Training staff in AI or hiring AI talent: a cost-benefit analysis for HR

Why the question arises in 2026

Two trends meet. On one side, the World Economic Forum's Future of Jobs Report 2025 ranks AI and big data as the fastest-growing skills. The IBM Institute for Business Value reported in August 2023 that the executives it surveyed estimated that 40% of their workforce would need to reskill within three years because of AI and automation.

On the other side, the French market for IT profiles has eased. DARES (the French labour ministry's statistics department) observed a clear drop in tensions in IT and telecommunications in 2024, and France Travail's 2026 Labour Needs Survey records fewer planned hires of IT engineers.

Finally, Article 4 of the EU AI Act requires companies that use AI systems to take measures promoting AI literacy among their staff (see AI Act Article 4). This need concerns every user and cannot be met through hiring.

Usage skills or technical skills?

This is the first question to settle:

	AI usage skills	Technical AI skills
Definition	Using existing AI tools in one's job, checking and correcting their output	Designing, training, integrating, deploying or technically supervising AI systems
Examples	Assisted writing, document analysis, information search, simple automation	Data science, model engineering, MLOps, integrating language models
Population concerned	A large share of employees	A few specialists
Natural lever	Training	Hiring, a service provider, or long training for people who are already technical

Most organisations need both, in very different proportions.

How much does a hire cost? The method

Recruitment costs vary too much by profile, region and channel for a market average to be useful. Calculate them with your own data:

Component	How to quantify it
Acquisition cost	Job ads, recruitment agency (contractual fees), time spent by recruiters and managers in interviews
Vacancy cost	Vacancy duration × value of work not done, or cost of temporary staff
Onboarding cost	Time to full autonomy × unproductive share of salary, plus tutoring time
Pay	Any gap between the market salary and the internal pay scale
Departure risk	Probability of leaving within two years × cost of a new hire

For duration, Apec (the French executive employment association) provides a benchmark for managerial and professional staff: according to its 2024 barometer, the average duration of the last executive hire stabilised at around 12 weeks in 2023, and reached 15 weeks in manufacturing. 18% of SMEs and 20% of large companies took more than 16 weeks to complete it. This time does not include the onboarding period, which comes on top.

How much does training cost? The method

Component	How to quantify it
Training fees	Provider quote, or platform cost per learner
Training time	Hours × employee's fully loaded hourly cost
Support	Time of tutors, internal experts and managers
Time to application	Time needed to reach the expected level on the job
Available funding	Depending on OPCO rules and French funding schemes (see Funding workplace training in France)

Then compare the two options over the same period (two or three years), with a low and a high estimate for each component.

A three-question decision framework

1. What type of skill? For usage skills, training is the natural lever: employees already know the job, and applying AI to that job is what creates value. For advanced technical skills, hiring or using a service provider is often unavoidable in the short term.

2. How much time? If the technical skill is needed in the coming months, long training will arrive too late. If the horizon is more than a year, training people who are already technical (developers, analysts) can become competitive.

3. What lifespan? A skill defined by a specific tool ages quickly. Training on principles (checking output, data protection, choosing the right tool) produces a more durable skill. See Skills obsolescence.

Qualitative comparison

Criterion	Internal training	External hiring
Knowledge of the job and the company	Already there	Built during onboarding
Time	Depends on the target level: short for usage, long for expertise	Recruitment time, then onboarding
Cost	Training fees and training time	Acquisition, vacancy and onboarding costs, pay gap
Scale	Suited to large populations	Suited to a few profiles
Risks	Training not applied on the job	Hiring mistake, early departure

Good practice

• Start with usage skills. They concern the most people and determine the value the organisation will get from AI. See Which AI skills to train first.
• Identify internal champions. Employees who already use AI in advanced ways can become go-to people before any hire.
• Be clear about what cannot be trained quickly. Some technical expertise requires a hire or a service provider.
• Measure. Track skills progress and its effect on business indicators to adjust the balance (see How to measure the ROI of a training plan).