Minimum Viable CRM Data for AI: The 20 Fields You Need for Scoring, Enrichment, and Personalization

If you want AI to score leads, enrich records, personalize outreach, and predict deals, you do not need a perfect CRM. You need a consistent, validated “minimum viable” field set that your team can actually keep fresh, plus a few AI-specific fields like confidence and last verified timestamps.

TL;DR: Build your “minimum viable” schema around 20 fields across Account + Contact + Compliance + Activity. Enforce picklists, required rules, and freshness SLAs. Add enrichment confidence and last-verified dates so your AI can trust what it is using. Then map each field to an AI outcome: scoring, routing, email personalization, and pipeline predictions.

---

What “minimum viable CRM data” means for AI

Minimum viable CRM data is the smallest set of fields that:

1. Explains fit (who the company is, what they use, and whether they match your ICP)
2. Explains intent (signals that they are in-market)
3. Explains reachability and permission (can you contact them, and should you)
4. Explains momentum (last activity and lifecycle movement)
5. Explains trust (how recently it was verified, and how confident you are)

Without those five, AI features degrade fast:

• Lead scoring turns into “company size guessing.”
• Email personalization becomes generic and risky.
• Routing sends the wrong leads to the wrong reps.
• Deal predictions are built on stale stage data, not reality.

Related reading: Why AI Lead Scoring Fails (and How Enrichment Fixes It)

---

The 20 CRM data fields for AI (the actual list)

This is a pragmatic 20-field schema for B2B teams. It assumes you already have system fields like Record ID, Created Date, Owner, etc.

1) Account-level fields (10)

These power ICP matching, scoring, routing, and (for some models) win-rate predictions.

1. account_name (text)
2. account_domain (text, unique)
3. account_industry (picklist)
4. account_employee_count (number or range picklist)
5. account_annual_revenue (number or range picklist)
6. account_hq_country (picklist)
7. account_hq_region_state (picklist)
8. account_company_type (picklist: SaaS, agency, services, ecommerce, fintech, etc.)
9. account_tech_stack_primary (multi-select picklist: CRM, marketing automation, data warehouse, ads stack, etc.)
   • “Technographics” are data points describing the technologies a company uses. This is one of the highest leverage personalization inputs because it lets AI reference their stack without inventing it. See definitions and examples from HubSpot and Demandbase.
     Sources: HubSpot technographics explainer, Demandbase technographics FAQ
10. account_buying_signal (multi-select picklist)

• Examples: hiring in sales, recent funding, new executives, new product launch, competitor migration cues, “visited pricing page,” “requested demo,” “G2 intent,” etc.

2) Contact-level fields (7)

These power personalization, deliverability, routing, and sequence logic.

1. contact_first_name (text)
2. contact_last_name (text)
3. contact_email (email, unique per account domain rules)
4. contact_job_title (text + optional normalized title category)
5. contact_department (picklist: Sales, Marketing, RevOps, IT, Finance, Founder, etc.)
6. contact_seniority (picklist: C-level, VP, Director, Manager, IC, Owner)
7. contact_linkedin_url (url)

3) Source, consent, and compliance fields (2)

These reduce risk and prevent AI from recommending actions that violate your rules.

1. lead_source (picklist + optional “source_detail” text)

• Examples: inbound demo, webinar, referral, outbound list, partner, organic, paid search.

1. contactability_status (picklist)

• Suggested values:
  • ok_to_contact
  • do_not_contact
  • unsubscribed
  • bounced
  • complained
  • needs_consent_check (for edge cases like sole proprietors in certain regions)

Why this matters: in the US, the CAN-SPAM Act requires a clear opt-out mechanism for commercial emails, and opt-out requests must be honored within 10 business days. Source: FTC consumer guidance noting the 10 business day requirement

If you sell into the UK and EU, the rules differ by subscriber type and local implementation. For example, the UK ICO explains that PECR rules for electronic mail differ for corporate subscribers vs sole traders, and that UK GDPR still applies when processing personal data. Sources: ICO business-to-business marketing, ICO electronic mail rules

More compliance detail: Cold Email Compliance in 2026: SPF, DKIM, DMARC, One-Click Unsubscribe, and the 0.3% Complaint Rule

4) Lifecycle, activity, and AI trust fields (1 + 2 optional, but strongly recommended)

You asked for lifecycle stage, last activity, and enrichment confidence. These are where AI gets “ground truth.”

1. lifecycle_stage (picklist)

• Suggested: suspect, prospect, MQL, SQL, opportunity, customer, churned, partner

Strongly recommended (AI-specific fields that pay for themselves):

• last_activity_at (datetime)
• enrichment_confidence (number 0-100)
• enrichment_last_verified_at (datetime)

If you must stay at exactly 20 fields, treat last_activity_at and enrichment_last_verified_at as system fields if your CRM already has them. If it does not, add them even if it pushes you to 22. Your AI will be measurably better.

For more on “AI that takes action” and why these guardrails matter:

• Agentic CRM Checklist: 27 Features That Actually Matter (Not Just AI Widgets)
• Copilot vs AI Sales Agent in 2026: What Changes When Your CRM Can Take Action

---

CRM data fields for AI: what each field actually powers

Below is a simple mapping you can copy into your internal docs.

AI Lead Scoring: minimum inputs

AI scoring needs three buckets: fit, intent, and reachability.

• Fit: industry, employee_count, revenue, region, company_type, tech_stack_primary
• Intent: buying_signal, last_activity_at
• Reachability + risk: contactability_status, email validity/bounce history

Practical scoring recipe (example):

1. Start at 0 points.
2. Add points for ICP match (industry + size + region).
3. Add points for “stack match” (tech stack indicates compatibility or competitor displacement).
4. Add points for buying signals (funding, hiring, pricing page, demo request).
5. Subtract points if low confidence or stale verification.

Lead Enrichment: minimum inputs

Enrichment works best when you have a stable join key:

• account_domain is the most valuable “anchor” field.
• contact_email is the second anchor.
• linkedin_url helps disambiguate contacts with common names.

Store:

• enrichment_confidence
• enrichment_last_verified_at So you can decide when to refresh and when not to trust a value.

AI Email Writer (personalization): minimum inputs

Email personalization fails when it lacks concrete facts.

Your AI Email Writer should primarily reference:

• contact_first_name, job_title, department, seniority
• account_industry, employee_count, region
• tech_stack_primary
• buying_signal
• lead_source (tone: inbound vs outbound changes everything)

To avoid hallucinations, set a rule: the AI may only mention tech stack or signals if enrichment_confidence >= 70 and enrichment_last_verified_at <= 90 days.

AI deal predictions: minimum inputs

Predictions need a timeline and consistent stage definitions:

• lifecycle_stage plus your opportunity stages
• last_activity_at
• buying_signal
• firmographics (industry/size) because win rates vary by segment

---

Copy-paste “downloadable” schema: CSV header template

Option A: Exactly 20 fields (pure minimum viable)

account_name,account_domain,account_industry,account_employee_count,account_annual_revenue,account_hq_country,account_hq_region_state,account_company_type,account_tech_stack_primary,account_buying_signal,contact_first_name,contact_last_name,contact_email,contact_job_title,contact_department,contact_seniority,contact_linkedin_url,lead_source,contactability_status,lifecycle_stage

Option B: Recommended for AI (adds freshness + trust + activity)

account_name,account_domain,account_industry,account_employee_count,account_annual_revenue,account_hq_country,account_hq_region_state,account_company_type,account_tech_stack_primary,account_buying_signal,contact_first_name,contact_last_name,contact_email,contact_job_title,contact_department,contact_seniority,contact_linkedin_url,lead_source,contactability_status,lifecycle_stage,last_activity_at,enrichment_confidence,enrichment_last_verified_at

---

Field definitions and validation rules (copy into your CRM spec)

Account fields

• account_domain

  • Type: text
  • Rules:
    • required
    • unique
    • lowercase
    • strip http(s):// and paths
  • Freshness SLA: refresh if domain changes, or if enrichment confidence drops

• account_industry

  • Type: picklist
  • Rules:
    • required for scoring
    • restrict to a controlled taxonomy (avoid free text)
  • AI use: ICP Builder, scoring, routing

• account_employee_count and account_annual_revenue

  • Type: number or range picklist
  • Rules:
    • if unknown, store null, not 0
    • if range picklist, keep ranges consistent across teams
  • AI use: scoring, segment-based win-rate priors

• account_tech_stack_primary

  • Type: multi-select picklist
  • Rules:
    • store “known” tools only
    • do not allow reps to type random tools
  • AI use: personalization and competitor displacement
  • Reference: technographics definitions and use cases:
    HubSpot, Demandbase

• account_buying_signal

  • Type: multi-select picklist
  • Rules:
    • must include timestamp somewhere (either separate field like signal_last_seen_at or attach in your events system)
  • AI use: scoring, email triggers, prioritization

Contact fields

• contact_email

  • Type: email
  • Rules:
    • required
    • validate format
    • block role emails if your policy requires it (ex: info@, support@) or flag them
  • AI use: outreach, enrichment matching

• contact_job_title / department / seniority

  • Rules:
    • keep “job_title” as raw text
    • also normalize into department + seniority picklists for routing and scoring

Compliance and contactability

• contactability_status
  • Type: picklist
  • Rules:
    • required
    • automation must update it on bounce, unsubscribe, complaint
  • AI use: prevents an AI Sales Agent from emailing someone you should not contact
  • CAN-SPAM: opt-out must be honored within 10 business days (US). Source: FTC

If you email UK business contacts, note how rules differ for corporate subscribers vs sole traders, and that guidance can change. Sources: ICO B2B marketing, ICO electronic mail marketing rules

---

Data freshness: the simplest SLA that works

Most teams fail here by making rules too complex. Use a 3-tier SLA:

1. Tier 1 (14 days): buying signals, last activity

   • If this is stale, scoring is stale.

2. Tier 2 (90 days): job title, seniority, department, tech stack

   • Great personalization depends on this being current.

3. Tier 3 (180-365 days): firmographics like revenue and headcount

   • Slower moving, but still worth re-checking.

Operationalize freshness with two fields:

• enrichment_last_verified_at
• enrichment_confidence

Then implement one rule:

• If enrichment_last_verified_at is older than the SLA for that field category, queue re-enrichment.

---

Enrichment confidence: how to compute it (simple model)

You can compute confidence in a way that is transparent to reps.

Example (0-100):

• +40 if email verified recently (last 90 days)
• +20 if domain matches account domain
• +20 if LinkedIn URL matches name + company
• +10 if job title appears on LinkedIn and matches
• +10 if multiple sources agree (two providers or provider + public page)

Use it as a gate:

• AI Email Writer can reference technographics only if confidence is high.
• AI routing can assign “high confidence” records to senior reps first.

---

Recommended enrichment priorities by business type

You requested different priorities for agencies vs SaaS vs consultants. Here is a practical ordering.

For digital agencies

Goal: fast qualification, service-fit, budget hints, and urgency.

Enrich first:

1. industry
2. employee_count (proxy for marketing budget)
3. region/timezone (meeting logistics)
4. buying signals (hiring marketing, new funding, new site launch)
5. tech_stack_primary (CMS, analytics, ads platforms, marketing automation)

Why: agency offers often hinge on stack and maturity, not just company size.

For B2B SaaS companies

Goal: maximize ICP precision, competitor displacement, and product-fit triggers.

Enrich first:

1. company_type (is it actually SaaS, or services?)
2. employee_count + revenue bands
3. tech_stack_primary (CRM, marketing automation, data tools)
4. buying signals (pricing page visit, category intent, hiring RevOps)
5. department + seniority normalization

Why: SaaS outbound wins when you can say “we integrate with X” or “we replace Y” and route to the right motion (SMB vs mid-market vs enterprise).

For consultants (solo or boutique)

Goal: find the right buyer, the right problem, and get a meeting fast.

Enrich first:

1. job_title + seniority (you need decision-makers)
2. buying signals (leadership changes, growth indicators)
3. industry (so your offer is framed in their language)
4. region (time zone, language)
5. LinkedIn URL (for credibility and personalization)

Why: consultants win on relevance and trust. Great personalization beats huge datasets.

---

Implementation steps (so the schema actually sticks)

Step 1: Make picklists non-negotiable

At minimum, enforce picklists for:

• industry
• department
• seniority
• lifecycle_stage
• contactability_status
• lead_source

Free text equals untrainable AI and broken reporting.

Step 2: Automate the boring updates

Automate status updates from your sending and tracking systems:

• bounced -> contactability_status = bounced
• unsubscribed -> contactability_status = unsubscribed
• spam complaint -> contactability_status = complained

This matters because US compliance requires an opt-out mechanism and honoring opt-outs within 10 business days. Source: FTC

Step 3: Add routing rules tied to fields

Examples:

• If employee_count 1-50 -> SMB queue
• If department = RevOps and seniority >= Director -> senior SDR queue
• If buying_signal contains “pricing page” -> immediate follow-up sequence

Step 4: Put guardrails on autonomous actions

If you use an AI SDR, add hard gates:

• do not send if contactability_status is not ok_to_contact
• do not mention technographics unless enrichment_confidence >= 70
• do not reference events older than 90 days

If you are building agentic workflows, see: Genereti it for me

---

FAQ

What are “CRM data fields for AI”?

CRM data fields for AI are the specific account and contact attributes that an AI model uses to make decisions and generate outputs, such as lead scores, routing recommendations, personalized emails, and deal predictions. The best fields explain fit, intent, reachability, momentum, and data trust.

Do I really need technographics for personalization?

If you sell a product or service that depends on a prospect’s tools, yes. Technographics let your AI reference real stack context instead of guessing. Definitions and use cases are covered by HubSpot and Demandbase.

Which single field improves AI scoring the most?

If you already have basic firmographics, the highest leverage single addition is often account_buying_signal (with timestamps) because it separates “good fit” from “good fit right now.”

How should I store consent and unsubscribe status in the CRM?

Use a simple, enforced contactability_status picklist (ok_to_contact, unsubscribed, do_not_contact, bounced, complained). This prevents both humans and AI agents from contacting people who opted out. In the US, opt-outs must be honored within 10 business days. Source: FTC

What is enrichment confidence and why does it matter?

Enrichment confidence is a numeric score (0-100) indicating how trustworthy the record’s enriched fields are. It reduces AI hallucinations by gating what the AI is allowed to reference in emails, scoring explanations, and routing.

---

Put this schema into your CRM this week

1. Create the 20 fields (or the 23-field recommended version with freshness and confidence).
2. Convert key fields to picklists and lock them down.
3. Add two automations:
   • update contactability_status on bounce/unsubscribe/complaint
   • re-enrich records when last_verified_at exceeds your SLA
4. Map fields to AI features in your CRM UI so reps understand “why this matters.”
5. Run a 2-week cleanup sprint: enforce domain uniqueness, normalize industries, and backfill department and seniority.

If you want the fastest path to operational AI scoring, enrichment, and personalization, start with this minimum viable schema, then iterate based on what your AI models and reps actually use.