Staff training
Training in a small business usually looks like this: the new hire is handed a folder of playbooks, shown the job once during a shift, and everyone hopes it stuck. A month later it turns out it did not, and asking felt awkward.
The Training section turns your own documents into a working trainer: the system breaks playbooks and instructions down into processes, concepts and questions, hands them to the team in short sessions, brings back what was forgotten and shows you who can be put on a shift today.
Open: Admin panel → Training (admin.cenaly.com/learning). The section is visible by default — no extension to switch on. The document library, however, lives in the knowledge base, and that one is hidden by default: switch on the "Knowledge base" card in the app store if you have not yet.
What this is#
| Who | What they see |
|---|---|
| Owner, manager | the material build, the process map, a "practice yourself" mode and the team's progress. Tabs live in the address (?set=overview and so on): Overview, Practice, Staff, Processes |
| An employee with their own login | only their own training — the "My training" screen with the Session and My progress tabs. Source documents are out of reach: they see questions and explanations, not your library |
The split is automatic, there is nothing to configure: whoever signs in sees their own side. Employee logins are issued in the Employees section.
What you need to start#
| What | Why |
|---|---|
| At least one processed document in the knowledge base | there is nothing to teach from while the library is empty. The build button says so plainly: "Upload and process at least one document first" |
| A knowledge base collection (one shared one is fine) | material is built from the chosen collection — you can train on a narrow set of documents instead of the whole library |
| A collection with the "Training" purpose | required for the process map and scenario questions: edit the collection in the knowledge base and set that purpose |
| Logins for the team | without their own login a person gets neither sessions nor progress |
| Devices | nothing special: a session runs in the browser, a phone included |
Roles: the owner and the manager build material and watch progress; everyone with a login can train.
Step 1. Build the training material#
On the Overview tab press Build training materials. The build runs in the background and says what it is busy with: "Reading the knowledge base…" → "Extracting core concepts…" → "Grouping into topics…" → "Composing quizzes…" → "Finalising…". You can leave the page — pressing again during a build honestly answers "materials are already being built".
What comes out:
- Topics — one per knowledge area of your base, each with a summary;
- Concepts — terms and rules with a definition, examples and a link to the source document; if a concept is backed by a fragment that contains a diagram or a photo, the card gets that picture from the document;
- Quizzes — questions whose correct answer is a statement of the article, while the wrong options are drawn from its own "what not to say" and from neighbouring topics. That is why the questions test your playbook and not general ideas about politeness.
When documents change, a note appears above the material — "Your documents changed — rebuild to refresh the materials" — with a Rebuild button. The rebuild is careful with people: rephrasing a question resets nothing, changing the correct answer resets only the evidence for that question, and a changed rule of the playbook lands in the mandatory part of the employee's next session marked "changed".
Step 2. Processes: what exactly to teach#
The Processes tab is what makes this more than a pile of tests. Business processes are extracted from your documents: steps, branches, conditions, endings — each with a quote from the document.
| What is on the process card | Why |
|---|---|
| Status ready / draft / blocked | "blocked" means "cannot be taught, there are findings in the way" |
| The Description completeness, Backed by documents, Steps reachable and Can be checked gauges | you see what exactly your playbook is missing |
| Applies to | you confirm which workplaces need this process. Until you do, it is not assigned to anyone |
| Mentioned, but no procedure | the documents talk about it but never say what to do step by step: a list of holes worth filling |
| Findings: "blocking" / "needs fixing" / "worth a look" | split by severity, so you fix what breaks training first |
Next to it sits the document's training access: a document can be barred from producing questions and verbatim quotes, and then it stays a source of meaning without ending up in the trainer word for word.
Step 3. How an employee learns#
The employee opens Training and sees "My training".
- Session. A badge "Time to review: N" and a Start session button. A session is a server-side object: F5 or a closed tab bring them back to the same step, not to the beginning.
- My processes. A list of what the person is mastering, with statuses: "review needed", "a rule changed — needs a repeat", "theory verified", "theory needs refreshing".
- Time budget. Chips for 3, 5 and 10 minutes: the session adapts to how much time there is right now.
- Inside a session. A mini-map of the process areas, terms highlighted with a popover holding the definition and the quote (the next answer is then marked "with the reference open"), and after each answer an explanation: the correct option, the reasoning and the "why". The formats go beyond "pick an option": a decision in context, the next step, spot the violation, put steps in order by dragging, matching.
- A mistake is not drilled immediately. The question that was missed comes back later in the session or in the next one — memory works better that way than with an instant repeat.
- "Is this question ambiguous?" — if the employee thinks the verdict is wrong, they say so with one button. It is a note to the owner about the material and does not affect their result.
- My progress. Mastery per process, the streak, upcoming reviews and the Training history: past sessions with the date, the kind (practice / review / check), "N of M correct" and each mistake spelled out — which area, what the question actually said, "incorrect" or "partly correct", plus the marks "with the reference open" and "answered it later unaided". The correct answer is deliberately absent from the history: that question will come back as a review, and a list with keys would turn the repeat into recognising a string.
- "I know it — test me". An experienced employee does not have to walk a process from the start: the button hands out one of the hardest questions per area. Answered correctly on the first try and unaided — the area is confirmed and moves to a long interval; missed — the area goes into the normal route. There is no instant retry in this mode, and the screen says so up front: this is a check, not a lesson.
Step 4. Exam, practice and clearance#
Once enough theory has accumulated, Check theory appears on the process.
| What | How it works |
|---|---|
| The ticket | assembled from the process, with mandatory coverage of critical goals. There are no term popovers and no instant verdicts in an exam — the breakdown arrives after you finish |
| Threshold and attempts | the screen names the pass threshold in per cent and the attempt number; the weekly attempt limit and the cooldown are shown as the date of the next possible attempt |
| Safety rule | if a question tests a rule you must not get wrong, per cent do not help: the attempt is not passed, and the screen explains why |
| Validity | a pass is valid until a specific date ("Theory verified until …"); a changed rule of the playbook annuls the pass for that process |
| Practice | a separate kind of task on the shift task board: it is closed by an observer, not by the learner, and that closure becomes the proof of practice |
| Clearance | theory and practice together; for a critical process — plus the confirmation of the process owner. The clearance runs out with the shorter of the two halves |
Clearance is not decoration: a task on the board can carry a gate "requires a qualification in process X", and without valid clearance such a task will not close.
Step 5. Control: who is ready for the shift#
The Staff tab answers one question: who can be put on a process today.
- The "processes × team" matrix. The cell where an employee meets a process is a traffic light: ready to be checked / theory verified / waiting for the process owner / expired / nothing to learn from / not started. Two scales switch next to it: Retention — the forecast of whether the person would recall it right now (it fades by itself over time) — and Demonstrated: the fact, areas walked unaided and without hints (it does not fade).
- Needs a decision. Only what requires action: "N employees cannot be put on the critical process X", "N have clearance expiring within 14 days", "N have it expired". Next to it — Assign training: a track, a due date, who to — and the people get a message in the work chat.
- Questions in doubt. Three sources in one list: staff complaints, statistics (first-attempt error rate and "discrimination" — whether the people who miss it are exactly those who master the neighbouring areas) and the build's own objections. Findings are split: "The build refused to serve these" — employees never see them at all, fixed by editing the document; "Worth a look" — the question works, the build merely has a doubt.
- Schedule calibration — a check on the planner: "we promised recall with a probability of 0.85 — it happened in 0.71 of cases". It sits at the bottom of the tab because it is not today's work.
- Clearance and readiness on the Overview — the same facts on the first screen. If there is nothing to decide, the block is not there at all: an "all good" banner trains people to skip that spot.
- Emails. Once a week the owner gets a digest: what is expiring, what expired, what waits for their confirmation and which critical processes have nobody ready. Nothing to write about — no email. Before a shift the employee gets a reminder like "a shift in 3 hours — 4 questions, 5 minutes, 1 of them mandatory".
Limitations#
- Training grows out of the knowledge base. No documents, no material; a weak playbook yields weak questions. Scenario formats and the exam work where a process could be extracted from the documents; where there are no processes, sessions run on topics.
- The practice half of clearance depends on a platform setting. Until the signing of observer receipts is enabled in your cloud, a closed practice task is not counted: theory still counts, clearance stays incomplete (training itself works as usual). If practice is being closed and clearance is still not issued — write to support, it is a setting on our side.
- Nothing runs "on a timer". Recalculating validity, processing practice receipts and reminders happen when a person opens training (and some of it when the owner opens the statistics). An employee who never opens the section will not get an email about their expired clearance — but you will see it in the matrix.
- The pre-shift reminder needs a shift schedule. No schedule, no reminder: we do not write by the calendar. A moved shift moves the reminder with it.
- The matrix shows the first 120 employees and says so honestly; the "questions in doubt" statistics is likewise computed over the first 120 progress files.
- Calibration fills up gradually and only from fresh answers; a bucket with fewer than 30 answers shows "not enough data" instead of a zero.
- The clearance gate works only on the shift task board. Actions at the till and in the schedule do not check clearance.
- There are no per-person deadlines per process yet — assignments live as topic tracks, so "process X by Friday" cannot be set.
- There is no dedicated till onboarding. Teaching work at the till is possible only through your own documents about it.
- Improvements to the questions themselves arrive with a rebuild. Until you press "Rebuild", training runs on the previous build of the material.
- Heuristic checks err both ways. "Two options say the same thing" and "this option came from another process" are hints to the owner, not a verdict: they do not block the question.
Troubleshooting#
| Symptom | What to check |
|---|---|
| The build button says "Upload and process at least one document first" | there is no processed document in the knowledge base library — upload a playbook and wait for processing |
| "No collection is marked as a training one yet" | open the knowledge base, edit a collection and give it the "Training" purpose |
| "No processes were found in these documents" | the documents describe facts and rules, not an order of actions. Add a playbook or a step-by-step instruction and build again |
| The employee sees "Training materials have not been built yet" | the owner builds the material — press "Build training materials" on the Overview |
| "No processes are assigned to you yet" | the process card does not confirm which workplaces it applies to |
| Answers were not counted, "the material was rebuilt" | the build changed while the session was running — start a new one, the questions will come from the fresh build |
| "There is nothing to build the check from" | some goals of the process have no questions yet; the owner sees this in "Questions in doubt" and fixes it in the document |
| A board task will not close: "a qualification is required" | the assignee has no valid clearance for the process — assign training or remove the gate from the task |
| A question looks wrong | the employee presses "Is this question ambiguous?", and the owner sees it in "Questions in doubt" together with the numbers about the question itself |
A symptom-by-symptom guide is in troubleshooting; if nothing helped, write to us — how to get help.
FAQ#
Do I have to switch on the knowledge base to train the team?#
The Training section is visible right away, but the material comes from documents, and the document library lives in the knowledge base — so its app store card does have to be switched on.
Will an employee see my documents?#
No. An employee only has their own training: questions, explanations and progress. The library, the collections and the process map stay with the owner and the manager.
Where do the questions come from — does the AI invent them?#
No. Every question is assembled from an article of your collection: the correct answer is a statement of the article, the wrong options come from its own "what not to say" and from neighbouring topics, and the question carries a link to the source document.
How long does a session take?#
As long as the employee chooses: chips for 3, 5 or 10 minutes. Questions are selected to fit that budget.
How is "theory" different from "clearance"?#
Theory is a passed exam on the process. Clearance is theory plus practice closed by an observer on the shift, and for a critical process plus the confirmation of the process owner. Put people on a critical process by clearance, not by a test score.
What happens if an employee simply never opens training?#
Nothing breaks, but nothing moves either: their clearance will not renew itself. In the "processes × team" matrix they show up as "expired" or "not started", and "Needs a decision" will list them under the critical process.
Can we train in several languages?#
Yes: material is built in the language of your documents, and the training interface is translated. Scenario formats are available for Russian, English, Georgian and Turkish.
We are a chain — can we train differently in different locations?#
A process is assigned to workplaces rather than to locations, and training is shared across the account. You can separate location tracks by keeping different knowledge base collections for them; how a chain is wired in general is in managing a chain.
Related articles#
- Knowledge base — the document library, collections and their purposes: the source of all training material
- Employees and shifts — logins, roles and workplaces: who training is assigned to
- Shift task board — the practice task under an observer and the clearance gate on a task
- Team chat and guests — where training assignments, pre-shift reminders and changed-rule notices arrive
- App store — where the knowledge base is switched on
- AI models and your own key — which model builds the material and the questions
- Subscription — balance and AI usage spent on building material
- How to get help — if clearance is not issued or a build keeps failing