agree a written 30-day plan, review it weekly, and decide at day 30 whether the hire or engagement is working.

• Week 1: Sign the NDA and IP terms, then set up accounts, data access and a codebase walkthrough. Meet the people who will use the system.
• Week 2: Build a test set from real examples and measure current quality.
• Week 3: Ship one small improvement behind a switch and measure it against the baseline.
• Week 4: Compare results with the success measure from step one, then agree the next 60 days.

What should US companies check before hiring?

US companies need to settle worker classification, visa status, code ownership and time zones before anyone starts. The rules differ for employees, contractors and offshore vendors. Common problems are treating a contractor like an employee and signing a contract that never transfers the code. Check each point with an employment lawyer.

• Employees: Every US employer must complete Form I-9 for each person hired to work in the US (USCIS).
• Visas: H-1B visas are capped at 65,000 a year, plus 20,000 for certain advanced-degree holders. A $100,000 payment on certain new petitions began in September 2025. A federal court vacated the guidance implementing it in June 2026, and DHS says it will comply for now (USCIS). A September 2026 proclamation extended it to September 2027, but a second federal judge has since struck down the fee and that proclamation (SHRM). Take current legal advice.
• Contractors: The IRS weighs behavioural control, financial control and the type of relationship (IRS). California's ABC test treats a worker as an employee unless the hiring company meets all three of its conditions (California LWDA).
• Code ownership: Under US law, a "work made for hire" covers an employee's work. Commissioned work only qualifies in nine listed categories, with a signed written agreement (US Copyright Office). Software is not on that list, so contractor agreements should assign the rights in writing.
• Offshore vendors: Foreign firms use Form W-8BEN-E to document their tax status (IRS). Ask for it before the first payment.
• Time zones: Write the overlap hours into the contract. A team in Asia usually needs an evening shift to cover US mornings.

What should UK companies check before hiring?

UK companies need to check IR35 for contractors, and right to work and visa sponsorship for employees. They also need to budget for employer costs, apply UK GDPR to any data reached from abroad, and secure code ownership. Get the contract right before work starts.

• IR35: Under the off-payroll working rules, the client usually decides a contractor's employment status and issues a status determination statement. For small clients, the contractor's own company decides instead (GOV.UK).
• Who counts as small: A company is medium or large if it meets two of three tests. They are over 50 employees, turnover above £15 million and a balance sheet above £7.5 million. The money limits rose for financial years starting on or after 6 April 2025, with transitional rules (HMRC).
• Overseas engineers: HMRC says a client "does not need to consider" the rules where there is no UK tax or National Insurance liability. A worker living outside the UK is one of the factors it lists (HMRC).
• Right to work: Check every employee's right to work in the UK. The fine is up to £60,000 for each illegal worker (GOV.UK).
• Visas: You usually need a sponsor licence to employ someone from outside the UK (GOV.UK). Skilled Worker pay must usually be at least £41,700 or the job's going rate, whichever is higher (GOV.UK).
• Employer costs: Employer National Insurance is 15% on earnings above £5,000 a year in 2026-27 (GOV.UK). Pension auto-enrolment adds at least 3% on earnings between £6,240 and £50,270 (GOV.UK).
• Data protection: Letting a provider abroad access personal data on your systems is a restricted transfer. The ICO's own example uses a provider in India (ICO). It needs a transfer mechanism. If you rely on a safeguard such as the IDTA or the Addendum, you also need a transfer risk assessment.
• Code ownership: A contractor "will usually retain copyright" in what they produce unless the contract says otherwise (GOV.UK). Put the assignment in writing.

What does hiring an AI engineer cost in total?

On our arithmetic from BLS and IT Jobs Watch data, a median US developer costs about $194,000 a year with benefits. A UK employee at the machine learning median costs about £86,800 with employer National Insurance and pension. Contractors cost more per day. Offshore staff augmentation costs least per hour but needs more management time.

US example: take the BLS median for software developers, $135,980 (BLS). Wages make up 70% of private-sector employer compensation costs, and benefits the other 30% (BLS). On that ratio, the total is about $194,000. A 20% recruiting fee adds about $27,000 more.

UK example: take the £75,000 median for permanent machine learning jobs (IT Jobs Watch). Employer National Insurance at 15% above £5,000 adds £10,500. The minimum pension contribution adds about £1,320. The total is about £86,800 before any recruiting fee.

Contractor example: at the UK median of £600 a day (IT Jobs Watch), 20 working days cost £12,000 a month. In the US, KORE1's $140-240 bill rate (KORE1) means $22,400-38,400 for a 160-hour month.

Offshore example: at Accelerance's $31-41 senior rate for Asia, a 160-hour month costs $4,960-6,560 (Accelerance). Add your own management time, and any premium for overlap shifts.

How long does it take to hire an ML engineer?

Expect roughly six to eleven weeks for a permanent AI or ML engineer, before any notice period. In Ashby's data, the median technical hire took 40 days from application, and 76 days from opening the job to the first hire (Ashby). Platforms and staff augmentation firms advertise much faster starts.

In EMEA, which includes the UK, Ashby's average time to hire has stayed between 37 and 43 days since 2021 (Ashby).

Then add notice. In the UK, the employment contract sets how much notice a person must give (GOV.UK). For a sponsored hire in either country, add time for the licence or petition and the visa.

Other routes are faster. Toptal says you can typically hire in about 48 hours (Toptal). Ask any staff augmentation firm for a written timeline to named profiles and a start date.

What are the red flags when hiring AI engineers?

The main red flags are no production experience and no way of measuring quality. Others are vague answers on security and data, and identity gaps in remote interviews. With vendors, watch for refusals to let you interview the actual engineers, sign an NDA or assign code ownership in writing.

• Demos only: every project is a notebook or a prototype, and none had real users.
• No testing: they cannot say how they measured answer quality.
• Model first: they propose fine-tuning before trying retrieval, prompts or a simpler tool.
• Blind trust in AI code: they paste generated code without tests or review.
• Security gaps: they have not heard of prompt injection, or give agents broad permissions.
• Identity gaps: reluctance to appear on video, details that do not match, or referees you cannot verify independently.

Where does Webify.AI fit?

Webify.AI is one option if you choose staff augmentation: named AI engineers in India who work inside your team. It suits companies that want lower hourly costs and can manage remote engineers. It does not suit companies that need on-site staff, or a vendor with security certifications.

Webify.AI is an AI and software company headquartered in Ahmedabad, India, with its delivery centre there. It is an IBM Partner and builds AI agents. Its rates are below, and the pricing page has the details.

These are its terms, stated plainly:

• Process: two named profiles within 48 hours. You interview the engineers before anyone starts, and an NDA is signed before the first call.
• Trial: two weeks, with no charge if you stop within the first five working days.
• Replacement: free, within five working days.
• Overlap: 09:00-13:00 UK time. For US Eastern time, 09:00-12:00 ET through an evening shift at 15% extra.
• Ownership: you own all the code.
• Security: Webify.AI holds no security certifications, such as ISO 27001 or SOC 2. If your procurement requires one, choose a vendor that has it. Read our security page and ask about anything it does not answer.

To see profiles, start at hire AI engineers. There are separate pages for UK companies and US companies.

Sources

• Ashby, Recruiter Productivity, 2026 Talent Trends Report
• Ashby, EMEA Recruiter Productivity
• Ashby, Are referred candidates more likely to get hired?
• Accelerance, 2026 outsourcing rate trends
• Arc, pricing
• BLS, Data scientists, Occupational Outlook Handbook
• BLS, Software developers, Occupational Outlook Handbook
• BLS, Computer and information research scientists, Occupational Outlook Handbook
• BLS, Employer Costs for Employee Compensation, June 2026
• California Labor and Workforce Development Agency, ABC test
• CIO.com, IT hiring sees a boost as software development jobs slowly bounce back
• FBI / IC3, North Korean IT workers, alert I-012325-PSA
• GOV.UK, Understanding off-payroll working (IR35)
• GOV.UK, Penalties for employing illegal workers
• GOV.UK, UK visa sponsorship for employers
• GOV.UK, Skilled Worker visa, your job
• GOV.UK, Rates and thresholds for employers 2026 to 2027
• GOV.UK, Workplace pensions, what you, your employer and the government pay
• GOV.UK, Ownership of copyright works
• GOV.UK, Handing in your notice
• HMRC Employment Status Manual, ESM10006A
• HMRC Employment Status Manual, ESM10020
• HR Dive, Gartner on fake job candidates
• HumRRO, Is cognitive ability the best predictor of job performance? (Sackett et al., 2022)
• ICO, International transfers, a guide
• IRS, Independent contractor or employee?
• IRS, About Form W-8BEN-E
• IT Jobs Watch, AI Engineer contracts, UK
• IT Jobs Watch, Machine Learning Engineer contracts, UK
• IT Jobs Watch, Machine Learning permanent jobs, UK
• IT Jobs Watch, Artificial Intelligence permanent jobs, UK
• IT Jobs Watch, Software Developer contracts, UK
• KORE1, Tech contractor hourly rates 2026
• Lemon.io, AI engineer rates
• LinkedIn, Jobs on the Rise 2026, US
• ManpowerGroup, 2026 Global Talent Shortage
• OWASP Top 10 for LLM Applications 2025, LLM01 Prompt Injection
• PwC 2026 Global AI Jobs Barometer press release (via AAP)
• SHRM, $100K H-1B fee extended through 2027
• Stack Overflow, 2025 Developer Survey press release
• Stack Overflow, 2025 Developer Survey, AI section
• Toptal, Hire AI engineers
• Underdog.io, Recruitment agency costs
• US Copyright Office, Title 17, Chapter 1 (section 101 definitions)
• USCIS, Form I-9
• USCIS, H-1B specialty occupations
• Veracode, AI-generated code security research