Both of us sell agents. Only one of us lets them act on their own.
Absorb's pitch for agentic AI is work that completes without an administrator opening a queue. That is the right trade for a support backlog and the wrong one for releasing a grade. Lurno's agents do the same work and stop — a person applies it, and the record keeps what was proposed beside what was accepted.
Should you choose Absorb or Lurno?
Absorb is a capable corporate LMS that has put agentic AI at the centre of its 2026 positioning, and for an enterprise L&D team with a large routine-support load it is a sound choice — automation that closes loops unattended is exactly what that volume needs. Lurno suits an institution where being wrong is expensive: a school group, a publisher, a regulated employer, an academy training client companies. The agents do the same work — drafting a course from your documents, marking against your rubric, assembling a report — and none of them can apply its own output. That is architecture rather than a setting: there is no path from an agent to a released grade or a published course. Absorb publishes no pricing; its pricing page asks for your learner count and contact details before any number appears. Lurno publishes four tiers, from $119 a month, with the member and sub-organisation caps beside each. SCORM, xAPI and LTI 1.3 runtimes are in development at Lurno, so plan around that if your library is already packaged.
- Agents that propose, versus agents that act
- An agent that acts completes a task and applies the result — a course is published, a grade is released — and you find out afterwards. An agent that proposes completes the same task and stops; what it produced arrives as a draft or a queued item, and a named person applies it or does not. The work is identical. The difference is whether a model error is a rejected draft or an incident.
What decides this one
The rows that matter once someone has to sign a paper saying what the system may do unattended.
| Capability | Lurno | Absorb LMS |
|---|---|---|
| What an agent may do alone | Draft, suggest, assemble — never apply. No agent can release a grade or publish a course | Positioned around resolving routine work without an administrator opening a queue |
| Is approval a policy or a wall? | A wall — there is no path from an agent to the live object | Automation is the product; ask where the checkpoints are and whether the model decides when to pause |
| What the record keeps | What was proposed, what was applied, by whom, and what was rejected | Ask to see whether rejected proposals are retained |
| Published pricing | Four tiers, on the [pricing page](/pricing) | None published |
| How you get a number | Read it before the first call | Choose a learner band on a form, then a sales conversation |
| Entry point | $119 a month, up to 1,000 members | Quote only |
| Separating organisations | Multi-tenant to the root — nested sub-organisations, each with its own domain, branding, roles and administrators | Sub-portals, designed for an enterprise distributing to divisions or resellers |
| Guardians and parents | Built in, scoped by data category | Not a case the product is built around |
| Competency frameworks | Built in, with evidence and automatic mastery | Skills features present; check depth against your framework |
| Scoped for | Institutions — school groups, publishers, academies training client companies | Corporate L&D |
| SCORM, xAPI, LTI | In development | Supported |
Three differences you would notice in the first month
A suggested grade is not a released grade
Lurno's agent reads a submission against your rubric and writes out its reasoning, anonymised. What it produces is a suggestion sitting in a queue. An instructor opens it, agrees or does not, and releases the result under their own name. Nothing reaches a learner otherwise.
- The person who releases is recorded, and can be asked why
- A rejected suggestion stays in the record — that is the entry proving the control was doing something
- The learner never sees a score no one accepted
Separate organisations, not separate audiences
Sub-portals separate who sees what inside one organisation. Multi-tenancy separates the organisations. If you run fifteen client companies or forty campuses, each needs its own administrators, its own domain and its own reporting boundary — and the parent still needs a view across all of them.
- Nested sub-organisations, each with branding and a custom domain
- Administrators scoped to their own organisation
- One account, one contract, one place to look
You can budget before you talk to anyone
Absorb's pricing page asks for your learner count and contact details, then routes you to sales. Lurno's four tiers are published with the caps next to them, so an evaluation can start with a spreadsheet instead of a discovery call.
- $119, $299 and $599 a month, plus a quoted Institution tier
- Member and sub-organisation caps listed against each tier
- No implementation project — your own administrators configure it
Where Absorb is the better choice.
If most of your AI workload is routine support — enrolment questions, certification chasing, password-adjacent admin — then autonomy earns its keep and an approval step is just a slower product. Those questions are repetitive, the answers come from approved content, and being wrong costs somebody a second email. Absorb has built for that and reports real deflection from it. Absorb is also further along on SCORM, xAPI and LTI, which matters if your library is already packaged; ours are in development. And if you are one enterprise distributing to your own divisions rather than an operator running separate organisations, sub-portals may be all the separation you need. The case for Lurno is narrower on purpose: it is about the handful of surfaces where being wrong is expensive.
What you get either way
Agents grounded in your documents
Upload the source material and every draft cites where it came from, so a reviewer can check the claim rather than trust it.
Ask your data in plain language
Describe the report you want. The agent writes the definition and never reads your data — the query runs through the same permission checks as one built by hand.
Guardians, scoped by data category
Parents and guardians see attendance without seeing wellbeing, or progress without seeing behaviour. Set per organisation.
Competency frameworks with evidence
Mastery backed by what the learner actually did, and stackable credentials anyone can verify on a public page.
Your brand, your domain
Each organisation runs on its own web address with its own branding, sign-in page and administrators.
An audit log that cannot be edited
Hash-chained, append-only. Access reviews, GDPR tooling and a FERPA addendum sit alongside it.
The rest of the shortlist
Before you book the call
Bring the surface where being wrong is expensive.
Show us the workflow you would not let run unattended, and we will show you exactly where the agent stops.