Entrar na lista de espera

Lançamento no início de junho por ordem de chegada
Obrigado! Seu envio foi recebido!
Ops! Algo deu errado ao enviar o formulário.
2 September 2026
AI Use Policy

1. Purpose and scope

This AI Use Policy explains how Salesforge uses artificial intelligence across the Salesforge platform and related products, including Agent Frank, Primebox, Leadsforge and Mailforge (together, the "Services"). It describes the AI providers and feature areas we use, how customer data is handled, the level of human control available, and the choices customers have when using AI-enabled features.

This Policy supplements our Terms of Service, Privacy Policy, Data Processing Agreement and Subprocessor List. If an applicable agreement with a customer contains different or more specific terms, that agreement controls.

2. How Salesforge uses AI

Salesforge uses large language models and related machine-learning services supplied through commercial APIs. We do not develop or host our own foundation models. AI is used to assist with content creation, communication workflows, classification, search, research and quality checks.

Provider Role Feature areas
OpenAI Language models Email/follow-up generation; subject lines; writing-style matching; message variation; personalised openers; suggested replies in Primebox; out-of-office detection; LinkedIn message drafting; ICP persona generation; Agent Frank reply handling; knowledge-base search; lead research and qualification in Leadsforge.
Anthropic Language models Re-engagement message generation for Agent Frank sequences.
Google Embedding model Mailforge spam and content checks.
Groq Lightweight model infrastructure Intent detection in Leadsforge.
Cohere Reranking model Result ranking for knowledge-base search.
Langfuse AI observability Processing of prompts and responses for quality, reliability, troubleshooting and performance monitoring.

Specific model versions may change as providers release new or improved models. Salesforge selects models based on functionality, reliability, security and performance. Current model-version information may be provided on request where reasonably available.

3. Customer data and model training

Salesforge does not use customer data to train or fine-tune its own models or third-party foundation models. This includes email content, replies, contact records, contact lists, sequences and other customer-provided content.

Salesforge's AI providers are accessed through commercial API services that do not use submitted API data to train their general-purpose models. Customer data may still be processed to provide the requested feature, generate an output, maintain security, monitor service quality and comply with applicable law.

Salesforge may use customer-specific data within the customer's own workspace to provide retrieval, context or personalisation. This does not train or fine-tune a shared model and does not make one customer's data available to another customer.

4. Prompts, outputs and retention

When a customer uses an AI-enabled feature, relevant prompts, instructions, context and outputs may be sent to the applicable AI provider and processed solely to provide and support that feature. Salesforge seeks to limit the information sent to what is reasonably necessary for the requested function.

OpenAI and Anthropic may retain API inputs and outputs for up to 30 days for abuse monitoring and security, subject to the provider's applicable commercial terms and approved retention configuration. Salesforge is pursuing zero-data-retention arrangements with those providers. Unless and until such arrangements are confirmed as active for the relevant service, customers should not assume zero retention.

Langfuse processes prompts and responses for observability purposes, including monitoring quality, reliability, errors and performance. Salesforge applies access controls and retention settings to this processing in accordance with its internal security and data-governance practices.

Retention periods and processing arrangements may differ between providers and may change. Additional information may be provided through our Subprocessor List, Trust Center or on request.

5. Human oversight and automated operation

AI-generated content may be incomplete, inaccurate or inappropriate. Customers remain responsible for reviewing their configurations, approved sequences and use of generated content, and for ensuring that their outreach complies with applicable law and their own policies.

Human review is available for AI-generated content, is required for defined higher-risk categories, and is optional elsewhere according to the customer's chosen configuration. Auto-pilot functions may operate without case-by-case human review within configured limits. AI-generated sequence emails may also send on schedule after the customer has reviewed and approved the sequence.

Salesforge uses safeguards, workflow controls and monitoring intended to support reliable use, but does not guarantee that AI-generated content will always be accurate, complete or suitable for a particular purpose.

6. Customer choice and control

  • Agent Frank is optional. Customers may use Salesforge for fully human-led outreach without enabling Agent Frank.
  • Reply-as-human mode removes AI from reply handling.
  • AI reply classification is disabled by default.
  • AI agents are disabled by default at account level and are enabled only on request.

Some individual AI-assisted features may be part of the normal product experience. Customers who need a particular feature disabled or require a specific processing configuration should contact Salesforge to confirm the options available for their account and plan.

7. Transparency when interacting with AI

Where a person interacts directly with an AI system through the Services, Salesforge provides or enables clear disclosure that the interaction is AI-powered, unless this is already obvious from the context. This is intended to support informed use and the transparency requirements applicable to interactive AI systems, including Article 50 of the EU AI Act.

Customers deploying AI agents or AI-generated communications are responsible for using the available identification and disclosure controls appropriately and for adding any further notice required by the context in which they deploy the Services.

8. AI providers and subprocessors

AI providers that process personal data on Salesforge's behalf are treated as subprocessors where applicable. The AI-related providers currently used by Salesforge are OpenAI, Anthropic, Google, Groq, Cohere and Langfuse.

  • OpenAI: AI-assisted email and follow-up generation, subject lines, writing-style matching, message variation, personalised openers, suggested replies, out-of-office detection, LinkedIn message drafting, ICP persona generation, Agent Frank reply handling, knowledge-base search, and lead research and qualification.
  • Anthropic: Re-engagement message generation for Agent Frank sequences.
  • Google: Embedding services supporting Mailforge spam and content checks.
  • Groq: Intent detection in Leadsforge.
  • Cohere: Result ranking for knowledge-base search.
  • Langfuse: AI observability, including processing prompts and responses to monitor quality, reliability, errors and performance.

Salesforge maintains its current Subprocessor List separately so that provider, location and service information can be updated without republishing this Policy. Customers should refer to that list for the most current information and any applicable subprocessor-notification process.

9. Security and responsible use

Salesforge applies technical and organisational measures designed to protect data processed through AI-enabled features, including access controls, vendor assessment, monitoring and contractual protections. No AI system is error-free, and customers should avoid submitting sensitive data that is not necessary for the relevant feature.

Customers must not use the Services or AI-generated outputs for unlawful, deceptive, discriminatory, harmful or abusive activity, or in a way that violates the Terms of Service, third-party rights or applicable law.

10. Updates to this Policy

We may update this Policy as our Services, providers, legal obligations or data practices change. We will publish the current version with its effective or last-updated date. Material changes will be communicated where required by applicable law or contract.

11. Contact

Questions about this Policy, current model versions, AI processing arrangements or account-level controls may be submitted through the Salesforge support or privacy contact listed on our website and in our Privacy Policy.

Hire Agent Frank to augment your sales team

Automated prospecting
Auto-pilot meetings booking
Long-term learning