An offline-first AI product may need a substantial first-run download. Tell users why it is required, what it enables, and what remains possible before it finishes.
The explanation should support a decision, not expose every runtime detail. Users need to know the transfer, storage, connectivity, and timing implications that affect when and where they can prepare the app.
Name the required preparation clearly
Separate application installation, model acquisition, optional assets, verification, and loading. State the expected transfer and storage based on the actual artefact, with an honest unknown or estimate where conditions vary.
Review current runtime documentation and model distribution requirements. Do not promise that every future update will fit the original download or that ready means every task is supported.
Walk through a fictional low-bandwidth setup
Suppose a teacher installs the app on a limited connection before travelling. The setup screen explains that local generation requires the model, allows postponement, and preserves imported material or manual drafts while installation waits.
Test user decline, interrupted transfer, restart, metered-network choice where supported, storage shortage, verification failure, and a required compatibility update. The interface should distinguish paused, failed, and ready instead of presenting one endless progress state.
After completion, disconnect and restart before the intended task. Verify the model and required source assets remain usable under the promised workflow. No download-duration or successful hardware result is claimed here.
Keep ongoing expectations consistent
Explain which future actions may need connectivity: updates, optional services, account checks, or obtaining new source material. Offline operation should be described by tested task boundaries rather than a blanket label.
Make cancellation and cleanup specific. Removing staged model data should not remove user-created lessons. If deleting the active model disables generation, state that consequence before the action.
The tradeoff is simple onboarding versus prerequisite transparency. Hiding a large download can make the first screen look easy while creating failure later. Too much technical setup can also overwhelm users. Present the minimum meaningful choice and make detailed requirements available when needed.
The ChalkLens case study documents a local teaching workflow. This first-run explanation is proposed guidance and does not describe its current installation screens or promise a particular transfer size.
Keep that distinction visible during recovery.
Questions about first-run model downloads
Does offline-first mean no setup internet?
Not necessarily. Separate installation prerequisites from steady-state disconnected use and verify both under the actual product design.
Should users be blocked from all work during download?
Only when the task truly depends on the model. Preserve independent viewing, importing, or manual editing where supported.
What should a readiness review inspect?
Inspect the explanation, download states, preserved work, cleanup, and a disconnected restart test. Bring hidden prerequisites to an offline onboarding review.