Pipeline Hygiene Agent

1. 01

   Issue detection

   Lists unhealthy opportunities.

2. 02

   Owner follow-up

   DMs the owner, parses the reply, writes confirmed changes.

3. 03

   Manager summary

   Friday counts-only post.

• Never change Amount. Never set a closed stage from a Slack reply; send the owner a record link instead.
• One DM per owner per day, one contact per opportunity per 3 days, at most 5 issues per DM.
• Write only fields the owner confirmed. Manager summary shows counts and names only.
• All three detection rules create issues on seeded data and re-runs create none.
• No more than 5 issues per DM.
• Updates occur only after an explicit yes.

Runs on Sales Cloud, Agentforce, Slack. Confirm editions and licenses with your Salesforce account team.

Salesforce products

Sales CloudAgentforceSlack

Editions and licenses

Agentforce must be enabled in your org. Confirm editions and licenses with your Salesforce account team; we check them in the first workshop.

Agent type

Autonomous agent

From you

Slack admin approval and user email matching. · Your sign-off on the stage names and the 14 day threshold.

Language

English. The agent writes in your locale and follows your wording settings.

Where it runs

Inside your Salesforce org. Records are read and written through the agent’s approved actions only.

Not included

Forecast category changes and closing opportunities. · Coaching or call analysis. · Email or SMS reminders.

Designed in full before a line is built, with 8 scripted test conversations. The numbers become verified results when the agent is delivered.

• Deployable sourceThe agent, its actions, objects and screens as Salesforce metadata, in your repository.
• Solution designA Solution Design Document with UML: architecture, data model, sequence and security.
• Test suiteApex tests on every action, plus scripted conversations replayed before every release.
• Least-privilege accessIts own permission sets, so the agent and your staff see only what the job needs.
• Runbook and handoverHow to run it, change its rules and roll back, with a walkthrough for your team.

• Least-privilege access

  It runs as its own user with a permission set written for this agent, so it can reach only what the job needs.

• Changes only through approved actions

  The agent cannot edit records freely. Every change goes through one of its 4 actions, each with one job.

• People make the big calls

  Never change Amount. Never set a closed stage from a Slack reply; send the owner a record link instead.

• Runs inside Salesforce

  Agentforce runs on the Salesforce platform and its Einstein Trust Layer. Confirm which Trust Layer features apply to agents in your org with your Salesforce account team.

• Every change is on the record

  What the agent does lands as normal Salesforce records with their full history, so you can see who changed what and when.

• The review is free

  A 60-minute call with an architect about the Pipeline Hygiene Agent and your org, with a written plan you keep.

• A fixed-scope proposal first

  Before any work starts you get a written proposal with milestones. You approve it, or you do not.

• Timing follows scope

  How long it takes depends on your setup and your rules. The proposal states the milestones up front.

• Or inside a monthly plan

  Agents can be built and run through one of our monthly delivery plans, alongside your other Salesforce work.

Salesforce licenses, including Agentforce, are bought from Salesforce and are not part of our fee. See our plans and pricing →

• One-page brief (PDF)

  What it does, the rules it follows and what you need, on one page you can forward.

  Get the brief →

• Slide deck (PDF, 6 slides)

  For your team or steering group: the problem, what it does, the rules it follows and how we deliver it.

  Get the slides →

• Talk to an architect

  A free 60-minute session on how it would work in your org, with your processes, your data and your rules.

  Book a session →

• The whole library

  Every agent we offer, by business area.

  Browse agents →

Further reading from our architects

• Agentforce for sales: where it helps, and where reps quietly ignore itA fourth test for sales AI candidates beyond value, data readiness and error cost: workflow fit. What genuinely helps reps, what they route around, and why review is not a control.
• Claudeforce: what actually changes for the people who run the orgSalesforce and Anthropic announced Claudeforce on 26 August 2026. What the sources actually state, what nobody has published yet, and what to do in your org before the beta reaches you.
• The grounding checklist we run before an agent talks to a customerTwelve pre-flight checks we run before an AI agent is allowed near a real customer, each with the failure it catches and the fix. A go or no-go list, not a readiness score.

• What does the Pipeline Hygiene Agent do?

  An autonomous Agentforce agent that finds unhealthy opportunities each weekday, asks each owner in Slack to fix them, writes the confirmed reply back to the record and posts a counts-only summary for the sales manager. A cleaner forecast without managers chasing reps.

• Which Salesforce products does it run on?

  Sales Cloud, Agentforce, Slack. Confirm editions and licenses with your Salesforce account team.

• Where do people use it?

  Slack DM, manager channel #sales-pipeline.

• What does it not do?

  Forecast category changes and closing opportunities; Coaching or call analysis; Email or SMS reminders.

• Is there a demo?

  Not yet. This agent is in our build program and is coming soon. We can scope it for your org now, and show you the agents that are already live.