Always-On Claude: 3 Routines Every B2B Sales Team Should Run By End of Quarter

On 14 April 2026, Anthropic released Claude Code Routines in research preview. It closed a gap that every team running Claude Code autonomously had been patching with duct tape.

Before Routines, making Claude run on its own looked like this: a cron job on a server, a webhook receiver, a custom script that picked up the event and called Claude, glue code to handle retries, a logging layer because you could not see what was happening. Routines collapses all of that into a native feature.

Three trigger types, all available on day one:

• Schedule - cron-style recurring, for example every weekday 7am.
• API - HTTP POST from any system that can call a webhook.
• GitHub - repo events like PRs, pushes and issues.

The question is not "is this interesting". The question is "what should your sales team run first".

Three routines that we think every B2B sales team should have operational by end of quarter.

Routine 1: Morning inbox triage

Trigger: Schedule - every weekday at 6:30am local time.

What it does: Runs through inbound leads that arrived overnight, applies qualification logic, and by the time your SDR team starts work the inbox is sorted into "respond now", "schedule follow-up", "route to a specific rep" and "not a fit". Each email already has a draft response attached for review.

Why it matters: The first 30 minutes of an SDR's day is usually spent triaging. Done by Claude overnight, those 30 minutes go straight into conversations with qualified leads. Multiplied across a team of six SDRs over a year, this is roughly 780 hours of recovered selling time.

Pairing: This is a good candidate for a Routines trigger handing off to a Managed Agent that can hold a session long enough to work through a full inbox.

Routine 2: Weekly deal-at-risk scan

Trigger: Schedule - every Friday at 4pm.

What it does: Pulls every open opportunity in your CRM that has gone dark for more than seven days, cross-references with recent account activity (email, meetings, calls logged) and produces a prioritised list of deals that need a re-engagement touch. Each deal comes with a recommended next action and a draft outreach message tailored to where the conversation left off.

Why it matters: Most B2B sales teams do deal hygiene weekly at best and often quarterly. Deals quietly rot between reviews. A Friday scan surfaces every rot-risk account while there is still time to re-engage before the next forecast call on Monday.

Pairing: Connects to your CRM via an MCP connector. Runs inside a Managed Agent session because the work involves reading many records and producing a structured output. See the Code vs Cowork guide for the split of who builds this vs who uses the output.

Routine 3: Named-account mention monitor

Trigger: API - inbound POST from your news and social monitoring tool.

What it does: When a named account shows up in the news (funding round, executive change, product launch, layoffs), the event fires a Routine. Claude pulls the article, summarises the relevance to your sales motion and drafts a context-aware outreach message addressed to the right contact on that account.

Why it matters: Timing matters in account-based selling. A funding announcement opens a three-week window where your message is massively more likely to get read. By the time an AE finds out through a Monday newsletter, the window is closing. A trigger-based routine compresses the reaction time from days to minutes.

Pairing: This is an API-triggered routine, so the work happens in response to an external event rather than on a schedule. The source system might be a media monitoring tool, a Slack alert, or a custom script watching a feed.

What Routines do not replace

Worth being clear, because this is where teams waste time.

Routines do replace:

• Custom cron-plus-glue setups where Claude was already the main actor.
• Simple self-hosted n8n workflows where the Claude step was 80% of the logic.

Routines do not replace:

• Multi-vendor orchestration where the work involves passing data between five SaaS systems. n8n still owns that job.
• GitHub Actions CI/CD. Routines has GitHub triggers but it is not a CI platform.
• Zapier for pure SaaS-to-SaaS automation without a reasoning step.

The rule: if Claude is the main thinker in the loop, Routines is the right home. If Claude is one of ten steps, keep the orchestration tool and just use Routines for the Claude part.

The "always-on Claude" combination

The single most useful pattern that emerges from Routines + Managed Agents:

Routines handles the when. Managed Agents handles the how. An event fires, a Routine picks it up, a Managed Agent session runs long enough to do the work, the result lands wherever your team needs it.

Before April 2026, this pattern existed as a DIY assembly. From April 2026 onward, it is two official Anthropic products that are designed to pair. We covered the full ladder in our Five Levels of AI in Revenue Ops guide.

Where to start

Pick one of the three routines above. Do not pick all three.

If your biggest bottleneck is inbound, start with Routine 1. If you are losing winnable deals to neglect, start with Routine 2. If account-based motion is your go-to-market, start with Routine 3. Whichever you pick, get it to "runs every day without breaking" before you add the next one.

TL;DR

• Routines launched 14 April 2026 in research preview.
• Three trigger types: schedule, API, GitHub.
• Best pairing is Routines + Managed Agents - the "always-on Claude" pattern.
• Three starter routines for B2B sales: morning inbox triage, weekly deal-at-risk scan, named-account mention monitor.
• Routines replaces cron + glue, not n8n or Zapier for multi-vendor orchestration.