Greenhat IMS template · DIY operations guide · Procedure 1
So you’ve built an IMS. Now what?
A services group stood the template up in a week, reloaded the clause maps, and then asked the question this guide answers: we have all the documents, what do we do now?
The planning was done before the structure existed. You worked the questionnaire with the people who make the calls; the agent roughed the system in from the transcript. What you have is a rough-in: every wall framed, nothing yet sharp. This procedure takes the operative parts one at a time and chisels each against how the firm actually works, in the structure that now exists. At the end of it the system is in use, and the records prove it.
This is the first procedure in the DIY operations guide: system start-up. Ongoing operation and external audit preparation follow as their own procedures. Nothing here prepares you for an external audit, and nothing here teaches you to run an internal audit or a management review; the template’s own procedures do that. This guide gets the system going.
Step one
Where you are
Take stock before you touch anything. The build did a great deal; it decided nothing.
01What the build gave you
docs/records/, every one empty. The template says empty folders are honest. They are also the gap this procedure closes.02How thorough you were
Tick what is true of your questionnaire.
All three ticked: the steps below are review and sharpening, and the depth is yours to set. Any unticked: the same steps are the planning you have not yet done. Do them with the same people in the room, budget the time, and expect each to take twice as long as it reads.
03What done looks like
At the end of the first quarter these records exist. Everything below is in service of this list.
main behind it.docs/records/acknowledgements/.docs/records/consultation/.docs/records/management-reviews/ and one in docs/records/internal-audits/.Plans are worthless, but planning is everything. Eisenhower said it about war and it holds for management systems. The firm that asked the question at the top of this page had every plan, and had done the planning too, in its questionnaire sessions. What it lacked was the join between the two: nobody had yet picked up a document and asked whether it described the firm. That join is this procedure.
Step two
Review and approve every document
The rough-in becomes a system here. Each document is read by the role that owns it against how the firm actually works, changed through the agent, reviewed by a second person and approved by a third. The record is the pull request and the merge.
01Put names against the five roles
The questionnaire gave the roles titles. The People register needs names. Every document carries three of them in its frontmatter: owner, reviewer and approver.
- Top Management
- Chairs the monthly review. Approves 20 of the 21 documents.
- Quality Manager
- Owns 13 documents. Approves the Management Review procedure.
- OHS Coordinator
- Owns 6 documents. Standing reviewer on most of the rest.
- Document Controller
- Owns Document & Record Control. Runs the acknowledgement campaigns.
- Worker Rep
- Reviews the Policy, Consultation, Incident Reporting and the Emergency Plan. Sits in the risk workshop.
One rule, and the build enforces it: a document’s owner and its reviewer can never be its approver. tools/check_docs.py fails the build if they are, and branch protection on main gives that teeth. Because Top Management approves 20 documents, the director cannot own or review those 20. The check requires two people; three is the honest minimum, one to own, one to review, one to approve. In a firm this size these are hats, not jobs. Decide who wears which now, and put the names in the People register.
02Split the set
Owners drive. Each owner takes their documents in the order below, two a week; at that pace the set is through in six weeks. The reviewer and approver for each are in its frontmatter and in the Document Index.
Order. The documents people live under first: the Policy, Incident Reporting, the Emergency Plan, Document & Record Control, Operational Control, Customer Requirements. Then the rest. Any document with worker_consultation_required: true needs a consultation record in docs/records/consultation/ linked from its pull request before it can be approved; the Policy and the OHS procedures are the usual ones.
03Learn the lifecycle once
Every owner, reviewer and approver reads the Document Lifecycle for Authors page, twenty minutes, and logs it on the Training Completion form. Those are the first rows in the Training Log. The Author, Reviewer & Approver Guide is the reference; keep it open. Reviewers and approvers can work entirely from GitHub in the browser. Anyone who prefers the editor installs the GitHub Pull Requests extension from DIY Step three.
04Review with the agent in VS Code
The owner opens the repository in VS Code, opens the document and reads it against how the firm works. Where it is wrong, instruct the agent; do not edit by hand. Name the section, say what it claims, say what you actually do:
Open the Operational Control & Service Delivery procedure. Section 3.2 says the job lead verifies the deliverable. Here the job lead drafts and a second senior verifies. Rewrite 3.2 to match. Resolve the [ORG-DECISION] at 3.4: job files live in SharePoint under Projects/<job number>. Keep status Draft, bump version 0.1 to 0.2, add a Revision History row.
Read the diff before you accept it. Every [ORG-DECISION] and every LATER left in the document is decided now, or assigned a name and a date in the pull request; do not approve a document around an open marker. A reviewer who wants changes rather than comments checks out the same branch and instructs the agent the same way.
05Push for review
When the document reads true, have the agent set status: Under Review, then push. Branches are named doc/<ID>-<slug>. The agent is barred from git writes by the repository’s CLAUDE.md; people push.
git checkout -b doc/ORG-011-service-delivery git add -A git commit -m "Review ORG-011: verification by a second senior; job file location" git push -u origin doc/ORG-011-service-delivery
On GitHub, open the pull request, complete its checklist and request the reviewer named in the frontmatter. The reviewer reads Files changed and either comments or approves. A reviewer’s approval finishes the review; it does not approve the document.
06Approve
The approver checks four things, then merges.
- Review
- The named reviewer has approved the pull request.
- Checklist
- Every item in the pull request template is ticked.
- Consultation
- If the frontmatter requires it, the record in
docs/records/consultation/is linked. - Checks
- The build checks pass.
Then the approver sets status: Approved, approved_date and next_review_date and merges to main. The merge commit is the approval record; the template writes no other. In the same change the Document Register row moves to Approved with its Version, Approved Date, Last Reviewed Date and Next Review Date. Ask the agent to do it over MCP, or do it by hand in Airtable:
Update the Document Register row for ORG-011: Status Approved, Version 0.2, Approved Date today, Last Reviewed Date today, Next Review Date twelve months out.
Pages rebuilds on the merge. The approved document is live behind the login, and the intranet now says something you decided.
07Run the campaign and the induction
Once the Policy and any other document with acknowledgement_required: true is Approved, the Document Controller opens the acknowledgement campaign. The automation is already wired: staff receive the form, rows land in Policy Acknowledgements, and a snapshot goes to docs/records/acknowledgements/. Do not run it on a Draft. People acknowledge decisions, not proposals.
In the same week every staff member works through the IMS Induction page, twenty to thirty minutes, and logs it on the Training Completion form. A 30-minute all-hands is optional. If you hold one, say where the intranet is, what the two forms are for, and end with one sentence that promises thinking rather than paperwork: next, we define our first offer.
The agent drafts and edits on instruction. It does not review, approve, merge or push. The build check stops an owner approving their own document. A document approved by the person who wrote it is not approved; it is published.
Step three
Choose the offer to define first
One service, one market, chosen by margin and recorded by decision.
01Why one offer
A management system exists to define quality and then deliver it. Quality cannot be defined for the firm as a whole. It is defined where a particular service meets a particular buyer, and it differs across that grid even when the service name stays the same.
The template is built on this. Product & Market Offer Definitions holds one subsection per offer, an offer being a service in a market, and every job draws its baseline requirements from there. Beyond the eight company-wide objectives, objectives exist only as a sleeve attached to an offer that a management review has selected. There is no place in the system to define quality in general, because there is no such thing.
The company-wide objectives, on time, on cost, to specification, are table stakes. Every competent firm claims them and every client assumes them. They do not tell a consultant what a good report for this client looks like, and they will not move your work from acceptable to chosen.
If you are not with me here, the rest of this procedure will not land. The long-form argument is in Every sector is a buyer. None of them is your market.
The Lamborghini name sits on tractors and on supercars. Put “reliable and distinctive” in the quality policy and it is true of both and defines quality for neither.
The farmer is buying hours. A tractor that starts every morning for twenty years, takes any implement, and can be fixed in the shed with parts the dealer keeps on the shelf is a good tractor. Distinctive is a paint colour. Reliability is the whole product.
The supercar buyer is buying something else. Reliability barely registers: the engine and the parts are expected to have a short life, the service bills are part of the deal, and a car that spends a month a year in the workshop is normal. What is being paid for is the sound, the shape, the numbers, and what it says when it pulls up. I have seen two cars burning on the side of the road in my life. Both were Ferraris. Nobody who buys one is surprised by that, and none of them buys a Corolla instead.
Two products under one name, two definitions of good, and the tractor procedure would fail the supercar buyer as surely as the supercar procedure would fail the farmer. Your services page has the same problem, one line at a time.
02Rank by gross profit
You cannot do this for every offer, so pick the one that pays. Pull the last twelve months of closed jobs from your job costing. The IMS keeps no jobs table by design; job files live in your own job-tracking arrangement. For each job: revenue less direct cost (hours at cost, subcontract, laboratory, travel). Group by service and market. Sort by the total. Revenue is the wrong column; a line that turns over the most and earns the least is the one to fix last, or drop.
Be honest about cost, and count the overruns. Many firms have a habit of not recording hours once a budget is blown; the timesheet stops, the job closes on budget, and the loss disappears. If that is your culture, fix it now: every hour goes against the job it was spent on, invoiced or not. Until the habit is fixed the ledger is dirty, and it flatters exactly the lines that hurt you. Read it with a qualitative eye before you rank: ask each job lead which jobs ran over and by how much, and adjust the numbers before they decide anything.
A fifth of the lines will carry most of the margin; Perry Marshall’s 80/20 Sales and Marketing is the best short introduction to why that is and what to do about it. Take the top one. If the data does not exist, take the director’s estimate for the last quarter, minute it as an estimate, and start capturing direct cost per job in the job system this week; the objectives in Step four need it.
03Find the painful tail
The same ledger shows the other tail: the lines that produce the overruns, the reworked reports and the bounced submissions. Some are fixable. The cheaper move for most is to stop chasing them: take them off the website, stop quoting them cold, and still do them when a good client asks. Note them in the minutes with the decision.
04Start the post-job feedback now
The objectives plan leaves the CW-1 feedback mechanism for you to decide. Decide it now, because it is the cheapest evidence the system will ever produce and it comes from live work. At every job close from this week:
- Two questions
- Ask the client what worked and what they would change. Log both in the Customer Feedback & Complaints register.
- Rework or bounce
- If the deliverable was reworked, or an assessor sent it back, the job lead raises a row in the Nonconformity & CAPA Register.
Five minutes a job. It feeds CW-1 to CW-3, the customer-feedback slice of the monthly review, and the nonconformity stream in Step seven.
05Record the decision
At the next management-review session, whichever slice is scheduled, table three items: the offer selected, the sleeve opened for it, and the process owner named. The procedure requires every session to end in decisions, so this one qualifies. Minutes go to docs/records/management-reviews/YYYY-MM-<slice>.md, a row goes in the Management Review Log, and the offer gets its row in the Product & Market Offers register. The objectives plan is explicit: without that recorded decision no sleeve exists, whatever the register says.
Step four
Ask the people who use it
Quality for this offer is whatever its users need it to be. Find out from them, not from the proposal template.
01Draw the job map
A job map is the sequence of things the customer is trying to get done before, during and after your deliverable: work out what they need, find who can do it, brief them, get the work, check it, use it, deal with what comes after. Draft it yourselves first, from how the work actually arrives and leaves. The pattern is Bettencourt and Ulwick’s customer-centred innovation map. Then let the agent take a pass:
Read the Operational Control procedure and the last three released reports for this offer. Draft a job map for the customer: the stages before we are engaged, during the job and after delivery, one line each, with what the customer is trying to achieve at each stage.
02Interview the customers and the regulator
Good customers in this market, not the loudest. Show them the map and ask them to break it. Then, stage by stage: what are you trying to get done here, what gets in the way, what does a good result look like, how well is it met today.
The questions come from Outcome-Driven Innovation and Jobs to be Done; Strategyn and Christensen explain the method better than this page can. Record the conversation.
The people who assess your reports. The first answer is always that it is all in the guideline. Push past it: what is confusing when it arrives, what do you check first, what would make it faster to assess. Where the agency has probity rules on meeting consultants, use its published engagement channels or an industry forum.
One side effect: assessors remember whose submissions are easy to assess.
Every user of the deliverable, not only the payer. Most environmental reports have two readers with two jobs. If the assessor is unhappy with the half they read, the client who paid for the other half will hear about it. Quality is both.
An erosion and sediment control plan is bought by a contractor and assessed by a council. The contractor tears the drawings out of the back, bins the rest and builds what the drawings show. The assessor reads the methodology, the lab results and the risk assessment, and barely looks at the drawings. Two readers, two definitions of good, one document. A plan that satisfies only the contractor gets sent back, and the contractor pays for the delay.
Tell people the conversation is recorded and why; consent rules differ by state. Strip names, sites and job numbers from the transcript before it goes anywhere near an agent, and check the tool’s data terms before you feed it a client’s material. The model-training setting is covered in DIY Step two. Keep the transcripts outside docs/ and add their folder to .gitignore, as you did with the questionnaire.
03Extract the outcomes
Give the agent the de-identified transcripts, the job map and the framework, and let it do the first pass:
Read the seven de-identified transcripts in interviews/ and the job map. For each stage, list every outcome the interviewees described, score importance and current satisfaction from what was said, and flag the outcomes that are important and unmet. Then draft the "Jobs & outcome statements" subsection for this offer in Product & Market Offer Definitions.
The agent is consistent across seven transcripts in a way a tired reader is not. It is not the judge. The process owner reads every scored line against what they heard in the room and changes what is wrong. What the regulator and the customers said they need also lands in the interested-parties table of Context, Interested Parties & Compliance Obligations, which until now held the template’s generic entries.
04Draft the sleeve
Three to five objectives for this offer, each aimed at an outcome that scored important and unmet. Each one carries:
- Measure
- What is counted. First-pass acceptance by the assessor; days from brief to release; rework hours.
- Baseline
- Where it sits now, from the last twelve months or an honest estimate marked as one.
- Target
- Where it should sit by the end of the year.
- Data source
- An existing register or log. The plan forbids inventing a new instrument to measure an objective.
- Owner role
- The process owner, usually.
- Review cadence
- Monthly or quarterly, at the objectives slice of the review.
An OH&S objective needs two more things before it goes Active: a Linked Risks reference to the register row it addresses, and a Consultation Ref to the record where workers had their say. Table the sleeve at the next session; it is adopted by minute and mirrored as rows in the Objectives & Targets register. Meanwhile the company-wide objectives are measured from the first job closed, not left as decoration.
Step five
Productise it
Write the offer down so that a consultant who has never done it can produce a conforming result, then run real jobs through it.
01Capture what you do now
Before you improve anything, record it. The spreadsheet everyone uses, the report skeleton that lives on someone’s desktop, the checks the senior does in their head before anything leaves. Have the person who does the job best walk through one recent job from brief to release, aloud, with the recorder running. That transcript is the raw material; the agent drafts from it and from nothing else.
02Write the offer
From the walkthrough transcript, the job map and the scored outcomes, complete the subsection for this offer in Product & Market Offer Definitions under its four headings: Offer and market; Jobs & outcome statements; How the offer satisfies them; Claims. Include the deliverables, verification steps, safety preconditions and compliance obligations. Draft the delivery-plan template and the start-of-job checklist, and add or update the Document Register rows.
Read it as the new starter would. Every statement under Claims must be one Top Management will confirm can be substantiated; the template requires that before it is published. Then the offer goes through the Step two sequence like any other document: branch, pull request, review, approval, merge. The Product & Market Offers register row points at it, and Customer Requirements & Contract Review draws every future quote for this offer from it.
03Build the deliverable set
For most offers the key deliverable is a report, and a consultant is only as good as what they start from. Build three things from the best of what the firm has already produced, so a new starter edits down from a full draft instead of writing up from a blank page.
- Template
- The report itself, with its structure fixed and its words in. Headings and standard sections assembled from the best three or four released reports, each marked as required or conditional by job type; direct text for the standard passages; the standard tables; and a placeholder for each figure stating what it must show. The consultant edits it down; nobody writes it up.
- Guideline
- What has to be done to get the data behind each section: the field and desk steps, the procedures they rely on (a water quality monitoring procedure for the handheld probe, say: leave it in the sample for at least sixty seconds and wait for the pH to stabilise, because an early reading overestimates), and the report-specific hints a senior knows and a junior does not.
- Workbook
- If the report needs standard calculations, a spreadsheet template: inputs in open cells, every other cell locked, a version in the filename and a change log on the first sheet. Checked once by a senior, not re-derived by every reviewer. Fewer errors, and review becomes a comparison.
The guideline is where the unofficial knowledge goes. One consultancy I worked at decided whether excavated soil could stay on site or had to go to landfill by a number that sat between the health and ecological investigation levels and appeared in no published guideline. It lived in an email from the then head of the EPA’s contaminated land group, and in the memory of the two people who had read it. That is exactly what the guideline holds, email attached. Otherwise it leaves with them.
The agent assembles the first cut from your own released reports and the walkthrough:
Read the last four released reports for this offer and the walkthrough transcript. Assemble a report template that keeps the best structure of the four, marks each section as required or conditional by job type, and carries the standard text, tables and figure placeholders. Draft a guideline of the field and desk steps behind each section, the procedures they rely on, and the checks a senior would make. Where the reports contradict each other, list the differences and stop; I will decide.
None of these is a controlled document in the register’s sense, so control them the same way anyway: one owner, a version number, a change log, and a pull request for every change. Name them from the offer’s subsection so Customer Requirements & Contract Review and the delivery plan can point at them.
Building this set used to cost a senior consultant a month of evenings, which is why most firms never did it. With an agent working from your own released reports it takes days. Articles to follow in Insights: building templates and guides from your existing reporting, automating first-draft production from them, and holding the firm’s knowledge in a locally hosted vector database so the agent applies in-house IP without it leaving the building.
04Train and record
Add a Training Curriculum row for the offer. Everyone who delivers it works through the procedure and the templates and logs it on the Training Completion form. Those Training Log rows are the competence evidence for this offer, and the reason a new hire can be put on it in their second week.
05Run the next three jobs through it
Do not wait for the procedure to be perfect. The next three jobs in this offer run on it: a job file opened per Operational Control, requirements taken from the offer baseline per Customer Requirements & Contract Review, the deliverable verified by someone other than its author and marked Released, and the two close-out actions from Step three. Records come from work, not from exercises.
The first job will find something wrong with the procedure. Fix it through the same branch and pull-request sequence. That first revision after a live job, with its merge commit, is the proof the system is alive.
06Extend along the job map
The map from Step four shows the stages above and below yours. The more of the customer’s job an offer covers, the harder you are to replace, and a small firm can go where a large one will not. If a partner, a laboratory, a driller or a subconsultant does one of those stages for you, they are part of the offer: a row in the Supplier & Contractor Register and an evaluation snapshot in docs/records/procurement-evaluations/, per Procurement & Contractor Management. What you have written down in this step is the firm’s knowledge of how to do the work. It used to leave when people did.
A regional geotechnical firm added a soils laboratory and two drill rigs to what had been a desk-based consultancy. Consultants who had been engaging them for the field work found it cheaper to have them do the drilling, the testing and the logs, then hand the whole thing over; before long the sensible move for a client was to buy the entire job from them. A marine consultancy replaced diver inspections with an underwater drone and removed the most dangerous task in the offer altogether. Neither firm was large. Both had asked what the customer was actually trying to get done.
Closer to the desk: one firm had a written procedure for geo-rectifying scanned plans. People fetched it every time, because fetching it was easier than remembering. That is what a procedure is for.
Step six
Run the risk workshop
The agent drafted the registers from your answers. Rating them is a decision, and decisions are made by people in a room.
01Why
Nobody expects to crash on the drive to site. On any one trip the seatbelt feels like a formality, and a driver deciding trip by trip would often skip it. Governments do not decide trip by trip. They hold the data across millions of kilometres and decades, they know the crashes will come, and so the belt is required on every trip. That is a risk assessment: the individual odds set aside, the population odds counted, a cheap control chosen against a certain event.
The workshop is where the firm reasons like the government rather than the driver, and the register is what it writes down: the event, how likely and how bad, and the control chosen against it. Your own history is too small a sample to reason from, so borrow the population: your peers have had the incidents you have not yet had, and the industry’s are on the public record.
The rarest events, good and bad, the ones Taleb calls black swans, stay invisible until someone sits down to think about them. The template’s matrix has no special box for them: a Rare likelihood against a Severe consequence lands in the Medium band, which means treat it where a proportionate treatment exists, otherwise accept it with the owner’s documented sign-off.
02Calibrate before you rate
The methodology leaves two decisions to the executive, and neither can be delegated to the agent. Make them at the top of the workshop and minute them.
- The scales
- What Rare, Unlikely, Possible, Likely and Almost certain mean in years for a firm your size, and what Insignificant to Severe mean in dollars, days and injuries.
- Risk appetite
- The default: a residual Medium is accepted with the row owner’s sign-off; a residual High or Extreme needs Top Management’s acceptance or a treatment. Confirm it or change it.
The decisions go into the Risk, Opportunity & Hazard Methodology through the Step two sequence; the OHS Coordinator owns it.
03Set it up
- Time
- Half a day, once. Shorter reviews follow at the cadence each band sets.
- Who
- As broad across staff as you can afford. The Worker Rep is required, not optional; the methodology says assessment is done in consultation with the workers who do the work. The people who are not executives will name the risks the executives have stopped seeing.
- On the table
- The agent’s draft Risk & Opportunity and Hazard registers, printed, and the methodology.
- Out the door
- Every row rated, owned, treated and dated. An attendance list.
04Work the register
Row by row, aloud. Each row leaves the room with:
- Owner Role
- One name, who will answer for it at review.
- Existing Controls
- What is already in place, honestly.
- Inherent, Residual
- Likelihood and consequence before and after those controls; the rating is computed.
- Acceptance
- Within criteria and accepted, or escalated. A human decision, recorded as one.
- Treatment
- What will be done, with its level on the hierarchy: eliminate, substitute, engineer, administer, protect. Administrative or PPE controls alone cannot close a High or Extreme residual.
- Next Review
- Set by band: Low yearly, Medium six-monthly, High quarterly, Extreme monthly until it is not.
Add what is missing. Close what is not real. Rate the opportunities as carefully as the risks; the good swans are on the same register. Touch every cell. The hierarchy of controls is the one in Safe Work Australia’s model Code of Practice, which is where an auditor will expect it to have come from.
05File it
The attendance list and the notes go to docs/records/consultation/; that record is the consultation evidence the OH&S objectives and several documents point at. A snapshot of both registers goes to docs/records/. Before the room empties, take thirty minutes with whoever handles compliance and walk the Compliance Obligations register: every row a real obligation from a real instrument, none invented.
The agent wrote the registers from what you said in the questionnaire and from what the template knows about firms like yours. It has never seen your site, met your staff or driven your roads. Treat every row as a question.
A change to state legislation that requires continuous monitoring on every licensed site is an opportunity that arrives overnight, and only the firm that has already written it down is ready. A contaminated-land investigation that misses the contaminant because the sampling plan was copied from the last job is a risk that has not happened yet, and would be severe if it did.
Not having happened yet is not a rating.
Step seven
Start the machine, then stop
The system improves itself only if things are fed into it. Three homes, one session a month, and a short list of what not to do yet.
01Feed the three homes
- Hazards, incidents
- Near misses, hazards and injuries go on the Hazard & Incident Report form and land in the Incident Register; investigation records in
docs/records/incidents/. Staff learn it on the Hazard and Incident Reporting page. - Nonconformities
- A defective deliverable, rework, a report sent back by an assessor, an audit finding. Whoever finds it raises the row directly in the Nonconformity & CAPA Register: Description, Correction, Root cause, Corrective action, and an Effectiveness Check before it can close. The Quality Manager confirms the check.
- Feedback, complaints
- The Customer Feedback & Complaints register, fed by the two close-out questions. A complaint that reveals a nonconformity is linked to a CAPA row and resolved there.
The first stream is safety and everyone knows it. The second and third are quality, and without them the first quarter yields safety records only. The job lead owns them.
02The monthly session
The Schedule already holds twelve dated slices, one a month, seeded at instantiation:
- 01
- Prior actions and system changes
- 02
- Quality objectives and KPIs
- 03
- Customer feedback and satisfaction
- 04
- OH&S performance
- 05
- Audit results
- 06
- Risk and opportunity effectiveness
- 07
- Compliance evaluation results
- 08
- Consultation and participation
- 09
- Objectives and sleeve review
- 10
- Nonconformity and corrective action
- 11
- Resources and external providers
- 12
- Improvement and strategic alignment
Top Management chairs; the Quality Manager and the OHS Coordinator attend; others are invited by slice. A standing calendar entry, forty-five minutes. Minutes go to docs/records/management-reviews/YYYY-MM-<slice>.md and a row to the Management Review Log. The procedure’s rule: no session ends without explicit decisions and actions. Ask the agent for the pack the day before:
This month's review slice is Customer feedback & satisfaction. From the Schedule, the Customer Feedback & Complaints register, the Nonconformity & CAPA Register and the open actions in last month's minutes, draft the session pack and a minutes file with the decisions and actions sections left blank.
03Internal audit is a conformance check
The Schedule also holds twelve audit slices, and the Internal Audit procedure says how each is run; records go to docs/records/internal-audits/. One thing about them belongs here. An audit checks whether people do what the procedure says. A procedure nobody follows has no value, and the usual reason nobody follows it is that remembering is easier than fetching it. When an audit finds that, the fix is the procedure, not the people: make it shorter than the memory it replaces.
Anyone who has made spaghetti bolognese fifty times still opens the recipe, because the recipe is a hundred words and the outcome matters. Nobody opens a forty-page recipe. Write procedures people will fetch.
04Leave these alone this quarter
05Then
When the first offer is running on its procedure and its objectives have a quarter of numbers, go back to Step three and select the next one by a new recorded decision. With every document Approved, the registers rated in consultation and a quarter of records from live work, you hold the evidence an external auditor asks to see. When you book one, ask what they want to see running, and show them the list in Step one. The next procedure in this guide covers keeping it running.
Add the clause maps
All 104 obligations tied to the document, section and statement. One-time.
GuidedWe instruct, you build
Two set-up sessions, clause maps, registers built, a monthly check-in.
ManagedWe run it
A named Greenhat quality manager carries every duty above.
Documentation, not legal, safety or compliance advice. No ISO standard text is reproduced; you need licensed copies of the standards. Certification depends on your arrangements and your certification body’s assessment. The template is governed by its disclaimer and terms of use.