Every vendor deck now says "personalised learning powered by AI," and the phrase has stopped meaning anything at the exact moment buyers have real money to spend. The adaptive learning software market reaches roughly $2.97 billion in 2026, according to Precedence Research, growing near 17 percent a year, with K-12 the single largest segment.
This guide is for school admins, EdTech founders, and university L&D directors comparing options. You will get the adaptive-versus-personalised distinction vendors blur, an honest DreamBox versus ALEKS head-to-head, a 2026 cost comparison, an evaluation sequence, and a clear-eyed look at when a custom AI personalised learning platform beats both. Published 4 June 2026. Last updated 4 June 2026.
- Adaptive learning adjusts difficulty inside fixed vendor content. Personalised learning adapts content, pace, language, and goals around the learner. Most products sell the second and deliver the first.
- DreamBox is the K-8 math and reading pick (school pricing around $20 per student per year). ALEKS is the mastery-based math engine through higher ed ($19.95 per month individually).
- Both platforms personalise only within their own content libraries. Your curriculum, your language mix, and your data-residency rules sit outside their walls.
- A custom build makes sense when two or more constraints (curriculum, language, data rules, product ownership) rule the licensed tools out, typically landing in the $20,000 to $100,000 build range.
Adaptive vs personalised learning: the distinction vendors blur
The two terms get used interchangeably in sales decks, and the difference decides whether a product fits your need. Adaptive learning is an algorithm adjusting difficulty and sequence within a fixed content library. The student struggles with fractions, the system serves easier fraction problems. Useful, proven, and narrow.
Personalised learning is the broader promise: the system adapts what is taught, how it is presented, in which language, at what pace, and toward whose goals. That requires a learner profile that spans subjects, content that can be generated or recombined, and teacher controls over the path. Almost no licensed product delivers all of it, because their business model depends on their own content library.
Keep that frame as we compare the two licensed leaders, because their strengths and their ceilings both come from the same place: ownership of the content.
DreamBox vs ALEKS: what each AI personalised learning platform actually delivers
DreamBox (by Discovery Education) is adaptive K-8 math and reading courseware. Lessons adjust in real time to student responses, and the program is rated "Strong" by Evidence for ESSA with third-party validation across grade levels. School pricing reportedly starts around $20 per student per year, with home plans from $12.95 per month. Its ceiling: it ends at grade 8, it teaches its own curriculum sequence, and personalisation means lesson choice inside DreamBox content.
ALEKS (McGraw Hill) takes a different approach called knowledge space theory: it maps exactly which topics a student has mastered and serves only what they are ready to learn next. It runs from K-12 through higher education in math, chemistry, statistics, and accounting, and individual subscriptions run $19.95 per month or $179.95 per year, with school and district pricing quoted per contract. Its ceiling is the same shape: mastery within ALEKS content, with a utilitarian interface younger students find dry.
The honest head-to-head: for elementary math intervention, DreamBox wins on engagement and price. For mastery-based math placement and progression through high school and college, ALEKS wins on rigor. Neither personalises across your whole curriculum, neither speaks your second language of instruction well, and neither hands you the learner data model. Those three gaps define the custom conversation later in this guide.
One implementation reality both vendors underplay: usage decays after the launch term. The pattern shows up consistently in buyer reviews on G2 and TrustRadius, where teachers praise the first semester and report drift once the novelty fades and dashboard check-ins stop. Whichever platform you pick, assign a named owner for weekly usage review in the contract period. An adaptive engine nobody opens adapts nothing, and the per-student fee bills either way.
Third Rock Techkno helps education teams run structured pilots and scope what licensed tools can and cannot cover. Talk to our education team →
AI personalised learning platform cost comparison 2026
Price models differ so much by category that per-student comparisons mislead unless you anchor them to scale. Here are the 2026 numbers side by side, sourced, with the custom-build bands included for the same budget conversation.
The arithmetic worth running: licensed tools price per student per year, forever. A custom build is a one-time cost (plus 15 to 20 percent annual maintenance, in our delivery experience at Third Rock Techkno) and the platform is yours. At 2,000 students, an ALEKS-class per-seat spend can cross a professional custom build's cost in two to three years. Below a few hundred learners, licensing wins on pure economics almost every time. The crossover point, not the sticker price, is the number that should drive the decision.
"Students learn significantly more in less time with AI-based tutoring systems than with in-class active learning."— Finding reported in Nature Scientific Reports research, summarised by the Brookings Institution review of AI tutoring evidence, 2024
How K-12 districts and L&D teams should run this evaluation
Buying personalised learning software for K-12 school districts or a university L&D program fails most often at process, not product. Run the sequence below before any contract, and insist every vendor answers in writing.
For the wider integration picture (rostering, SIS sync, and the workflow plumbing around any learning platform), Third Rock Techkno's guide to the LMS features worth demanding pairs well with this checklist.
We turn this sequence into a scored vendor comparison for your specific curriculum and constraints, in about two weeks. Book an evaluation sprint →
When neither fits: the custom AI personalised learning platform option
Licensed adaptive tools should win most single-subject evaluations. The custom conversation starts when the constraint is structural. In our platform work at Third Rock Techkno, four triggers come up repeatedly: a curriculum no vendor library matches, a language of instruction the tools handle poorly, data-residency rules that forbid the vendor's cloud, and EdTech founders for whom the platform is the product, where licensing someone else's engine caps the company's value.
What a custom build looks like in practice: a learner-profile model you own, AI generation grounded in your curriculum content rather than a vendor library, teacher dashboards shaped to your workflows, and integration with your existing systems.
That is the approach behind Learnly AI, which turns a school's own textbooks and PDFs into lessons, assessments, and tutoring material, and the platform builds delivered through the custom learning platform practice. The build lands in the professional to enterprise band ($20,000 to $100,000 one-time per the 2026 benchmarks above), with India-based engineering keeping rates at roughly a third of comparable US agencies.
For EdTech founders the calculus is different again. If personalisation is your product's core claim, renting the engine from a vendor puts your differentiation in someone else's roadmap and your margin in their price list. Founders we work with usually license third-party content where it is commoditised and build the personalisation layer (the learner model, the path logic, the teacher controls) as owned IP, because that layer is what acquirers and investors actually price.
The honest counterweight: a custom platform has no Evidence for ESSA rating on day one, and you carry the efficacy burden yourself. The right pattern for most buyers is hybrid. License DreamBox or ALEKS for the subject gap they demonstrably close, and build custom only for the parts of personalisation no vendor sells: your curriculum, your languages, your learner data model.
Run the pilot before the purchase order
Strip the branding and the decision is short. Name your gap in one sentence. If a proven adaptive tool covers it, pilot DreamBox for K-8 math and reading or ALEKS for mastery-based math, measure one cohort for one term, and buy on the evidence.
If two or more structural constraints (curriculum, language, data rules, ownership) rule the licensed tools out, price the custom AI personalised learning platform route against your multi-year per-seat spend and let the crossover point decide. The buyers who regret this purchase are the ones who bought the word "personalised" without checking what, exactly, gets personalised.
