Learnlytica/ The Readiness Report
Issue 02 · Weekly

The 90-day
reskilling lie.

Every consultant and platform is selling 90-day AI reskilling. The data says skill decay starts at day 45 without continuous practice. Either the program has a 12-month retention scaffold or you’re paying for theatre.

#02 / June 01, 2026 / 9 min read /
Previously, in Issue 01: Your CFO doesn’t believe your training worked, and she has a point. Completion is an activity. Readiness is an outcome. Read it →

Every vendor sells 90 days. The forgetting curve doesn’t care.

Open any enterprise AI reskilling proposal from the past twelve months and you will find the same number: 90 days. It is the length procurement approves without escalation. It is the sprint that fits inside a quarter. It is also, according to a growing body of evidence, a fiction.

Hermann Ebbinghaus demonstrated the forgetting curve in 1885. The principle has not aged out: without spaced retrieval, humans lose 50–80% of new knowledge within 30 days. Apply that to a 90-day AI reskilling bootcamp and the arithmetic is brutal. By the time the program wraps, the skills taught in week one are already degrading. By day 180, without reinforcement, the average learner retains less than 28% of the material.

The vendors know this. The 90-day model persists not because it works, but because it is what procurement will sign. A 12-month program requires a different budget line, a different approval chain, and a different relationship between L&D and the business. Most organizations are not set up for that conversation — yet.

But the organizations that are having it are seeing dramatically different results. The 12-month scaffold model — where initial intensive training is followed by monthly lab exercises, quarterly skill assessments, and continuous micro-reinforcements — retains 92% of skill at the six-month mark. The 90-day sprint retains 28%.

You cannot reskill someone in a two-week bootcamp. In 2026, the most successful programs are 6 to 12 months long. impress.ai workforce intelligence report, April 2026

The cost difference is smaller than most L&D leaders assume. A 90-day sprint that produces 28% retention and requires re-training is more expensive per competent employee than a 12-month scaffold that produces 92% retention once. The real question is whether your organization measures cost per seat or cost per ready employee.

If you measure cost per seat, the 90-day sprint wins every time. If you measure cost per ready employee, it never does.

The same skill, two delivery models
Head to Head
90-Day Sprint (Theatre)

Fast, forgettable

  • Fits inside one quarter
  • Procurement-friendly pricing
  • 28% skill retention at day 180
  • No post-program reinforcement
  • Completion = success metric
  • Re-training required within 6 months
12-Month Scaffold (What works)

Slow, durable

  • Intensive phase + 9-month retention arc
  • Monthly labs, quarterly assessments
  • 92% skill retention at day 180
  • Spaced retrieval built into cadence
  • Readiness = success metric
  • One program cycle per skill
Figure 01

Ebbinghaus Decay Curve — applied to enterprise reskilling

0% 25% 50% 75% 100% Day 0 Day 45 Day 90 Day 180 Day 365 -50% by day 90 -72% by day 180 92% retained Without scaffold With 12-month scaffold

Source: Composite from Ebbinghaus (1885), updated with enterprise cohort data from impress.ai and internal Learnlytica benchmarks. Scaffold includes spaced retrieval, monthly labs, and quarterly skill assessments.

The Readiness Paradox
82%
of enterprise leaders say they provide AI training. 59% still report a persistent AI skills gap.
Source: DataCamp State of AI Literacy 2026
VK
Vikram K.
CHRO, Top-5 Indian IT Services (220K+ engineers)
We killed our 90-day Databricks sprint in Q1. Replaced it with a 12-month program: 6-week intensive, then monthly lab challenges, quarterly proctored assessments, and a peer-teaching requirement. Retention at month six went from 31% to 89%. The CFO stopped asking why we spend money on training. She started asking how fast we can scale it.

Vikram’s team now measures one number per cohort: days from enrollment to first independent production deployment. The average dropped from 174 days to 68 days after switching to the scaffold model. The program costs 40% more per seat — but 62% less per production-ready engineer.

This Week’s Playbook

The “12-Month Test” — five questions to ask your vendor

Before you sign the next reskilling contract, run these five questions past the vendor. If they can’t answer all five, you’re buying a 90-day sprint dressed up as transformation.

  1. What is your day-180 proficiency rate? Not completion. Not satisfaction. What percentage of learners can demonstrate the skill independently six months after the program ends? If they don’t track it, that tells you everything.
  2. What happens after the intensive phase? Spaced retrieval? Monthly labs? Peer cohorts? If the answer is “they have access to the content,” you’re paying for a library card, not reskilling.
  3. How do you measure skill decay? Do they re-assess at 90, 180, and 365 days? If not, they have no idea whether the program worked — and neither will you.
  4. What is your cost per production-ready employee? Not cost per seat. Not cost per completion. The number that matters is: how much did you spend per person who can now do the job?
  5. Can you show me a cohort that finished 12 months ago? Ask for retention data from a real cohort, not a pilot. If the program has only been running for 90 days, you are the pilot.

Four things we’re tracking this week

Vendor · Correlation One

Correlation One extending flagship programs to 9 months

After internal data showed sub-40% retention at day 180, Correlation One is restructuring its enterprise data-science programs from 14-week sprints to 9-month scaffolded cohorts. The market is starting to correct.

Research · BCG

BCG publishes “The Retention Cliff” study

New BCG research documents the “retention cliff” in enterprise AI training: skill proficiency drops 54% in the 60 days following a bootcamp with no follow-up. Programs with monthly reinforcement retained 4.1× more skill.

Regulation · EU AI Act

EU AI Act mandates ongoing competency verification

Under the EU AI Act, organizations deploying high-risk AI systems must demonstrate ongoing workforce competency — not just initial training. One-and-done 90-day programs will not meet the compliance bar.

India · NSDC

NSDC draft mandate: continuous assessment for AI credentials

NSDC’s draft framework for AI skill credentials requires re-certification at 6-month intervals. Vendors without longitudinal assessment infrastructure will not qualify as approved providers.

From the team at Learnlytica

If your vendor can’t show you day-180 proficiency, they’re selling theatre.

Learnlytica’s platform tracks skill retention at 30, 90, 180, and 365 days — with spaced labs, proctored assessments, and decay alerts built into every learning path. One dashboard. Real numbers. No theatre.

See day-180 data in action or explore the platform