In short: People keep a second spreadsheet because the new system does not answer the question they are afraid of getting wrong. Ask that question, fix the gap, let them pilot it, and only then make the new path the default. Banning the file on day one hides the requirement you missed.
A new workflow that nobody uses is not a training failure until you have asked why the old path still answers a question the new one does not. Public posts ask the manager’s version of this: what to do when someone refuses to automate, how to handle pushback on a new HR system, how to get a team to use a new CRM. Status of those posters was not established. One of them, in a finance-flavoured forum, answered their own question by saying the person feared the automation would break and make them look bad. Treat that as one explanation to test, not as the motive in your office.
There is a local illustration, and it should stay small. On HardwareZone in June 2025, a poster who described themselves as in charge of process and projects said a colleague still keyed every invoice into the accounting system and into Excel, years after the system arrived. A reply suggested the system might not make totals easy, so the spreadsheet remained the working report. This article does not quote that thread. The point to steal is the question, not the tone: what does the sheet do that the screen does not?
The plan below is four weeks for one process. It is not a change-management certification and it does not promise that people will feel good about it.
Week 0: find the reason, with five questions
Sit with the person who keeps the second copy. Not a survey. A conversation you write down. Ask:
- What do you check in the spreadsheet that you cannot see in the system?
- What are you afraid will go wrong if you stop keeping it?
- What is slower now than it was before?
- When it has gone wrong, who fixed it, and how did you find out?
- What would you need on the screen before you would stop the sheet for a week?
Do not argue with the answers in the meeting. The fear of looking bad is useful information: it means the last failure had no owner. The missing total is useful information: it means a report was never built. Write both down. If several people keep sheets, ask each. Do not let the most fluent person answer for the others.
Job security will come up. Do not promise that a workflow will not change anyone’s job if you cannot keep that promise. You can say what this rollout will not do: it will not be used as a covert timing study, and it will not send customer or payment actions without the approval rules you already published. If you cannot say that, stop and fix the approval design first. A rollout of an unattended money movement is not a training problem.
Week 1: fix the gap the answers named
Build the smallest thing that replaces the reason, not a new platform.
- If the sheet exists to total a column the system hides, build that view. Do not add an automation that copies the same rows into a third place.
- If the sheet exists because errors have no owner, name the owner and the queue. The maintenance guide is the standard for “what we do when it is wrong”.
- If the sheet exists because the person does not trust a figure, show the source and the time it was taken.
Give the most sceptical person the exception role for the pilot: they log anything that would have made them reopen the sheet. That is a paraphrase of a practical move, not a quotation from a guru. It lasts for the pilot. It is not a permanent veto.
Training in this week is “what you do when it is wrong”, not a tour of buttons. A click-through that never shows a failure teaches people that failure is your problem, which is why they keep a private copy.
Week 2: pilot on their own work
Pick a slice they recognise: one customer, one week of invoices, one joiner, one inbox. Agree, in writing, what “it went wrong” means. Examples: a duplicate, a missing field, a total that does not match the sheet they still hold for comparison. During the pilot the sheet is allowed. It is the check, not the system of record.
The person who does the work runs the new path. You watch one failure with them and use the runbook. If you cannot produce a failure on purpose, you do not know the path well enough to ask them to trust it. The nine signs include “you cannot tell when it is wrong”. If that sign is still true, go back. More training will not invent a definition of correct.
New starters should see the new path, not the legacy sheet, from their first day. The onboarding checklist is where that task belongs, so you do not teach the workaround to the next person while you are trying to retire it.
Week 3: make the new path the default
The manual path stays available and becomes an exception you log: who, why, and whether the system was actually wrong. The poster who feared looking bad had, in that same self-answer, the outline of a rule: the old way remains possible and is recorded. Use the rule. Do not announce a ban you will not enforce.
Customer-facing and payment steps do not become unattended in this week. Default means “this is how a new item is created”. It does not mean “this is how money moves”.
If usage of a CRM is the actual fight, the buying criteria sit in how to choose a CRM. This rollout assumes you have already chosen. It will not rescue a system that cannot hold the fields the team needs. That is a selection problem. Say so, and stop the rollout rather than blaming the team.
Week 4: read the log, then retire the parallel sheet
Read the exception log with the people who wrote it. Sort the rows:
| What the log shows | What you do |
|---|---|
| The system was wrong or could not show the figure | Fix the system. The resistance was right |
| The person did not know the click | One more session on that click, not a new tool |
| The person did not trust an owner to fix it | Keep the named owner, and show a closed exception |
| Habit, after the above are done | Set the retirement date |
Retire the shared working copy on that date. If history matters, keep a read-only archive and say how long, using the same thinking as the retention guide: a copy kept “just in case” with no purpose becomes a second system again. Do not announce a productivity percentage. Count open exceptions and whether the sheet was reopened. Those are the numbers you actually have.
Say this, not that
| Instead of | Say |
|---|---|
| “The spreadsheet is unprofessional.” | “I need to know which total you are rebuilding, and I will put that total on the screen.” |
| “Automation will free you up.” | “This path creates the record. You still decide the exceptions on this list.” |
| “Everyone else is fine with it.” | “I am asking you because your sheet is the one finance relies on when the system and the sheet disagree.” |
| “We will monitor who refuses.” | “We will log items that could not go through the new path, and why.” |
Monitoring keystrokes or reading personal mail is not part of this plan. If you later want to measure employee use of a system, that can involve personal data and employment rules. This article does not state those rules. Ask someone qualified before you build a surveillance report and call it adoption.
When you should lose the argument
If week 1 cannot produce the missing total, or the system cannot represent the exception the team actually hits, the rollout is the wrong project. Revisit when not to automate. Adding a workflow on top of a system people bypass will create a third place for the same fact. That is the parallel-running trap. The spreadsheet guide is for when the sheet itself is the system of record and you are deciding what to move. This page is for when you have already moved, and the sheet came back.
A feedback log with four columns
Keep this next to the exception role. One row per item that could not stay on the new path. If the log is empty, you have not looked. If it is huge, you are not ready for week 3.
| Date | Item | What they needed the sheet for | System wrong, training, or trust |
|---|---|---|---|
| Invoice or job id | The missing total, the missing field, the fear | One of the three, not “resistant” |
Review it on the Friday of week 2 before you announce a default. Add a column for “who closed it” in week 4. An unclosed row is the reason the sheet comes back. Do not add a column for speed. You are not timing people.
Scripts you can actually say, when the conversation stalls:
- “Show me the last time the sheet saved you. We will put that check in the new path or we will not switch.”
- “For two weeks the sheet is the referee. After that it is read-only if the referee had nothing to correct.”
- “I will not ask you to stop a check I cannot replace.”
Those sentences are editorial. They are not from the HardwareZone thread. They exist so the interview in week 0 has somewhere to go.
Illustrative scenario
Illustrative scenario, not a client and not a result. A 20-person distributor in Singapore has an accounts executive who keeps a parallel invoice sheet. Week 0: she says the system cannot show a customer’s open invoices with the delivery date on the same line, and she was blamed last year when a statement went out wrong. Week 1: that view is built, and she is asked to log every statement she would not send from it. Week 2: one customer’s invoices run both ways; two mismatches are the system, one is a delivery date nobody entered. Week 3: new invoices start in the system; the sheet is only for the log. Week 4: the log is short enough that she agrees a date to stop adding rows. The story stops there. It does not claim the sheet disappeared, or that statements got faster.
Limits
The four weeks are a sequence you can run, not a measured duration from a study. The forum posts are single voices, one of them an employee in Singapore describing one colleague, the others of unclear role. None of them is a portrait of your team.
To look at one sheet people will not drop, use the workflow service or WhatsApp +65 9071 3976 (chat).
Frequently asked questions
Should I just ban the spreadsheet?
Not on day one. A ban without a replacement view pushes the sheet into a personal laptop, where you cannot see it and they still need it. Set a date to retire the shared copy after the replacement has been used on real work. Keep a read-only archive if you still need history.
How long does adoption take?
This plan is four weeks for one process. That is a schedule, not a measured average. A process that is genuinely unfit will not be adopted on a timetable. Revisit the design instead of repeating the training.
What if a senior person is the blocker?
Interview them with the same five questions. Seniority does not change the method. It changes the cost of ignoring the answer. If they are right that the system cannot produce the figure they are judged on, fix the figure. If they are protecting a private workaround, the exception log in week three makes that visible without a confrontation in week one.
Should the sceptical person be punished or put in charge of exceptions?
Put them in charge of the exception log for the pilot. They already know how it fails. A public post framed as a manager’s question suggested fear of looking bad if the automation broke. Giving that person the log uses the fear as a job, for a limited period, instead of arguing with it.
