Insurance companies train their people on rules and procedures that regulators examine, and they need a record of who completed which training, and how each person did.
Syllabium is an enterprise learning management system (LMS) for regulated industries, made by Athantic AI LLC. Insurance technology is one of the ten regulated industries an organization in Syllabium can work in. Tuning for insurance, with AI prompts and a reflection risk screen written for insurance work, is planned. Everything else on this page is the platform itself, and an insurance organization uses all of it.
It gives an insurer one system to write its compliance courses, assign them to cohorts on a schedule, assess and grade them, and keep the record, with AI working only under the authority an administrator grants.
| Step | In Syllabium |
|---|---|
| Write | Courses are written once in a versioned library in Syllabium Studio, and every change passes one publishing gate |
| Assign | Each cohort has a roster, a schedule and paced release, with a pre-course survey and prerequisites gating what follows |
| Assess | Knowledge checks, module quizzes and course finals, with pass marks set by policy |
| Follow up | A daily engagement check raises overdue lessons and unseen learners for a staff member to act on |
| Record | Learning activity, AI actions and publishing decisions land in tamper-evident records, which an administrator exports for an auditor |
The courses are your organization’s own, written by your team in Syllabium Studio, and each organization owns its content, whatever the subject. Syllabium is built to SCORM 1.2 and 2004, the long-standing format for course packages made in authoring tools, so packages your team already owns run beside the courses written in Studio, and their results land in the same learning record. Authors build lessons slide by slide on a canvas, compose curricula from library entries, and preview a slide as learners read it. Each slide can carry a facilitator’s note, which instructors see in the facilitator guide and learners never receive. Uploaded media reaches only a signed-in person entitled to it, and a file in use cannot be deleted. Courseware describes the studio in full.
Assessments draw their questions for each attempt, shuffle the options, limit attempts and apply a pass mark set by policy. A timed assessment saves answers as the learner goes and closes on time, and extra time can be granted as an accommodation. After a lesson, a quick check of two ungraded practice questions explains each answer, and one try opens the next lesson. Every question cites slide text word for word, and a verified question goes stale, and is no longer drawn, when a slide it cites changes. Blueprints set which questions an assessment draws, fixed or at random, and show a sample draw and a coverage report. Instructors see every attempt in the cohorts they staff, with a queue of what needs a person, and each attempt keeps the questions as asked and the pass mark applied. Flagging an integrity concern, clearing a flag, resetting an attempt and clearing it for credit each take a recorded reason, and none can be done on one’s own attempt.
AI agents in Syllabium act through the same recorded route as people, and an administrator sets what each one may do, capability by capability. A capability left unset is refused, and on content flagged safety-critical an AI agent’s work always waits for a person. The assistant beside the editors runs in the mode each author chooses, Off, Ask, Suggest or Act; it works on text only and does not publish. Every model call is logged with its model, tokens and cost, and an emergency stop at the front of the route can stop all AI, one agent or one kind of work. No person’s identity is sent to a model.
Syllabium uses Claude models, through Amazon Bedrock and Anthropic's API. An organization can also bring its own model (BYOM), and the answers of Syllabium’s agentic RAG, retrieval-augmented generation that looks up sources before it answers, are grounded in each field’s own library. Governed AI and AI review before publish describe the controls.
An auditor can check records in which any change shows: each entry carries a fingerprint of itself and of the entry before it, and once a day the newest fingerprints go to write-once storage outside the database.
| An auditor asks | Where the answer is |
|---|---|
| Who completed the training? | The learning record, kept as a tamper-evident chain, and each learner’s transcript as a CSV file |
| How are course versions controlled? | Each curriculum pins every entry at a version and is told when a newer one exists, learners stay on their version until the next is published, and versions are compared block by block |
| Did they pass? | Each attempt, with the questions as asked and the pass mark applied |
| Who changed the course, and who approved it? | The publishing gate’s decisions, which download as CSV, Markdown or PDF once closed |
| Who signed in, and who read a person’s record? | The sign-in record and the record of each reading |
| Has anything been altered? | Every chain and seal, recomputed nightly and on demand |
An administrator exports all of it as one archive for a period, with a verification report and a list of file fingerprints, and every export is logged. Most records are kept seven years, and a nightly purge records what it removes. Training records for audits and Evidence go further.
A company is a tenant, licensed for a number of organizations, and each organization owns its content and works in its own field. The database lets a session read only the platform’s shared rows and its own tenant’s and organization’s rows. Instructors and supervisors see only the cohorts they staff, and support staff read every cohort. AI settings exist at the platform, tenant and organization levels, and a lower level can only be stricter. Nobody edits, deletes or acts as an account above their own. Sign-in has email verification, lockout after failed attempts and an idle limit, with multi-factor sign-in and single sign-on through the organization’s own identity provider. Platform shows the three levels in full.
Insurance-specific tuning is planned: AI prompts and a reflection risk screen written for insurance work, as MedTech has for clinical training. A field is a setting of the organization, fixed when the organization is created, and it picks defaults, grants no access, and claims no compliance. Syllabium is built to ten standards, among them SOC 2 Type II, ISO/IEC 27001 and ISO/IEC 42001, all listed on Standards. Regulated industries lists every field.
Enroll your organization online, see the pricing, or talk to us. Syllabium is made and operated by Athantic AI LLC.
Contact us