Back to Legal

AI Policy

Last updated: May 2026

Overview

Artificial intelligence tools are a meaningful part of how we work at 37SOLUTIONS. We use AI tools to write more clearly, review code, conduct research, and solve technical problems faster. This policy explains which tools we use, how we use them, what guardrails we apply, and what that means for your data.

We believe in transparency about AI use. If you have questions about how AI tools were used in work we performed for your project, please ask.

AI Tools We Use

We currently use the following AI tools in our operations:

  • Claude (Anthropic): Our primary AI assistant for writing, code review, documentation, technical research, and problem-solving. We access Claude through claude.ai and the Anthropic API.
  • Claude Code (Anthropic): An agentic coding tool used for software development tasks, codebase navigation, and implementation work. Runs locally via the command line.
  • ChatGPT (OpenAI): Used for general research, writing assistance, and cross-referencing AI outputs on complex or nuanced topics.
  • Perplexity: Used for research and fact-finding tasks that benefit from web-grounded answers.
  • Duck.ai (DuckDuckGo): Used for lightweight queries where additional search privacy is preferred.
  • Zendesk AI: Integrated into our support platform for automated ticket acknowledgment and self-help suggestions. When a support ticket is received, Zendesk AI may send an automated reply acknowledging receipt and surfacing potentially relevant help articles while the team reviews the ticket. All substantive responses are written and sent by a human.
  • Grammarly: Used for grammar, clarity, and tone review on client-facing written communications.
  • SparkMail AI (Readdle): Used for summarizing email content. Not used to draft or send replies.
  • QuickBooks AI (Intuit): Present within our QuickBooks accounting environment. We have not been able to identify opt-out controls for this tool at this time (see Training and Opt-Out Settings below). Its use is limited to our internal accounting operations; client data is not entered into QuickBooks beyond what is necessary for billing.
  • Local LLMs (Ollama): We occasionally run locally hosted large language models for tasks where keeping data entirely off third-party servers is preferred.

This list may change as the AI landscape evolves. We will update this policy to reflect material changes in the tools we use.

Automation Tools

We also use n8n, a self-hosted workflow automation platform, for internal process automation including support ticket triage and accounting integrations. n8n is not an AI tool — it executes rule-based workflows rather than generating AI output — but it does interact with data from support requests and billing systems. It runs on our own infrastructure and does not send data to third-party AI services.

Training, Tracking, and Monitoring Settings

We have disabled all available training, telemetry, tracking, and conversation monitoring settings across every AI tool we use, to the extent those controls are reasonably available to us. This includes opting out of model training on our inputs and outputs wherever such opt-outs are offered.

Exception: We have not identified a way to disable AI features or data use within QuickBooks at this time. We will update this policy if that changes. QuickBooks AI access is limited to our internal accounting environment.

We cannot guarantee that every AI provider honors these settings in all circumstances. Clients with strict data sensitivity requirements should let us know so we can adjust our workflows accordingly.

How We Use AI Tools

AI tools are productivity aids that help us work faster and more thoroughly. They are not replacements for human judgment, technical expertise, or client relationships.

Specific uses include:

  • Writing and documentation: Drafting client-facing content, internal documentation, reports, and communications. All AI-assisted writing is reviewed, edited, and approved by a human before delivery.
  • Code assistance: Generating, reviewing, or debugging code. All AI-generated code is reviewed by an engineer before use in production. AI-generated code is not deployed without human review and testing.
  • Research and analysis: Researching technical topics, comparing solutions, and synthesizing information. AI-generated research is verified against primary sources.
  • Problem-solving: Working through technical problems, diagnosing issues, and exploring solutions. AI suggestions are always validated against actual system behavior.
  • Support operations: Zendesk AI may send an automated acknowledgment when a ticket is received, including links to potentially relevant help articles. No substantive support response is sent without human review.
  • Email summarization: SparkMail AI is used to summarize incoming email content. It is not used to draft or send replies on our behalf.
  • Internal accounting: QuickBooks AI features are present in our accounting environment. Their use is limited to internal financial operations.

Client Data and AI Tools

We do not input confidential client data into AI tools. Specifically, we do not share:

  • Client passwords, API keys, access credentials, or secrets
  • Non-public business information, financial data, or personally identifiable information about a client's customers
  • Proprietary client code, content, or data that has been explicitly designated as confidential

When we use AI tools for tasks involving client systems or projects, we use only the minimum information necessary — typically anonymized or generalized descriptions of the problem rather than actual client data.

If a specific project requires AI-assisted analysis of client data or content, we will discuss this with you first, explain which tools would be used and under what terms, and obtain your explicit agreement before proceeding.

AI on Client-Hosted Infrastructure

Some clients host AI-integrated applications on infrastructure we manage. Our hosting and management services are infrastructure-agnostic, and we do not restrict clients from building or hosting AI-powered applications on their infrastructure.

Clients are responsible for ensuring their AI-integrated applications comply with:

  • The terms of service of the AI providers they use
  • Applicable data protection and privacy laws, particularly when processing personal data through AI systems
  • Our Acceptable Use Policy, which prohibits hosting applications designed to harm, deceive, or defraud

If you are building AI-integrated applications, we are happy to advise on infrastructure, compliance, or security considerations.

AI Disclosure Standards

We are committed to transparency about how AI contributes to our work:

  • We do not represent AI-generated content as entirely human-written when directly asked
  • Where AI tools played a significant role in producing a deliverable, we will disclose that upon request
  • We do not use AI to impersonate individuals or generate deceptive content
  • We do not use AI-generated voices, images, or video in ways intended to deceive

Quality and Accuracy

AI tools can produce incorrect, outdated, or misleading information. Every AI-assisted output is reviewed by a qualified human before use. We do not deliver work product to clients based solely on AI output without human verification.

We remain responsible for the accuracy and quality of the work we deliver. Using AI tools does not change that responsibility.

Our Position on AI

We think AI tools are powerful, useful, and here to stay. We use them because they help us deliver better work faster. The technical judgment, problem-solving experience, client relationships, and accountability that define 37SOLUTIONS are human — and that will not change.

We are thoughtful about where AI helps and where it does not. We will not use AI to cut corners on quality, replace genuine expertise, or deliver work we have not actually thought through.

Changes to This Policy

AI tools and best practices are evolving rapidly. We will update this policy as our practices change. Material updates will be communicated to active clients by email.

Contact

Questions about this policy or how AI tools are used in work we perform for your project: