The daily agent & overnight research

Diurno HQ works while you don't. Two things run on their own: the daily agent (housekeeping and re-scoring, every morning at 6:00 a.m. Central by default) and the overnight research agent (works through what was queued on the overnight lane).

What the daily agent does

Each morning, for every live workspace (archived ones are skipped), it re-scores and builds the Daily Brief — and, on the paid plans, also queues background research. (The Free plan gets the scoring and the brief but no background research: its small monthly allowance is yours to spend, and a lapsed pilot gets none either.) Each step of a workspace's pass is its own transaction: if one fails, it is named in the run's summary and in an info alert in that workspace (“Part of the nightly pass couldn't run for this workspace”), and the rest of the pass and every other workspace carry on. When the rescore or the quiet-account follow-ups could not read a setting (automation rules, sales plays or score-decay settings) the alert points you there; every other failed step says the problem is on our side, that it has been logged, and that the steps are tried again on the next daily run. For a paid workspace it:

• Refreshes accounts in Call now or Call this week whose research is more than 7 days old — hottest first, up to 25 a day per workspace; an account whose refresh failed in the past week is skipped.
• Researches accounts that have never been researched — up to 10 a day per workspace, oldest first. Customers, Disqualified accounts, and anything whose research failed in the past week are skipped.
• Looks for up to 5 new companies a day that fit your profile, each backed by a public source. They're added as New accounts (within your plan's account limit), their first research is queued, and an info alert lists them. For customer organizations it runs on the overnight lane, like the rest of the upkeep.
• Queues people checks for accounts not checked in 14 days (top scores first, up to 10 accounts per workspace), plus a weekly job-change check for your starred champions (up to 5).
• Re-scores everything with time decay — old evidence gently fades, so scores stay honest without anyone touching them — and raises an alert when an account crosses into Call now or Call this week, or cools back below Call this week.
• Fires your “No touch for N days” rules for accounts that have gone quiet.
• Pre-builds today's Daily Brief for every workspace, so the morning list is waiting.
• Sends the Daily Brief email digest — one email per teammate per morning, only to people who turned it on (SettingsTeamYour preferences; it starts off for new accounts), covering only the workspaces that person can see, and only when there's something to contact. Each email carries an unsubscribe link that switches off just that person's morning email — it doesn't touch the do-not-email list or anything else.
• Posts the same list to your Slack channel or webhook, if you've connected one (SettingsIntegrationsSlack & webhooks). A post that doesn't reach the channel isn't recorded as sent and is tried again on the next daily run.
• Runs a pass of the sequence engine — the dedicated runner already checks every 15 minutes — so every enrollment whose next step has come due sends, advances, or stops (reply detected, unsubscribed, account became a customer), with the reason recorded.

The numbers above are per day (counted in the scheduler's timezone, US Central unless changed), across every run that day: a catch-up after a restart, or a second press of the button, queues only what is left of the day's batch. Research you ask for yourself always runs ahead of the agent's own refreshes, first-time research, and discovery. Want the agent now instead of waiting? ResearchRun daily agent now — one run at a time per organization (a second click while one is running is refused), at most 10 an hour, and the summary lists only the workspaces you can see. When it finishes it reports what it queued and, if any step couldn't run, names the steps. The morning email and Slack posts still go out only once a day — each delivery is reserved before it is sent, so a run that overlaps the scheduled one can't send the same email twice.

The overnight research agent

Everything you queue on the free lane — imports, enrichment columns, outreach batches, people checks — waits as a held job until the overnight research agent claims it. It works through the waiting jobs in batches of at most 25 and takes turns between organizations: each organization's jobs are ranked by priority and then oldest first, and a batch takes every organization's first-ranked job, then every second-ranked one, and so on until it is full (Free-plan jobs after paid ones within the same round, not after all paid work), so one account's refresh wave doesn't block everyone else. Agent runs that start at the same moment are given different jobs. There is no guaranteed turnaround: a long queue takes more than one night, and the Research screen shows a banner if nothing has drained the lane for 36+ hours. It follows the same evidence rules as the instant lane and doesn't count against your instant allowance. The daily agent's own upkeep rides this lane too — stale refreshes, first-time research, people and champion checks, and the first research of newly found companies — so background work doesn't spend your allowance. Jobs waiting for it show as Waiting · overnight (free) on the Research screen, where ▶ promotes any of them to instant if you can't wait (while an instant lane exists).