Configuring AI

Pact's AI capabilities — email drafting, contact enrichment, and lead scoring — run against a configurable provider. You can use Anthropic, OpenAI, or Azure OpenAI.

Supported providers

Provider	Models	Notes
Anthropic	Claude 3.5 Sonnet, Claude 3 Haiku	Best quality for long-form drafts
OpenAI	GPT-4o, GPT-4o mini	Wide ecosystem, function-calling support
Azure OpenAI	GPT-4o (hosted)	Data residency in your Azure region

Setting your API key

Go to Settings → AI & Enrichment. Enter your API key for the provider you want to use and click Save. The key is encrypted at rest and never exposed in the API response.

If you are an admin managing multiple tenants, per-tenant keys are set in Admin → Tenants → [Tenant] → AI Settings.

Key scoping

API keys are scoped to the tenant. Workers that process enrichment and sequence jobs always use the key for the tenant that owns the job — cross-tenant key usage is not possible by design.

Customizing system prompts

Every AI feature ships with a default system prompt. You can override these prompts per tenant to match your brand voice, tone guidelines, or industry context.

Go to Admin → AI Prompt Library (staff only) or Settings → AI Prompts (tenant admin). Available prompt slots:

Slot	Default behavior
`email_draft`	Drafts a cold outreach email for a given contact + context
`email_followup`	Drafts a follow-up when no reply detected
`subject_line`	Generates 3 subject line variants for A/B testing
`enrichment_summary`	Summarizes a contact's enrichment data into a one-line insight
`lead_score_rationale`	Explains why a contact received a particular score

Prompt variables

Default and custom prompts can reference template variables:

code

{{contact.name}}          Full name of the contact
{{contact.company}}       Company name
{{contact.title}}         Job title
{{sender.name}}           Name of the sender (from their profile)
{{sender.company}}        Sender's company
{{sequence.name}}         Name of the sequence the step belongs to

Variables not present in the contact record are replaced with a sensible fallback (e.g., {{contact.title}} becomes an empty string if the field is blank).

Example custom prompt

code

You are a concise sales development representative at {{sender.company}}.
Draft a 3-sentence cold email to {{contact.name}} at {{contact.company}}.
Focus on the value Pact provides for consent-native outreach.
Do not include a subject line.

AI drafting in the sequence editor

In the sequence step editor, click Draft with AI to generate a starting point. The draft incorporates:

The contact's firmographic data (company, industry, employee count).
Prior steps in the sequence for context continuity.
Your custom email_draft prompt if one is set.

You can regenerate the draft or edit it inline before saving.

Lead scoring

Lead scoring runs nightly and assigns each contact a score from 0–100 based on:

Engagement signals (email opens, clicks, replies).
Firmographic fit (industry, company size) against your ideal customer profile.
Recency of activity.

Scores are visible on the contact detail page and available as a filter in list views and playbook rules.

To configure the ideal customer profile (ICP) used for scoring, go to Settings → Lead Scoring → ICP.

Pact does not mark up AI token costs. You pay the provider directly at their listed rates, using your own API key. Usage statistics (tokens in/out by feature) are available in Settings → AI & Enrichment → Usage.

Disabling AI features

If you prefer not to use AI features, leave the API key blank. Draft buttons will be hidden and enrichment jobs will skip the AI summarization step. Everything else (CSV import, sequences, consent management) works without an AI key.