Account Enrichment

Account enrichment automatically appends company data, decision-maker contacts, verified addresses, and AI intelligence scoring to each account, giving your team the context they need to prioritize outreach and close deals. The enrichment pipeline is shared with leads: same stages, same behavior, same review tools. This page describes it from the account perspective.

​

What Gets Enriched

​

Company Data

​

Contact Data

For each decision-maker contact discovered on an account:

​

Account Intelligence

After company and contact data are populated, AI generates: These appear on the Intelligence sub-tab of the account drawer.

​

How the Fit Score Is Calculated

The Fit Score is produced by an LLM synthesis pass over five weighted insight categories. Each category gets its own 1–100 sub-score and short reasoning paragraph; the headline Fit Score is the synthesis output, not a hand-tuned linear formula. Score-to-verdict mapping used across the UI:

​

Enrichment Triggers

​

Auto-Enrichment on Import

When importing accounts via the Import Wizard, the Auto-enrich imported accounts toggle on the Column Mapping step runs enrichment automatically as soon as rows land. Matched/duplicate rows are skipped. You can run enrichment on them manually if needed.

​

Manual

Enrich on demand from any account row:

1. Open the account drawer
2. Click Enrich
3. Wait for data to populate

​

Bulk: Selection-Driven Enrich Button

The Enrich button at the top of the Accounts (and Leads) table retargets based on what’s selected: Server-side BATCH_LIMIT = 50 is intentional. Pick the 50 highest-priority rows and run again rather than raising the cap. The button is disabled while any visible row is already queued or enriching.

​

Per-Column Fill

You don’t have to re-run the whole pipeline to refresh a single value. Each enrichable column on the Accounts and Leads tables has its own Fill action: an icon next to the column header (and a per-row variant where available) that re-fills just that field for the targeted rows. Four pipelines share the registry; each owns a distinct slice: Each cell tracks its own lifecycle state in the DB (<field>_fill_status = queued → filling → NULL) so the cell shows a Queued… or Filling… badge live without polling other columns. Each pipeline only asks for the slice it owns: “Fill domain” never touches website, name, or address; “Fill industry” never touches revenue.

​

Prerequisite Check

Per-column Fill only runs when each row has enough input for the pipeline to ground the lookup. The column-header button is disabled with a tooltip if any visible row is missing a prerequisite: The tooltip names the missing field(s) so you can fix the gap or narrow the selection.

​

Hover for Source Citations

Enriched Revenue, Employees, and Industry cells carry per-field citations from the original lookup. Hover any of those cells to reveal a tooltip with:

• The citation URL that grounded the value (typically a LinkedIn company page or industry profile)
• The page title so you can confirm the source before clicking through
• The model’s confidence when reported (high / medium / low)

If a row was enriched before the citations layer shipped, the tooltip stays hidden. There’s no “Sources unavailable” placeholder cluttering the cell. Re-enriching (or per-column Fill) will populate citations going forward.

​

The Enrichment Pipeline

Each enrichment run flows through several stages.

​

Stage 0: Domain Resolution

Before any lookup runs, an AI step resolves messy company strings (abbreviations, misspellings, customer-list short codes) into a canonical domain. This dramatically improves match rates downstream and lets enrichment work on rows that only have a company name.

​

Stage 1: Decision-Maker Discovery

Searches the company for decision-makers and surfaces the most senior contact available, attaching them to the account’s Contacts tab. If a primary contact already exists on the account, newly discovered contacts land as secondary contacts rather than overwriting the primary. Phone-only primaries created from import are preserved.

​

Stage 2: Company Data

Populates verified company data including industry, employee count, revenue, technologies, headquarters address, and a specific segment classification.

​

Stage 3: Phone Reveal

Requests a direct mobile number for the primary contact. If the initial lookup doesn’t return a number quickly, the system automatically falls back to a secondary source so the Mobile column populates reliably.

​

Stage 4: CRM Sync, Intelligence & Embeddings

The pipeline finishes with CRM sync (if a CRM is connected), AI intelligence scoring (fit score, timing signal, and priority actions) and vector embeddings that power natural-language semantic search across your account base.

​

Estimated Run Time

Enrichment runs in batches of up to 100 leads or accounts at a time, with the actual batch size adjusted in real time based on available resources. This applies to all enrichment: auto-enrichment on import, manual single-row enrichment, and bulk re-enrichment. Plan for roughly 12 minutes per batch end-to-end. Rows that match an existing record (duplicates) are skipped by auto-enrichment but can be enriched manually.

​

Revenue Formatting

Revenue is stored as a raw integer and rendered everywhere through a single formatRevenue helper: The web-research prompt is constrained to return a clean integer (no $, no K/M/B, no ranges, no “approx”) and Fill-revenue paths route through the same helper. Pre-EMA-263 rows with messy strings are normalized on read.

​

Overwrite Behavior

Enrichment never overwrites existing data. If a field already has a value, it’s preserved; if it’s empty, enrichment fills it in. This applies to every enriched field: name, email, phone, title, LinkedIn URL, company info, and so on. Addresses are handled with an extra layer: a verified billing address that differs from the existing one is stored as an alternate address for you to Accept or Keep Existing rather than being applied silently. Decision-maker contacts follow the same principle: newly discovered contacts land as secondary contacts when a primary already exists, and never replace the primary automatically.

​

Verified Addresses & Alternate Addresses

When enrichment returns a verified billing address that differs from the address already on the account, the new address is stored as an alternate instead of silently overwriting the existing one. The drawer’s Address section grows an Alternate Addresses card with Accept and Keep Existing buttons so you can promote or dismiss each candidate.

​

Enrichment Review

After a batch run completes, the Enrichment Review dialog highlights every account with discrepancies between previously stored data and the newly enriched values. Each row is tagged: Synthetic alternate-address cards are surfaced inline so you can resolve them in a single review pass.

​

The needs_review Status

Accounts that require human attention move to a dedicated needs_review state, surfaced as an amber pill in the table and an N need review count badge next to the enrichment ETA. Filtering the status column shows friendly labels for each state. Common reasons an account enters needs_review:

• Enrichment matched a domain already claimed by another account (potential duplicate)
• An alternate address was returned that disagrees with the existing primary
• Confidence on a critical field is below threshold

​

Duplicate Detection

If enrichment resolves to a domain that’s already claimed by another account, the second account is flagged needs_review and the Duplicate Review dialog gives you three choices: Duplicate detection runs on both leads and accounts, with parity in the review experience.

​

Semantic Search

Vector embeddings produced in Stage 4 power natural-language search across your account base. Type a query like “mid-market manufacturers in the Midwest hiring engineers” and the system returns the closest-matching accounts ranked by similarity.

​

Enrichment Quality

​

Match Confidence

Enrichment results carry a confidence label that’s surfaced on alternate-address cards and contact matches:

​

Handling Low Confidence

Low-confidence fields are tagged uncertain in the Enrichment Review dialog, and accounts containing them move to the needs_review state so they’re easy to spot and action manually.

​

Monitoring Enrichment

​

Dashboard Metrics

​

Enrichment Status

​

Troubleshooting

​

Best Practices

​

Enrich Strategically

​

Review Alternate Addresses Promptly

When enrichment surfaces alternate addresses, action them quickly so your account records reflect ground truth. Stale alternates can clutter the drawer and obscure what’s authoritative.

​

Monitor Quality

Regularly review enrichment quality:

• Spot check enriched accounts
• Track match rates and the needs_review count
• Action needs_review accounts promptly so the queue doesn’t pile up

​

Next Steps