AI for repetitive browser work: describe it once, get it every week
Every Monday, the pipeline report compiled from the CRM and last week’s calls, in the same format, waiting for you. Every morning at eight, the overnight inbox triaged with replies drafted.
The twenty-minute jobs never get automated. Each is too small for a script and too frequent to drop, so the Monday report, the daily inbox pass and the lead-list refresh stay on you and quietly cost more than anything you did choose.
A routine takes them. Describe the job once in plain language and a companion runs it on schedule in your own logged-in browser, then hands you the result instead of the task. Free to download on Mac and Windows.
| By hand | Strawberry routines | |
|---|---|---|
| Cost per repetition | The same 20 minutes, every time | Described once, runs on schedule |
| When it happens | When you remember and have time | Every morning, whether the week is busy or not |
| Consistency | Drifts with attention and mood | Same format, same rigor, every run |
| Your role | Operator of every step | Reviewer of results; approver of sends |
| Improvement | You get faster, slightly | Corrections persist and compound |
What do people actually schedule?
The first routines people keep are unglamorous and specific:
- Weekday mornings: new leads matching the profile sourced into the sheet, skipping anyone already in it. The recurring form of sales prospecting.
- 8 a.m.: the overnight inbox classified, replies drafted for approval, the two threads that need you flagged.
- Monday: the pipeline report compiled from the CRM and last week’s calls, same format every time.
- Before every external call: a brief on who is in the room and what changed since you last spoke, with the transcript and drafted follow-ups landing after.
- Weekly: the competitor pricing check re-run and the moves noted, so the answer is already written when someone asks.
Described once, corrected forever
A routine beats re-delegating the same task because your corrections persist.
Tell a companion once that the report leads with churn and it leads with churn from then on. Context carries across sessions, so the tenth run is better than the first.
Not everything recurring is calendar-shaped. The vendor-comparison format you always want, or the screening pass you run whenever a role opens, saves as a skill instead. “Run the screening pass on these twelve applicants” is the whole instruction the second time.
Is scheduled the same as unsupervised?
No. Routines read, gather and draft freely, and sending, posting and changing records stop at an approval card that names what is about to happen. Drafted outreach waits for you. See the security page.
Keep Awake holds the machine up so a run that starts at eight finishes with the lid shut, and the results are sitting there when you open it.
How we check that unattended runs come back right
Real workflows, one written pass condition each, marked against a fixed key rather than reviewed by a model reading its own write-up. It re-runs every month from September, and the spec is here.
What it measures is whether the job finished without a nudge, which is the only version of a routine worth scheduling.
Experience Strawberry for free
DownloadTrusted by fast-growing companies worldwide
Frequently asked questions
Strawberry is free to download and includes AI credits to start. Paid plans begin at $20/month. See pricing. · Reviewed · Canonical facts for AI agents