Almost every multi-site operator has one: a manual, a handbook, a procedure binder, somewhere between 120 and 300 pages, containing the answer to nearly every operational question. It is usually accurate. It is usually maintained by someone competent. And it is read, in full, by almost nobody who works a shift.
The instinct when that becomes obvious is to blame the manual: too long, too dry, too dense. That diagnosis leads to a rewrite, and the rewrite produces a slightly shorter manual with the same problem. The manual is not the failure. The failure is the assumption sitting on top of it, that handing someone a document and recording that they received it counts as instructing them.
Reading is not instruction, and two statutes say so plainly
This is not a pedagogical opinion. It is what the law in both countries actually describes.
Codex art. I.2-21 requires the employer to ensure that every worker receives sufficient and adapted training in relation to welfare at work, and it specifies what the training has to be aimed at: his workpost or function. Not the organisation, not the policy set, not the manual. The workpost or the function. It also names the triggers, at hiring, at a change of function, when new work equipment or new technology arrives, and requires the training to be repeated and adapted as risks evolve.
Arbowet art. 8 lid 2 says the same thing from the other direction: the employer ensures that workers receive instruction that is effective and adapted to their distinct tasks. Lid 4 then requires the employer to supervise compliance with the instructions given.
A 200-page manual is, by construction, not adapted to anyone's distinct task. It is adapted to the organisation. Every role in the building reads the same document, most of which does not apply to them, and the parts that do apply are separated by forty pages of material that does not. Distributing it is a perfectly reasonable act of documentation and a poor act of instruction, and an inspector reading either of those articles is asking about the second one.
Being precise here also protects you from over-claiming. Neither article says a manual is forbidden, neither says a course is required, and neither prescribes a format. They ask for a result: this person, in this job, has received instruction adapted to what they actually do. The manual is evidence that the content exists. It is not evidence that it arrived.
Sorting two hundred pages into three piles
Before converting anything, split the document. Almost every operations manual is three documents wearing one cover, and they have different audiences, different lifespans and different destinations.
Reference. Product and SKU lists, supplier contacts, equipment specifications, escalation numbers, org charts, opening hours by site. Nobody should be memorising these. They need to be findable in fifteen seconds by someone standing at a counter with a customer waiting.
Task procedure. The opening sequence, the cash handling protocol, the changeover, the cleaning cycle, the hygiene log, the equipment check. This is the part that is genuinely instruction material, and it is also the part with a legal duty attached, because these are the tasks the Codex calls the workpost and the Arbowet calls distinct tasks.
Policy. Codes of conduct, privacy, absence rules, disciplinary procedure. This is real content that everyone needs once and then needs to be able to find again.
The test for the middle pile is simple and worth writing down: does someone have to act on this without looking it up, in the moment, while something else is happening? If yes, it is instruction. If they can look it up, it is reference.
Why the reference pile has to survive the project
The most common way this project goes wrong is that reference material gets swept into courses because it is in the manual and the manual is being converted.
That produces the modules everyone skips: fourteen screens of product codes with a quiz at the end. Worse, it hides the information. A person who needs to know whether an item contains celery does not want a training module, they want an answer. Keep the reference layer searchable and separate, and let the training draw from it rather than swallow it.
This is also the honest answer to anyone defending the manual. You are not deleting it. You are demoting it to what it was always good at, and lifting out the twenty or thirty percent that has to be taught rather than looked up.
Cut by role before you cut by chapter
Once the task procedures are isolated, resist the urge to convert them chapter by chapter. Chapters are an artefact of how the manual was written, not of how the work is done.
Cut by role first. A front-of-house colleague and a back-of-house colleague draw from the same library and need different sequences of it. A shift supervisor needs a superset with different emphasis. Someone who covers two stations needs both paths, not a merged one.
Then order each path the way the work unfolds rather than the way the binder is numbered: what the person touches on their first shift, then the safety or hygiene rule attached to that task, then the next task. A rule delivered before the task it protects has no anchor, and it is the first thing to go when the site gets busy.
The failure this prevents is the one every operator recognises. Assign all twenty-five modules to every new hire and completion collapses, not because people are unwilling but because the volume is visibly unreasonable. Assign eight that are obviously about their job and they finish them.
The policy chapters, and the honest thing to say about them
Policy content is where handbooks and operations manuals overlap, and it is the part most likely to be handled by an attestation form in a shared drive: a signature saying the person has read a document they scrolled past.
Convert the behaviour-changing parts into short modules with knowledge checks, and leave the legalese as reference that is signed once. The distinction is the same one as before. What someone has to act on gets taught; what they have to have been told gets recorded and stored where they can find it again.
What conversion does not fix
Three things, worth saying before anyone expects too much.
Converting an inaccurate procedure produces an accurate-looking course teaching the wrong thing. The conversion is a good moment to have the procedure owner read it, because for many manuals it will be the first read-through in years.
Conversion does not discharge supervision. Arbowet art. 8 lid 4 asks whether compliance is being watched, and a completion record answers the instruction question, not the supervision question.
And conversion does not stop the manual from ageing. If the source document is updated once a year in a Word file and the courses are regenerated from it, the courses are exactly as current as that file.
Running the conversion itself
Upload the document as it stands. PDF, Word or PowerPoint, without cleaning it up first. Embedded tables, headers and figures are read as they are. What comes back is a parsed structure with chapters mapped to candidate modules, which you correct inline where a heading has been misread. This is also where the three piles become visible, usually faster than reading the manual would have been.
Generate the task procedures first. Each procedure becomes a course with knowledge checks and short scenarios rather than a chapter with a quiz bolted on. A long document does not become a long course, because what a frontliner has to act on is a fraction of what the document contains. Review before publishing.
Reorder for the shift, not for the chapter. Front-load what changes behaviour and demote what is genuinely reference. Modules run three to five minutes so they fit the gaps that exist in a real shift, and spaced retrieval brings the operationally critical points back rather than leaving them in the module where they first appeared.
Assign per role and retire the PDF. Each role gets its own path, with auto-assignment for future hires. Then remove the static document from the onboarding folder and replace it with the path. From that point, "has read the manual" means completed, dated, per person, with a knowledge check result behind it, and it exports in that form.
Start with the chapter your last incident touched
Pick the procedure involved in your most recent incident, complaint or audit finding, and bring it. In a demo we convert that one chapter and show you the record it produces, which is a faster argument than a discussion about the other 190 pages.