A 30-person AI agency in Chicago was using their CRM as a glorified address book. Deals were entered sporadically. Stages were meaningless — "Contacted," "In Progress," and "Closing" covered everything from first email to contract negotiation. Pipeline reviews consisted of the founder asking each rep "how are your deals going?" and getting optimistic answers with no supporting data. When the agency hired a VP of Sales, the first thing she did was rebuild the CRM pipeline. She defined seven specific stages with clear entry criteria, added custom fields for AI-specific qualification data, and built automated workflows for stage transitions. Within three months, forecast accuracy improved from 35% (basically guessing) to 72%. Deal velocity increased because reps could see exactly what needed to happen to advance each deal. And the founder could finally answer the question "how much revenue will we close this quarter?" with confidence.
Your CRM pipeline is not a bureaucratic requirement — it is the tool that makes your sales organization predictable, manageable, and scalable. A well-designed pipeline tells every salesperson exactly where each deal stands, what needs to happen next, and whether the deal is on track. It tells leadership exactly how much revenue to expect and where bottlenecks are forming. And it creates the data foundation for continuous improvement.
Pipeline Architecture for AI Agency Sales
Multiple Pipelines for Different Sales Motions
Most AI agencies need more than one pipeline because they have more than one sales motion.
Pipeline 1 — New Business: Deals with new clients moving from first contact through signed contract. This is your primary pipeline and the one most of this guide addresses.
Pipeline 2 — Expansion: Deals with existing clients for additional AI use cases, department expansion, or scope increases. The sales process is shorter and the win rate is higher.
Pipeline 3 — Renewals: Upcoming contract renewals and subscription renewals. This pipeline ensures no renewal is missed and renewal negotiations begin early enough.
Pipeline 4 — Partners (optional): Deals sourced through or executed with technology partners. The process may include partner coordination steps not present in direct sales.
Stage Design Principles
Every stage must be objective. A deal's stage should be verifiable by anyone reviewing the CRM record — not dependent on the sales rep's subjective assessment.
Every stage must require action. Each stage transition requires the sales team to complete specific activities. Stages are not time-based ("Week 2") — they are activity-based ("Discovery Completed").
Every stage must have entry criteria. Specific conditions must be met before a deal can move to the next stage. These criteria prevent premature advancement that inflates pipeline estimates.
Stage count should be 5-8. Fewer than 5 stages lack granularity for pipeline analysis. More than 8 creates complexity that sales reps resist.
Detailed Stage Definitions
Stage 1 — Lead Qualified (5% probability)
Definition: A target account contact has been identified and initial qualification suggests a potential fit.
Entry criteria:
- Contact matches ICP for company size, industry, and role
- Initial signal of AI interest or need (content engagement, event attendance, outbound response)
- Basic company research completed
Required fields populated:
- Company name, industry, revenue, employee count
- Contact name, title, email
- Lead source (inbound, outbound, referral, event)
- Initial AI use case hypothesis
Activities in this stage:
- Complete company research
- Identify additional stakeholders
- Craft personalized outreach
- Schedule first meeting
Exit criteria: First meeting scheduled with a qualified stakeholder
Stage 2 — Discovery (15% probability)
Definition: Active engagement with the prospect through discovery meetings to understand their needs, qualify the opportunity, and assess fit.
Entry criteria:
- First meeting completed
- Prospect has confirmed interest in discussing AI solutions
Required fields populated:
- Business challenge description
- AI use case category
- Budget range (if known)
- Decision process description
- Decision timeline
- Stakeholder map (key stakeholders identified)
Activities in this stage:
- Conduct discovery meetings (business, technical, decision process)
- Complete qualification scoring
- Identify champion
- Assess technical readiness
Exit criteria: Qualification score meets threshold and the opportunity is formally qualified
Stage 3 — Solution Design (30% probability)
Definition: Designing the specific AI solution and building the proposal based on discovery findings.
Entry criteria:
- Discovery completed and documented
- Qualification score above minimum threshold
- Champion identified and engaged
- Budget range confirmed
Required fields populated:
- Qualification score (all dimensions)
- Champion name and assessment
- Proposed solution description
- Estimated deal value
- Target close date
- Competitive information (if competing)
Activities in this stage:
- Design the AI solution architecture
- Build the ROI analysis
- Develop the proposal document
- Prepare the presentation
Exit criteria: Proposal delivered to the client
Stage 4 — Proposal Review (45% probability)
Definition: The client is actively reviewing the proposal and evaluating the engagement.
Entry criteria:
- Proposal formally delivered to the client
- Presentation scheduled or completed
Required fields populated:
- Proposal delivery date
- Presentation date
- Stakeholder feedback notes
- Competitive situation update
Activities in this stage:
- Present the proposal to stakeholders
- Conduct technical demonstrations
- Provide references
- Address questions and objections
- Gather stakeholder feedback
Exit criteria: Client provides verbal commitment to proceed (or decides not to proceed)
Stage 5 — Verbal Commit (65% probability)
Definition: The client has verbally committed to moving forward and the contract process has begun.
Entry criteria:
- Verbal commitment from the economic buyer or authorized decision-maker
- Budget confirmed as available
Required fields populated:
- Verbal commit date
- Contract status
- Procurement contact identified
- Expected contract execution date
Activities in this stage:
- Send contract for review
- Engage with procurement
- Complete vendor registration
- Begin security review
Exit criteria: Contract in final negotiation with no major unresolved issues
Stage 6 — Contract Negotiation (80% probability)
Definition: Contract terms are being actively negotiated between legal and procurement teams.
Entry criteria:
- Contract delivered and in active review
- Procurement process initiated
Required fields populated:
- Legal review status
- Procurement status
- Security review status
- Open negotiation items list
- Updated expected close date
Activities in this stage:
- Negotiate contract terms
- Complete security review
- Resolve procurement requirements
- Obtain final approvals
Exit criteria: Contract signed by both parties
Stage 7 — Closed Won (100%)
Entry criteria: Contract fully executed by both parties.
Activities: Handoff to delivery team, kickoff scheduling, celebration.
Stage 8 — Closed Lost (0%)
Entry criteria: Deal definitively lost at any stage.
Required fields:
- Loss reason (primary and secondary)
- Competitor won (if applicable)
- Loss notes describing what happened
- Potential for future re-engagement
Custom Fields for AI Agency CRM
Deal-Level Custom Fields
- AI Use Case Category: Dropdown — Automation, Analytics, NLP, Computer Vision, Recommendation, Optimization, Other
- Technical Readiness: Dropdown — High, Medium, Low
- Data Availability: Dropdown — Existing and Clean, Existing but Messy, Partial, None
- Engagement Type: Dropdown — Project, Retainer, Subscription, Assessment, PoC
- Delivery Model: Dropdown — Fixed Price, T&M, Capped T&M, Value-Based
- Qualification Score: Number (auto-calculated if possible)
- Champion Strength: Dropdown — Strong, Moderate, Weak, None
- Loss Reason: Dropdown — Budget, Timing, Competition, Technical Fit, Champion Lost, No Decision, Internal Build
CRM Automation
Automated Workflows
Lead notification: When a new lead enters the pipeline, notify the assigned rep via email and Slack immediately.
Stage advancement alerts: When a deal moves to a new stage, notify the sales manager and update the deal record with the transition date.
Stale deal alerts: If a deal has not advanced stages in more than the expected duration for its current stage, alert the rep and manager.
Close date passed: If the expected close date passes without the deal closing, trigger a review — the date needs to be updated or the deal needs attention.
Renewal reminder: For existing clients, create a renewal opportunity automatically 90 days before contract expiration.
Win/loss documentation: When a deal is marked Closed Won or Closed Lost, trigger a required form for documenting the outcome and key learnings.
Data Quality Automation
Required field validation: Prevent deals from moving to the next stage unless all required fields for the current stage are populated.
Duplicate detection: Alert when a new contact or company matches an existing record.
Activity logging: Automatically log emails, calls, and meetings in the CRM through integration with your email and calendar systems.
Pipeline Health Metrics
Track these metrics to maintain a healthy pipeline:
Pipeline coverage ratio: Total pipeline value divided by quarterly quota. Target 3-4x coverage.
Average deal age by stage: How long deals spend in each stage. Identify stages where deals stall.
Stage conversion rates: What percentage of deals advance from each stage to the next. Identify where deals drop out.
Win rate by source: Do inbound leads close at higher rates than outbound? Does a specific referral source produce better outcomes?
Average deal size trends: Is your average deal size increasing, decreasing, or stable? Understand the drivers.
Forecast accuracy: Compare your quarterly forecast to actual results. Identify systematic bias (are you consistently over-forecasting or under-forecasting?).
Your Next Step
This week: If you have a CRM, audit your current pipeline against the stage definitions above. If you do not, choose a CRM and build the pipeline using these stages. Add the custom fields relevant to your AI sales process.
This month: Populate your CRM with every active opportunity. Train your team on stage definitions and entry criteria. Implement automated workflows for lead notification, stale deal alerts, and required field validation.
This quarter: Conduct weekly pipeline reviews using CRM data. Track pipeline health metrics and identify the biggest bottlenecks. Refine stage definitions and probabilities based on your actual conversion data. Achieve forecast accuracy within 15% of actual results.