Upskilling vs Reskilling: Which Path Fits You?

Upskilling vs reskilling is usually explained with definitions: upskilling deepens the skills for your current role, reskilling prepares you for a different one. The harder question is which one you need now. Two sources help answer it: the World Economic Forum's survey of employers about the rest of this decade, and 2025 Stack Overflow survey answers on how often tech and data professionals actually learn (Stack Overflow survey 2025).

import pandas as pd

so = pd.read_csv("results.csv", low_memory=False)          # 2025 Stack Overflow Developer Survey
roles = ["Data engineer", "Data scientist", "Data or business analyst", "AI/ML engineer", "Developer, full-stack", "Developer, back-end"]
pros = so[so.LearnCodeChoose.fillna("").str.startswith(("Yes, I am not new", "No, I am not new"))].copy()
pros["learned_new"] = pros.LearnCodeChoose.str.startswith("Yes")
pros["experience"] = pd.cut(pros.WorkExp, [0, 5, 10, 20, 100], labels=["up to 5 years", "6-10", "11-20", "21+"])

print("U1 experienced coders who learned a new technique or language in the past year")
for band, g in pros.groupby("experience", observed=True):
    print(f"   {band:<14}{g.learned_new.mean():>5.0%}   n={len(g):,}")

print("\nU2 by role: learned something new | learned AI tools for their job")
ai = so.dropna(subset=["LearnCodeAI"])
for role in roles:
    p = pros[pros.DevType.eq(role)]
    a = ai[ai.DevType.eq(role)]
    print(f"   {role:<26}{p.learned_new.mean():>5.0%} | {a.LearnCodeAI.str.contains('required for my job').mean():>4.0%}   n={len(p):,}")

print("\nU3 'Is AI a threat to your current job?' (data roles)")
data = so[so.DevType.isin(roles[:4])].dropna(subset=["AIThreat", "LearnCodeAI"])
print("   " + " | ".join(f"{k} {v:.0%}" for k, v in data.AIThreat.value_counts(normalize=True).items()) + f"   n={len(data):,}")
data = data.assign(learned_ai=data.LearnCodeAI.str.contains("required for my job"))
for learned, g in data.groupby("learned_ai"):
    label = "learned AI tools for the job" if learned else "did not"
    print(f"   {label:<30} say yes, a threat: {g.AIThreat.eq('Yes').mean():.0%} (n={len(g)})")

U1 experienced coders who learned a new technique or language in the past year
   up to 5 years   81%   n=11,443
   6-10            71%   n=9,385
   11-20           67%   n=11,753
   21+             65%   n=8,809

U2 by role: learned something new | learned AI tools for their job
   Data engineer               70% |  57%   n=745
   Data scientist              68% |  57%   n=548
   Data or business analyst    76% |  49%   n=312
   AI/ML engineer              70% |  67%   n=634
   Developer, full-stack       74% |  50%   n=12,149
   Developer, back-end         69% |  48%   n=6,357

U3 'Is AI a threat to your current job?' (data roles)
   No 63% | I'm not sure 20% | Yes 17%   n=1,976
   did not                        say yes, a threat: 17% (n=817)
   learned AI tools for the job   say yes, a threat: 17% (n=1159)

What employers expect by 2030

The World Economic Forum's Future of Jobs Report 2025 estimates that workers can expect 39% of their existing skills to be transformed or become outdated between 2025 and 2030 (WEF: Future of Jobs 2025). That is lower than the 44% in its 2023 report, which the report suggests could be due to more workers completing training: 50%, up from 41%.

Its picture of a workforce of 100 people is the clearest guide to the upskilling vs reskilling question:

• 59 would need training by 2030.
• 29 could be upskilled in their current roles.
• 19 could be retrained and redeployed elsewhere in their organisation.
• 11 would be unlikely to receive the training they need, leaving their prospects at risk.

So most of the expected change is upskilling, and most reskilling happens inside the same employer. The real risk is falling into the group that gets no training at all.

What tech and data professionals actually do

Most keep learning, but less with experience. Among people who were not new to coding, 81% of those with up to five years of experience had learned a new technique or language in the past year, falling to 65% after 21 years. Experience can make learning feel less urgent just when skills start to age.

Data roles learn at similar rates to developers. Around 70% of data engineers, data scientists and AI/ML engineers had learned something new, and 76% of data or business analysts.

AI learning is now part of upskilling. 57% of data engineers and data scientists, and 67% of AI/ML engineers, said they had learned AI-enabled tools for their job or career in the past year. Only about one in six data professionals (17%) saw AI as a threat to their current job, and that share was identical whether or not they had learned AI tools. Learning these tools seems to be treated as part of the job, not as a defence.

When to upskill

Upskill when your current role still has a future and you like the work. Signs: your team is adopting new tools you are not yet using; senior people in your role use skills you lack; or your role's job postings now ask for something new, such as cloud platforms, dbt or AI-assisted analysis. Upskilling is cheaper, it keeps your income, and your employer will often pay for it.

Good upskilling targets for data roles are the next layer of your current stack: an analyst adding Python and statistics, a data engineer adding streaming or cost optimisation, a data scientist adding deployment and monitoring.

When to reskill

Reskill when the core of your role is shrinking, when you have stopped enjoying the work itself, or when the role you want needs skills your current job will never teach. Start with roles close to your current one, which how to make a career pivot without a pay cut measures with real skill data, and look first for reskilling inside your organisation, where the WEF expects most redeployment to happen.

A simple test

Ask three questions:

1. Will my current role exist in a similar form in five years? If yes, upskill.
2. Do I want to be doing a more advanced version of my current work in five years? If yes, upskill. If no, reskill.
3. Can I learn the new role's skills while still in my current job? If yes, reskill gradually; if not, plan the money and time before you start.

Making learning stick

• Learn for a project, not a certificate: use each new skill on real work within a month.
• Set a fixed weekly time, even two or three hours.
• Ask your employer: training budgets, internal projects and mentoring are the cheapest routes.
• Review yearly: compare your skills with current job postings for your next role.

For the long view, see how to plan a 5-year career in data; for a concrete reskilling example, see moving from testing to data engineering.

Practical AI skills for data and tech roles are covered in the Generative AI course.

Sources

• Future of Jobs Report 2025 (World Economic Forum) — World Economic Forum
• 2025 Stack Overflow Developer Survey — Stack Overflow