Field Note / April 23, 2026 / 12 min read

The AI Upskilling Market Hit $97 Billion. Most Programs Still Don't Work.

The market for AI upskilling is one of the fastest-growing categories in enterprise software. The outcomes engineering leaders actually want, AI-augmented code that holds up to enterprise review, are still not arriving. This is a reading of the 2026 reskilling research, and what we think it actually says.

Vitaliy Filipov
Vitaliy Filipov
Camplight

In the first quarter of 2025, more than 70% of US venture capital flowed into AI companies. Some of that money went into models. A lot of it went into the layer above the models: tools to help organizations and their people learn to use AI at work. That layer is now a real market.

The market, in six numbers
$97B
AI in L&D market by 2034 (from $9.3B in 2024)
26.4%
CAGR for AI L&D, 2024 to 2034
$1.12T
AI-in-the-workplace market by 2029
85M
Jobs displaced by AI by 2025 (WEF)
81%
Employers hiring on skills, vs 57% in 2022
~20%
Annual growth in AI skills demand (Bain)

The demand is real. The budgets are unlocking. The platforms are scaling. And yet, when we walk into engineering organizations in 2026, we still find the same pattern. Engineers are using AI. The output is faster. The output is also less reliable, less reviewable, and more expensive to maintain than what those same engineers were producing two years ago.

The market is solving the wrong problem.

Section 01

What the research is actually telling us

Strip the funding announcements out of the 2026 reskilling research and three signals stand out.

Signal 1

The talent gap is widening

Bain: AI-skills demand growing ~20% a year, supply structurally limited. Hiring alone cannot close it. Internal upskilling is the only path with the right shape.

Signal 2

Skills-based hiring is real

57% to 81% in three years. Internal promotions, reviews, and comp will follow. Cultural resistance to AI training is dissolving on its own.

Signal 3

"Platform plus content" wins by default

Sana, SkyHive, Guild, Degreed, LinkedIn Learning, Cornerstone, Eightfold, Gloat. Different shapes, same shape. Works for breadth, not for engineering depth.

Section 02

Why "platform plus content" underperforms in engineering

Three structural reasons, all visible in the research, all confirmed in the field.

A

The unit of upskilling is wrong

Platforms optimize "the course" or "the learning path." For engineering, the unit that matters is how this person makes a decision in front of a real codebase, with a real deadline, with an AI in the loop. That decision produces vibe code or production code. No quantity of adaptive content reaches it.

B

The feedback loop is too slow

"Real-time analytics" in practice means updated weekly, surfaced to HR monthly. For an engineering manager whose senior is suddenly merging AI-written PRs that fail in staging, weekly is not a feedback loop. The loop has to live inside the workflow on the day the work happens.

C

The integration story over-promises

Analysts have called HRIS / LMS integrations "onerous and brittle" for years. Every vendor still claims seamless. The reality engineering leaders describe: the platform sits in HR's domain, engineering tools sit in engineering's, and the bridge is a quarterly export no engineer opens.

Two operating models
Dimension
Platform + Content
Cognitive Rebase
Unit of change
A learning path
A decision in front of real code
Personalization
By role and industry
By the engineer's codebase + observed behavior
Feedback cadence
Weekly, monthly to HR
Per PR, in the workflow
Owner inside the org
L&D / HR
Engineering leadership
Primary metric
Course completions
Defects shipped, time-to-merge, PR quality
What it sells
Delivery + content library
Methodology + coaching against it
Section 03

What the research gets right

Three points in the 2026 research are worth taking seriously, and worth designing around.

They got it right
  • +Personalization is non-negotiable. Tailored content beats one-workshop-per-team.
  • +Community beats content alone. Mentorship and peer learning drive engagement.
  • +Outcomes, not completions. Buyers want measurable capability change.
Where we push back
  • Personalize to the engineer's codebase and observed behavior, not their role description. The real differentiator is what they ship and how they decide, not the title on their HRIS profile.
  • Mentorship has to live inside the workflow, not in a separate community tab.
  • The outcomes that count are the ones a CTO sees in code review, not in a quarterly dashboard.
Section 04

What we'd buy in 2026, if we were a CTO

Four questions to evaluate any upskilling vendor against. They are deliberately uncomfortable.

Q1

Whose voice does the coaching speak in?

Generic instructor content has a ceiling. Coaching that sounds like your own engineering leadership clears it. Ask whether the platform can clone the methodology onto your CTO's voice and posture, or only deliver someone else's.

Q2

Adapted to our codebase, or our industry?

"Built for fintech" is marketing. "Built around your services, your tooling, your style guide" is upskilling. If the vendor cannot show role-specific examples that reference patterns from your repo, the personalization claim is theatre.

Q3

Where does the feedback live?

If the only artifact is a quarterly skills dashboard for HR, the engineering manager has no surface to act on. Insist on weekly capability signals at the level of the individual engineer, surfaced where engineering leaders already look.

Q4

What is the methodology?

Most platforms sell delivery and treat content as interchangeable. Ask which behaviors the program installs, and how those map to defects shipped, time-to-merge, and AI-augmented PR quality. If the answer is "we adapt to whatever you want," there is no methodology.

Section 05

The honest summary

Correct

The market thesis is right. AI is reshaping work, the workforce has to be reskilled at a pace no university can match, and enterprises are the buyers. The category will keep compounding through the decade.

Incomplete

The response is incomplete. Most dollars are flowing into platforms that scale content delivery, integrate with HR, and produce dashboards. Necessary, not sufficient, for engineering teams who have to ship code that holds up with AI in the loop.

Next

The next wave of value will not come from a better LMS. It will come from programs that treat the engineer's working day as the unit of change, embed the methodology in that day, and report capability in terms a CTO can actually use. That is the gap we built Cognitive Rebase to close.

Key takeaways
  1. 1The market is real. AI in L&D is projected to grow from $9.3B (2024) to $97B (2034) at a 26.4% CAGR. The AI-in-the-workplace market hits $1.12T by 2029.
  2. 2The skills gap is structural. 81% of employers now hire for skills (up from 57% in 2022). Internal upskilling is the only path with the right shape.
  3. 3Platform-plus-content underperforms in engineering for three structural reasons: wrong unit of change, slow feedback loop, and brittle HRIS/LMS integrations.
  4. 4What works: personalize to the engineer's codebase (not their role), embed mentorship inside the workflow (not a community tab), and measure outcomes a CTO sees in code review (not in a quarterly dashboard).

Sources cited

  • · District Angels, "AI-Powered Upskilling and Workforce Transformation Market Report" (Catherine McMillan, 2025)
  • · market.us, "AI in Learning & Development Market" projections
  • · Business Research Company, "AI in Workplace Market Outlook"
  • · Virtasant, "AI in Corporate Training & Learning"
  • · Forbes / McKinsey, US workforce automation projections (2030)
  • · World Economic Forum, "Future of Jobs Report 2025"
  • · Bain & Company, "Widening talent gap threatens executives' AI ambitions"
  • · Cornerstone OnDemand, SkyHive acquisition disclosures
  • · Crunchbase News, Guild Education funding history
YOUR TEAM, IN PRACTICE

See where your engineers actually stand.

A 30-minute Fit Call. We'll walk through your current AI workflow, the behaviors driving your code quality, and what a Rebase would change.

Book a Fit Call