Salesforge + Clay Workflow: Enrichment to Personalization to Sequences
How to run the full salesforge clay workflow: build and enrich a Clay table, generate AI personalization columns before export, and activate multichannel sequences in Salesforge with field-mapped variables intact.
TL;DR
The short version
The Salesforge Clay workflow does its most valuable work before a single email is sent. Every personalization variable Salesforge renders at send time was generated inside Clay first — which means the quality of your sequence copy is determined at the enrichment and column-building stage, not inside the sequencer.
How to structure a Clay table so Salesforge can import it cleanly. How to use Claygent and AI columns to generate first-line and context variables before export. How to map Clay columns to Salesforge custom variables and avoid field gaps. Which Salesforge plan tier you need for AI personalization at volume. The three most common failure points in this workflow and how to avoid them.
The Framework
The three-stage Salesforge Clay workflow at a glance
- Stage 1: Build and enrich in Clay
Import your contact list into Clay and run waterfall enrichment to fill verified emails, firmographics, and any technographic fields your sequences reference. Add Claygent columns to pull company-specific research (recent funding, tech stack, job postings) into structured fields. The output is a Clay table where every row has a complete, verified contact record with research fields ready to become personalization variables.
- Stage 2: Generate personalization columns in Clay before export
Add AI columns that use the enriched data to generate ready-to-send text: a custom first line, a one-sentence pain-point observation, or a company-specific reason for reaching out. Each AI column becomes its own field in the table. When this stage is complete, your Clay table contains both the contact data and the copy fragments Salesforge will inject into the sequence at send time.
- Stage 3: Push to Salesforge and activate the sequence
Export the Clay table as a CSV or push via the native Salesforge integration. Map each Clay column to the corresponding Salesforge custom variable. Assign contacts to the sequence and confirm that variable preview renders the correct field values before enabling sends. The sequence runs with personalization drawn from the Clay data rather than from generic templates.
Pushing a Clay table into Salesforge without generating AI personalization columns first means your sequence runs with generic variable fallbacks or blank fields. The enrichment in Stage 1 is only useful if Stage 2 converts it into text that the sequence can render.
Stage 1 in Depth
Structuring the Clay table so Salesforge imports it cleanly
Column naming is the most common source of import failures in the salesforge clay workflow. Salesforge maps variables by column header name, so any column you plan to use as a sequence variable needs a name that is lowercase, has no special characters, and matches exactly what you declare as the variable name in the Salesforge sequence editor. Set this convention before building your waterfall — renaming columns after enrichment breaks any downstream AI column that references the original field.
Run enrichment in waterfall order: email verification first (Clay credits consumed on a stop-at-first-match basis), then firmographics, then technographics, then Claygent research columns. Claygent costs more credits per row than standard enrichment, so run it last and only on rows that have passed verification. A contact without a valid email should not consume Claygent credits.
Add a Clay formula column that evaluates to TRUE only when email is verified, first line is generated, and no required fields are blank. Filter on this column before exporting to Salesforge. This prevents rows with incomplete personalization data from entering the sequence and rendering blank variables at send time.
Stage 2 in Depth
Building AI personalization columns in Clay before the Salesforge export
Clay's AI columns generate text by passing enriched field values and a prompt to a language model. The prompt instructs the model to write a specific type of output (a custom first line, a pain-point observation, a company-specific hook) using the data in the row. The output lands in a new column as a plain text string — which is exactly what Salesforge expects as a custom variable value.
Keep AI column prompts tightly scoped. A prompt that asks for a two-sentence first line with a specific tone produces more consistent output than a prompt that asks for "personalized outreach copy." Inconsistent output length causes variable rendering problems in Salesforge email bodies, particularly when one row generates three sentences and another generates eight words.
Salesforge's own AI personalization feature (Personalization Credits) is separate from Clay's AI columns. If you generate personalization in Clay before export, those values arrive as plain text and do not consume Salesforge personalization credits. If you let Salesforge generate personalization at send time instead, each send in "Overdrive" mode consumes 2 credits per preview. On the Pro plan, 100 credits per month supports roughly 50 personalized sends in Overdrive mode. The Growth plan includes 1,000 credits per month. Verify current credit limits at salesforge.ai.
Stage 3 in Depth
Activating the Salesforge sequence with Clay-enriched data
Salesforge accepts Clay data via CSV import or through the native Clay integration available on the Growth plan. The integration route pushes contacts directly from a Clay table into a Salesforge sequence on a trigger, removing the manual export step. With CSV import, the column mapping step runs inside Salesforge's import wizard: match each incoming column to a Salesforge contact field or custom variable.
| Clay column type | Maps to in Salesforge | Common failure |
|---|---|---|
| AI-generated first line | Custom variable (e.g. {{first_line}}) | Variable name in Salesforge sequence does not match the column header from Clay exactly |
| Verified email | Contact email field | Rows with catch-all or unknown verification status pass through and generate bounces |
| First name | Contact first name field | Name field contains full name instead of first name only, renders incorrectly in greeting line |
| Company research summary | Custom variable (e.g. {{company_context}}) | AI column output contains quotation marks or newline characters that break variable rendering |
| Claygent research field | Custom variable (e.g. {{pain_point}}) | Claygent returned a generic fallback for rows where research was inconclusive; fallback text renders in the sequence |
Salesforge's variable preview shows how each email will render for a specific contact. Before enabling any sequence that uses Clay-mapped variables, preview at least 5 contacts that represent the range of your list — including any rows where a Claygent field may have returned a fallback or a short answer. Fix field mapping gaps at this stage rather than after sends have gone out.
Common Questions
Salesforge Clay workflow: frequently asked questions
The native Salesforge integration (push contacts directly from Clay to a sequence) is available on Clay's Starter plan at $134/month (annual billing) and above. CRM integrations and native sequencer pushes are not available on the free plan. Verify current plan features at clay.com/pricing, as plan tiers and credit allocations change.
A minimal pre-send checklist for the salesforge clay workflow checklist has five items: Clay table filtered to verified-only rows; all required AI columns populated with no blank values; column names matching Salesforge variable declarations exactly; variable preview confirmed on at least 5 representative contacts; and daily send volume set within the domain's warmup ceiling for the active sending mailbox.
Agent Frank is Salesforge's AI SDR agent that runs outbound autonomously at $499/month on quarterly billing. It handles sourcing, personalization, and reply management independently. If you want Clay enrichment to feed Agent Frank's personalization, you would need to push the Clay data into Salesforge contacts before assigning them to Frank's workflow. Verify current Agent Frank configuration options at salesforge.ai.
A typical salesforge clay workflow example for agencies: one Clay table per client with client-specific enrichment waterfall and Claygent research prompts tailored to the client's ICP. Each table exports to a dedicated Salesforge workspace or sequence set, so client sends stay isolated. The agency's team shares one Clay plan but manages separate tables per client; Salesforge's unlimited users on the Growth plan means no additional seat cost for client-facing team members.
Custom variable support is available on both the Pro ($40/month) and Growth ($80/month) plans. The key difference between the two plans for this workflow is personalization credits: Pro includes 100 credits per month, Growth includes 1,000. If you are generating all personalization in Clay before export, credits matter less because Clay's AI columns produce plain text that Salesforge renders without consuming its own credits.
Enrichment is only valuable if personalization lands.
See how Salesforge handles multichannel sequences, AI personalization credits, and the Primebox unified inbox for teams running Clay-powered outbound.