BlueSky’s Modern CRM Playbook Series: AI Based Lead Scoring

Why lead scoring matters?

Your reps should start every day with the right list. Dynamics 365 Sales gives you two viable paths: Microsoft’s predictive lead scoring (quick to turn on, AI-driven from your CRM history) and a custom model with Power Automate (you control the signals, weights, and when/how scores update).

Path A — Microsoft’s Predictive Lead Scoring

What it is. A built-in model that analyzes your historical wins/losses to score open leads. You can see top influencing factors and use scores anywhere in Sales (views, dashboards, routing).

Licensing note. Predictive lead scoring requires the Dynamics 365 Sales Enterprise tier. Microsoft’s docs also note an allowance of 1,500 scored records per month with Enterprise.

Where to turn it on (Microsoft Learn). Sales Hub → Sales Insights settings → Lead scoring.

Good for

• Fast time-to-value if you have clean history in CRM.
• Teams that want “good default” AI without building/maintaining rules.
• UI graphics that shows lead score and grade overtime.

Watch-outs

• Limited control over the exact features/weights used by the model.
• Monthly scoring capacity is limited on Enterprise license; advanced Sales Insights features can require additional licensing.

Path B — Custom Lead Scoring with Power Automate + AI (recommended for control)

Why teams choose this route. You decide what “hot” means and can blend firmographics + history + live intent. A flow runs at lead create/update, computes your score, and writes it back to a custom field on Lead.

Example signals we include

• Company Size (e.g., employee count or revenue band → larger = higher weight)
• Job Title (C-level/VP/Director > Manager > Individual Contributor)
• Previous Purchase (frequency/recency/amount → upsell potential)
• AI-derived Intent from Request Description
  Use AI to read the free-text “How can we help?” and infer purchase intent (e.g., “Need pricing and delivery lead time for 200 units next month” → High).

How it works (flow blueprint)

1. Trigger: When a Lead is created/updated in Dataverse.
2. Enrich: Look up Account size, last order, role seniority, etc.
3. Intent AI step:
   
   • Use AI Builder (Category Classification/Prompt-based sentiment/Custom prompt) to classify purchase intent from the description; or
   • Call an Azure OpenAI endpoint via HTTP for a custom intent/confidence output.
4. Score: Calculate a 0–100 score with your weights.
5. Write back: Update custom fields: c_leadscore and c_scoregrade on the Lead.

Licensing/capacity: Custom scoring flows typically require Power Automate Premium for Dataverse/premium connectors, and AI Builder credits if you use AI Builder models.

Scoring sketch (example weights)

• Company Size (0–30)
• Job Title seniority (0–25)
• Previous Purchase recency/value (0–25)
• AI Intent from description (Low/Med/High → 0/10/20)

What does “better results” look like?

Whichever path you choose, treat scoring like a product: monitor conversion by score band, tune thresholds/weights monthly, and A/B test routing rules. Use predictive scoring as a baseline, then move to (or augment with) a custom flow when you need richer signals or tighter control.

Side-by-side comparison