Your workflow works on the administrator’s laptop. A branch employee opens it on a phone, cannot find the assigned task and sends the request through the old chat group. The manager replies there. Two processes are now running, and neither gives the whole team a reliable view.
Before extending a Lark rollout, test whether workers can complete the agreed task on their actual devices and whether managers act on what arrives. Adoption depends on both sides. Training workers to submit information has little value if the response still happens somewhere they cannot follow.
Choose one process for the first group. Agree who owns it, where a valid request belongs and how exceptions will be handled. Use the checks below to decide whether the setup is ready for ordinary use.
Separate the worker’s task from the manager’s task
A worker may need to find a checklist, submit a request and check its status. A manager may need to review records, assign work and manage exceptions across outlets. The person configuring the system has a third set of tasks.
Do not use one successful demonstration as evidence that all three roles can work effectively. Lark’s frontline positioning addresses mobile workers, but the exact plan and task still need checking. Its metrics-block guidance, for example, distinguishes mobile viewing from adding or editing those dashboard blocks.
Build a role-and-device list before choosing licences. Record which actions require a phone, browser or desktop application. Ask the supplier to demonstrate any task whose device support or entitlement is unclear.
Check access at the point of work
Establish whether workers use personal phones, shared devices or company-issued devices. Confirm the organisation’s rules for work information on those devices before asking staff to install anything.
For personal phones, consider who can see notifications and what happens when employment ends. For shared devices, check how the next worker gets appropriate access without receiving the previous person’s information. Avoid shared sign-ins unless a supported, approved design addresses accountability and access.
Test the connection where the task happens. A staff area, loading point or shop floor may behave differently from headquarters. Ask the worker to submit a record, verify that it arrived and find the result. Include an interrupted connection and an uncertain submission. The agreed recovery route should prevent workers from repeatedly submitting the same request.
Offline behaviour is task-specific. Establish what works in the chosen configuration before relying on it in an operating procedure.
Make the instructions usable during a shift
A worker should be able to recognise the process, understand the fields and tell what happens after submission. Use the language and terminology the team uses for that work. Check labels with representative staff rather than translating a long headquarters procedure word for word.
For a multilingual team, have appropriate reviewers check the instructions and exception messages. Automatic translation, where available, does not settle the meaning of an approval rule or a safety instruction.
Consider text size, attachment handling and the time available at the point of work. Ask workers to explain the task back and complete it without the administrator taking over. A question they ask twice may identify an instruction that needs rewriting.
Provide a short route for help. Name the person who can resolve a process question and distinguish that from a technical access problem. Staff should not have to guess which group chat will receive a response.
Run a role-based acceptance check
Use a test sheet with the actual roles and devices. The examples below are proposed checks, not claims about a preconfigured Lark deployment.
| Check | What to observe |
|---|---|
| Join and sign in | A new worker reaches the correct task with the right account. |
| Submit and confirm | The worker can tell that the intended record arrived. |
| Return for correction | The worker understands what is missing and how to resubmit. |
| Manager review | The assigned manager finds the request and can make the permitted decision. |
| Shift handover | The next responsible person sees the outstanding action. |
| Owner absent | The exception reaches someone authorised to act. |
| Access changes | A role change or departure produces the intended access result. |
Record the result and any assistance required. Distinguish a confusing instruction from missing permissions, unsupported device behaviour or an unsettled business rule. Each has a different remedy.
Agree when the old route stops being valid
During testing, the team may still need the old process as a fallback. Before ordinary use begins, state which route counts for the pilot process and how a temporary failure is recorded.
If managers continue accepting informal requests elsewhere, workers receive conflicting instructions. A staff member who follows the new form should not have to repeat the request privately to get a response.
Keep the fallback specific. Define when it can be used, who records the outcome and how the main record is reconciled when the system is available again. Do not leave workers without a way to handle time-sensitive work while a technical issue is being resolved.
Decide whether to continue, change or pause
Look at actual use by role and shift. Count expected tasks, completed tasks, incomplete submissions, help requests and unresolved exceptions. Ask managers whether the records contain enough information to act. Ask workers whether they can find the result.
Set the acceptance criteria before the pilot. The team might require every intended role to complete its task without administrator assistance, with no unresolved access issue. It should decide which failures require a correction before expansion.
A low submission count may reflect an unclear requirement, duplicate work or an unavailable manager. Investigate the cause before arranging more training. Extend the process when the access, instructions and response route work together.
MatrixC offers Lark licensing, workflow implementation and onboarding. Discuss your rollout requirements with a list of users, devices, sites and the first process. Include who will support workers and who will own the decisions once requests arrive.


