Set Up a Sales Pipeline with AI: Step-by-Step

A sales pipeline isn't just a list of deals. Done right, it's a system that tells you exactly what to work on next, surfaces deals that are going cold, and gives you reliable revenue forecasts. Done wrong, it's a data graveyard.

This guide walks through setting up a sales pipeline in DenchClaw that actually works — with AI doing the work of keeping it accurate and surfacing the right information at the right time.

What You're Building#

By the end of this guide:

• A Deals pipeline with Kanban view
• AI-powered deal enrichment and scoring
• Automated follow-up reminders
• Win/loss analytics
• Pipeline forecasting

Setup time: 1-2 hours. Install DenchClaw first with npx denchclaw if you haven't already.

Step 1: Design Your Pipeline Stages#

Before touching any software, decide on your stages. Most B2B pipelines have 5-7. Here's a solid default:

1. Lead — initial contact, unqualified
2. Qualified — confirmed fit, budget, timeline
3. Discovery — deep needs assessment, demo scheduled
4. Proposal — pricing/proposal sent
5. Negotiation — in active back-and-forth
6. Closed Won — signed
7. Closed Lost — dead

Fewer stages = less overhead. More stages = better visibility. Choose based on your sales cycle length.

Step 2: Create the Deals Object#

In DenchClaw's web chat or Telegram:

Create a Deals object with these fields:
- Deal Name (text)
- Company (relation to Companies)
- Contact (relation to People)
- Value (number)
- Stage (enum: Lead/Qualified/Discovery/Proposal/Negotiation/Closed Won/Closed Lost)
- Close Date (date)
- Probability (number, 0-100)
- Deal Source (enum: Outbound/Inbound/Referral/Conference/Other)
- Owner (text)
- Next Action (text)
- Next Action Date (date)
- Notes (richtext)

Set the default view to Kanban, with Stage as the kanban field:

Set the default view of Deals to Kanban, grouped by Stage

Your pipeline board is live. Deals appear as cards, organized by stage.

Step 3: Create Supporting Objects#

If you don't already have Companies and People objects:

Create a Companies object with: Company Name, Website, Industry, Size, ARR, Primary Contact (relation to People)

Create a People object with: Full Name, Email, Phone, Company (relation to Companies), Title, Last Contacted, Notes

Now your deals are linked to real companies and contacts. When you ask "what's our biggest deal?" DenchClaw can pull company size, contact details, and deal history in a single query.

Step 4: Import Your Existing Pipeline#

From HubSpot:

Import my HubSpot deals pipeline

DenchClaw opens HubSpot in your browser, exports your deals, and maps them to your Deals object. See the HubSpot import guide for detailed steps.

From a spreadsheet:

Import deals from this CSV, mapping Company to Deal Name, Amount to Value, Stage to Stage [attach file]

The AI handles field mapping and asks you to confirm ambiguous matches before inserting.

Manual entry:

Add a new deal: Acme Corp, $45,000, Discovery stage, close date end of Q2, primary contact is Sarah Chen

Step 5: Set Up AI Deal Scoring#

This is the feature that makes DenchClaw's pipeline different from a static kanban board.

Create an Action field for deal scoring:

Add an Action field to Deals called "Score Deal" that evaluates deal health based on: days in current stage, last activity date, value relative to average, whether close date has passed, and returns a score from 1-10 with a brief explanation

Now every deal has a score button. Run it weekly to identify at-risk deals before they slip.

For automatic scoring, set up a scheduled workflow:

Every Friday at 9am, score all deals in active stages and send me a Telegram summary of any deal that scores below 5

Step 6: Automate Follow-Up Reminders#

The most common reason deals die: someone forgot to follow up. DenchClaw fixes this.

When a deal moves to Proposal stage, set a Next Action Date of 3 days from now and send me a reminder on that date

If any deal hasn't had a stage change or note added in 14 days, send me a Telegram alert with the deal name and current stage

You can also use the Next Action and Next Action Date fields manually:

After every interaction with a prospect, log a note and set a Next Action Date

Ask for your daily action list every morning:

What deals need action today?

DenchClaw queries Next Action Date for today, pulls the deal names and next steps, and sends you a prioritized list.

Step 7: Build a Forecasting View#

Create a view called "Q2 Forecast" showing deals with close date in Q2, grouped by Stage, with columns: Deal Name, Company, Value, Probability, Close Date

For a weighted forecast:

Show me total weighted pipeline value by close month, where weighted value = Value × (Probability / 100)

This gives you a realistic revenue forecast, not just total pipeline value.

Build a chart showing monthly forecast for the next 6 months as a bar chart, with Closed Won as green and pipeline as blue

Step 8: Win/Loss Analytics#

Every lost deal is data. Set up a post-mortem workflow:

When I mark a deal as Closed Lost, prompt me to fill in a Loss Reason field (enum: Price/Competition/No Budget/Timing/No Decision/Other) and a brief note

Monthly analytics:

Show me win rate by Deal Source, average days to close by industry, and top 3 loss reasons this quarter

These three numbers — win rate by source, cycle length by industry, top loss reasons — will tell you more about your pipeline health than any dashboard.

Step 9: Connect to Your Email and Calendar#

Connect Gmail and Google Calendar

Once connected:

• Deal-related emails get logged to entry documents automatically
• Meeting prep: "Brief me on the Acme Corp deal before my call at 2pm"
• Calendar blocking: "Block 2 hours every Tuesday for pipeline review"

Maintaining Your Pipeline#

A pipeline that isn't maintained is noise. Build these habits:

Daily (5 minutes): Check "What deals need action today?"

Weekly (30 minutes): Move deals that progressed, close out dead deals, add notes to everything active.

Monthly (1 hour): Win/loss review, forecast vs. actuals, source performance.

The AI makes maintenance faster. Instead of clicking through every deal, you say "Update my pipeline" and DenchClaw walks you through each active deal with context.

For more advanced techniques, see the DenchClaw power user guide or the complete DenchClaw overview.

Frequently Asked Questions#

How many pipeline stages should I have?#

5-7 for most B2B sales. Fewer stages mean less maintenance; more stages give better visibility into where deals get stuck. Start with 5 and add stages only when you need to track a transition you're currently losing visibility on.

Can DenchClaw sync with my email to automatically log activity?#

Yes, via the Gmail integration. Once connected, you can ask DenchClaw to pull email threads related to a specific deal and log them to that deal's document.

How do I handle a team pipeline in DenchClaw?#

Add an Owner field to Deals and create views filtered by owner. Full team CRM setup is covered in the team CRM guide.

What's the difference between a personal pipeline and a team pipeline in DenchClaw?#

Architecturally they're the same. The difference is access — team setups use shared workspaces where multiple users can update the same DuckDB. DenchClaw supports multi-user team workspaces.

Can I run DenchClaw alongside my existing CRM while I'm evaluating it?#

Absolutely. DenchClaw doesn't replace your existing tools until you want it to. Run it in parallel, import a subset of deals, and test the workflow before committing.

Ready to try DenchClaw? Install in one command: npx denchclaw. Full setup guide →