Watch the current steps
Note who reads the slip, who types the items and who checks them. Include the calls needed to clarify handwriting or find missing information.
From a task that takes too long to software your team can use.
Start by watching how the task is done today and agreeing what needs to improve. A process change, ordinary software and AI are all options. Try a small piece with the people doing the work before extending it.
Follow a task from its input to its result.
Start with work that staff repeat, such as taking an order from a paper slip. Find where time goes before choosing software or AI.
Try a smaller change
A clearer order form may remove the problem. If reading varied slips is still slow, trial a model that prepares the items for someone to check.
Compare the completed work
Time the job through to a checked order, including corrections. Keep the change only if staff can finish with less effort and an acceptable error rate.
Phase 1 of 5
Understand the task as it happens.
Watch the people doing the task, including the steps kept in notebooks or message threads. Record where information starts, who handles it and what a finished job looks like.
What gets mapped
List the people and tools involved in the chosen task. Mark repeated entry, waiting and checks, including work done outside the official process.
Where the mapping happens
On the floor and in the field as well as in meeting rooms, including the spreadsheets and message threads people keep beside the ERP. Those hold the steps the ERP leaves out, and a workaround of that kind usually turns out to be a requirement.
What comes out
Write down the current steps, the time or errors worth reducing and the result the team needs. Use that record to agree a starting scope and a way to judge it.
Phase 2 of 5
Choose the smallest useful change.
Decide whether the task needs a process change, a connection, a screen or a model. Agree the choice with the people doing the work, including its cost and the steps they will still carry out themselves.
The records needed
Identify the records the task needs and where they are kept. Resolve conflicting names or numbers where they affect the result.
The screen people use
Choose where the result belongs so staff can use it during the task, with access appropriate to their role.
Keep, replace or build
Keep useful tools and record the reason for anything new. A replacement needs its own case, including the effort of moving records and learning different steps.
Where AI belongs
Choose tasks where a model could reduce reading, typing or searching. Compare it with a rule or a process change, counting the time staff spend checking results. Agree what needs confirmation and how the task continues if the model is unavailable.
Phase 3 of 5
Build a piece the team can try.
Release a small change that a team can use in its daily work. Try it with real inputs, check the result and resolve problems before adding more scope.
Check the essentials first
Set access, validation and confirmation steps before the first trial. Apply fixed checks in ordinary software so they do not depend on a model interpreting instructions.
Released in working slices
Each release is something one team can start using, so nothing waits for a single launch date at the end.
Built with your people
The team that will use a screen works with it while it is being made, and their changes go in before it is released.
Designed for all-day use
Screens that staff use for a full shift get the same design care as the ones customers see.
Keep useful tools in place
Start with the tools and records the team already uses. Test any connection before relying on it, and run a proposed replacement alongside the original while its results are checked.
Document the working change
Record how the change works, what it depends on and how to recover from a problem. Keep its limits visible to the people using and maintaining it.
Phase 4 of 5
Connect only what the task needs.
Join the relevant steps so information reaches the place it is needed. A connection should remove repeated entry or a handoff; it does not require every tool to be replaced.
Systems you keep
Leave accounting, payroll and specialist software in place when they serve the business. Connect only the records needed for the agreed task.
Connections with outside services
Check the service’s access, cost and failure behaviour before making the task depend on it. Include the time needed to handle a delayed or missing response.
Check a handoff end to end
For an order import, check that the destination accepts the file, keeps the right items and handles a duplicate safely. Compare the result with the source before staff rely on the connection.
Measure the steps removed
Compare the repeated entry and checking before and after the connection. Count any new monitoring or correction work as part of the result.
Agree how staff will spot a failed transfer and complete the task while it is being fixed.
Phase 5 of 5
Review whether the change still helps.
As the business changes, review the task with the team. Extend the software where there is a clear need, and simplify or remove steps that no longer earn their place.
Changes in the team’s work
Check whether new users or tasks need a change to the existing screen before adding another place for people to work.
Review the model’s value
Recheck accuracy, running cost and the time spent correcting suggestions. More data alone is no reason to add another AI feature.
Keep the instructions current
Update the working instructions when the process changes, including permissions, manual checks and the fallback steps.
Support that knows the system
The people who maintain it are the people who designed it. Nothing is switched off until you have seen its replacement run.
What 64ARCS will not do.
01
No rip-and-replace mandate
Systems that still serve the business stay, and are connected instead of discarded.
02
No one-size-fits-all ERP
The structure follows how your business works. Your business is not asked to reshape itself around a product roadmap.
03
No AI for the sake of AI
AI needs to reduce effort or help someone make a better decision. Its running cost, mistakes and review time all count when assessing whether it belongs.
To start with a task that takes too much time, write to hello@64arcs.com or use the form. A person reads it and replies within two working days. The first call is sixty minutes on how your operation runs, no deck.