Plank AI Policy
(Last modified on May 5, 2026)
This policy is an internal document for our team, our partners, and our clients, and we’re sharing it here in the spirit of transparency.
Plank and AI
Plank is a Web development studio and certified B Corp.
This policy sets out how we use AI tools. It applies to everyone who works with us: employees, contractors, and collaborators on client projects.
Our approach treats AI as a conversation, not a shortcut. You start the conversation, bring your brain, and set the direction. The tools help you explore, draft, research, or work through options. Then it comes back to you for judgment, expertise, oversight, and critical thinking. Repeat that cycle as many times as the work needs.
The person always stands at the end of the chain. Whatever the output is, whether it is copy, code, a design direction, or a deploy, a person double checks it before it goes anywhere. The tools never have the last word. You do.
Skill and judgment matter more than any tool. The goal is better work, not just faster work.
Some people have principled objections to using AI — whether for ethical, environmental, or personal reasons. Those objections are respected here. No one at Plank will be pushed, penalized, or made to feel behind for not using AI. Usage is not mandated.
Our Principles
As a B Corp, we evaluate AI the same way we evaluate everything else — across five areas of accountability.
Workers: AI should make work at Plank better, not erode the skills and craft that make it valuable. We make sure that AI is not a replacement for the high value work our team does. We care about long-term capability, not short-term output.
Customers: We use AI to serve clients better, not to cut corners at their expense. We bring our humanity, expertise & craft to using these tools. Our clients’ data, privacy, and trust come first. We are honest with clients about how AI contributes to their work.
Community: We consider how AI affects the people around us — including those whose creative work trained the models we rely on. We will not use AI in ways that are exploitative or harmful to the communities we serve or belong to.
Environment: AI uses real energy and water. We use it with purpose, not excess. We prefer tools that are transparent about their environmental footprint and avoid redundant or unnecessary use. This tension does not have a clean answer, but we name it and keep working on it.
Governance: This policy is publicly available and openly maintained. When we make decisions about AI, we can explain them. Accountability is not optional.
How We Work With AI
Experimentation and Learning
Anyone at Plank can try a new tool or workflow without asking permission first. Always try to create an account using your plank.co email, and check for data privacy and sharing before using a new tool.
Keep experimentation separate from client work — tools being explored have not been vetted for delivery, privacy, or client suitability. Document what you find in the AI directory on our shared workspace. If something is worth bringing into standard practice, bring it to the AI working group.
Share what you learn. A message in Slack, a mention in team standup — whatever works. What you figure out should not stay with just you, even if the tool ends up not being a winner.
Internal Work
Use AI where it genuinely helps — research, writing, analysis, planning, internal communications. Do not use it to skip the thinking. A first draft from AI still needs your expertise and your voice. “AI wrote it and I skimmed it” is not a finished deliverable.
See Protecting Data and IP below for what stays out of AI tools.
Client-Facing Work
AI can support client work. It does not replace the expertise or judgment that clients hire us for.
No private client data goes into a public AI model without explicit client permission. That covers project briefs, research, user data, copy, and code. Anything under NDA or confidentiality sits in the same bucket. If you are not sure whether something is confidential, treat it as confidential.
Our default tools for client work are Claude Team and Gemini. Both keep what you share with them out of the public models, so your data stays between you and the tool. You can use other AI tools, but only after you confirm how they handle your data. If a tool feeds your chats back into its training, it is off limits for client work.
Our standard contract gives us room to use AI this way. Some clients have stricter rules that go further, and those rules take priority. Check before you start, and if something is unclear, ask a project manager or leadership.
Every AI-assisted deliverable gets a real human review before it ships. That means reading it, checking it against the brief, verifying facts, and applying your professional judgment.
- Start the AI conversation with the humans on your team at project kickoff, write down any decisions in a project documentation, and revisit it as work evolves.
- Clients have different comfort levels, and those levels are not always obvious. A client in a regulated industry needs a different conversation than one who is actively experimenting with AI.
- Confirm whether the client has their own AI policies, whether their industry has compliance considerations, and whether they are comfortable with AI-assisted work in each area — code, copy, strategy. Their rules take precedence over ours.
Be honest about AI use when it matters. If AI played a real role in a deliverable — not just spellcheck, but actual structure, content, or strategic output — say so. Be straight when asked, and proactive when the contribution is substantial.
We ask the same of clients and partners. If AI plays a real role in how you engage with our work, whether that’s generating variations, reframing our designs, or producing reference material from what we’ve shared, we’d appreciate a heads up, so we’re all clear on what we’re looking at in review. If you’re planning to feed in-progress work into AI tools, please flag it at kickoff so we can talk through data handling together.
Before You Use AI
Ask yourself these five questions:
- Would you be comfortable explaining this AI use to our client or to leadership?
- Is there a person who will review, own, and stand behind the output?
- Does this involve confidential or client data? (if yes, confirm you’re using an approved tool before you start)
- Does this serve the quality of the work, or are you just moving faster?
- Are you sure that whatever output comes from the AI can be used freely by you, Plank, and (if applicable) the Plank client? (See below, “Legality of the Output” section)
Clear on all five: go ahead.
Something feels off: pause and ask.
Legality of the Output
For outputs from an approved tool used on a paid account with data training turned off, we are in the best defensible position available.
The terms of service for Plank’s approved tools: Claude Team, Gemini, GitHub Copilot – each confirm that outputs belong to the user, not the tool provider. Using an approved tool correctly is the check.
If we are using a tool outside the approved list, or on a free tier, we cannot make that assumption. Stop and confirm the terms in the project, or AI channel, before the work goes anywhere.
For client deliverables where IP ownership matters significantly, product work, branded content, anything the client will commercialize, we try to flag it at project kickoff so it can be confirmed in writing as part of the project setup.
When in doubt: pause and ask.
Protecting Data and IP
Regardless of which tool you are using, do not put any of the following into an AI tool that has not been explicitly approved for it:
- Client data of any kind, including project briefs, user research, copy, and code; client data also includes basic information about the client like their name or address
- Anything under NDA or a confidentiality agreement
- Plank financials, contracts, or business-sensitive information
- Personal information about employees, clients, or users; personal information here means any information about an individual, like their name, phone number or email, or anything else that could be used to identify them
- API keys, credentials, or access tokens
- Unreleased product details, strategies, or internal plans
A specific note on code: when you use an AI-assisted IDE like GitHub Copilot, the tool has access to the code in your editor. Without proper guardrails, that code — including client code — can influence training data or appear in suggestions for other users. Plank’s supported Copilot configuration has data sharing turned off by default. Confirm your setup is correct before working on client code, and check with the AI working group if you are unsure.
When in doubt: anonymize it, or leave it out.
Security
Some AI tools can act on your behalf in ways that are hard to reverse. That gap between what a tool can do and what you expect it to do is where security risks live.
Be careful with advanced features. Some tools offer features that let AI act without a human reviewing each step. Before enabling anything that lets AI act on your behalf — especially on company systems, client accounts, or anything external — confirm what it can access, what it might do, and who is accountable if something goes wrong.
Do not grant AI tools broader access than they need. If a tool asks for access to your file system, email, calendar, or company systems, give it only what the task requires. Apply the same least-privilege thinking you would with any software.
If you notice unexpected behaviour from a tool, a permission prompt you were not expecting, or output that suggests the tool accessed something it should not have — flag it to the AI working group or leadership immediately via Slack, and do not proceed until it has been approved.
Bias
AI models carry the biases of the data they were trained on. That shows up in writing, research, and how outputs frame people and problems.
Plank builds things for real people. Review AI output critically before using it, especially when it involves representing communities, framing audiences, or informing decisions. We have examples of these types of biases in our AI resources directory.
Do not treat “the model said so” as a sufficient reason to include something in a deliverable. Be especially careful when AI is involved in research, audience segmentation, or content strategy.
Bias awareness is part of any AI training we provide. If you have approaches that help you catch and address bias in AI output, share them in the Slack channel, or add them to our resources directory.
Our Tools
Plank maintains an official tools list so that everyone has paid access to vetted tools — you should never have to choose between good data practices and a free-tier workaround.
Plank Provides
Every full-time employee gets an opt-in paid Claude Team account. Additional tools are budgeted on a per month basis, worked out between directors and team leads based on role and need. Contact operations to get set up.
| Tool | What It’s For |
|---|---|
| Claude Team (Anthropic) | General AI assistant — writing, research, strategy, analysis. Primary tool for everyone who opts-in. |
| GitHub Copilot (Microsoft) | Code generation and review — opt-in for developers |
| Granola (Granola) | AI meeting notes — desktop app, does not join your calls. Opt-in for everyone who needs meeting notes. |
Also approved as long as you use an account, and turn off model/data training.
| Tool | Notes |
|---|---|
| Gemini (Google) | Available through Google Workspace — you must be signed into your Plank account to use this. |
| Figma AI | Available through your Figma subscription |
| ChatGPT (OpenAI) | Pro tier is fine for internal, non-client work as long as you’re signed in and have data sharing turned off. Do not use Free tier for any work. |
| Perplexity | Good for some non-specific research as long as you’re signed in and have data sharing turned off. |
Adding a New Tool
Bring a short description to the AI working group: what the tool does, how it handles data, and whether the cost makes sense. The working group reviews requests on an ongoing basis. Until a tool is approved, keep client data out of it.
Do not use tools where data handling is unclear, where inputs train models by default, or where enterprise-level privacy cannot be confirmed.
Hard Lines
These are not guidelines, they are commitments.
- We will not use AI to replace people’s roles at Plank. The goal is better, more interesting work.
- We will not present AI-generated work as entirely human-made.
- We will not mandate that any team member use AI if they have principled objections to doing so.
- We will not use AI as the first or final voice on decisions that require human judgment and accountability.
Governance and Support
A small AI working group at Plank holds this policy. Their job is to keep it current, gather input from the team, track what is working, and stay close to what is happening in the industry. Membership crosses disciplines. Everyone at Plank is welcome to be involved.
The policy is reviewed every six months, or immediately when something changes — a regulation, a shift in tools, or a concern from the team. The tools list is reviewed quarterly.
A note on law: AI law is still being written — copyright, IP, data protection, and industry rules are all moving. Plank’s legal counsel reviews this policy as part of each formal update cycle. If a project raises a legal question about AI use, bring it forward. Do not assume the landscape is settled.
For shared learning: We use the AI channel in Slack. Share good prompts, useful workflows, and tools worth knowing about. We run periodic prompting sessions as tools and use cases evolve.
We maintain an AI directory within Employee Experience with examples, resources, and experiments. Everyone at Plank is welcome to contribute to that directory.
For questions, concerns, or tool proposals bring them to the AI working group or to leadership directly.
—
AI Disclaimer: This policy was created with Claude CoWork over 3 initial versions, then went through multiple rounds of human refinement with all levels of leadership. It pulls from our initial planning, PAW 2025 discussion, 2026 AI survey summarization and notes, and leadership’s direct review and choices.
Plank is a certified B Corporation and a founding organization behind the Ethical Web Collective. We care about the environmental cost of the tools we use, the rights of the people whose work trained the models we rely on, and the trust our clients place in us. AI is a powerful tool. So is a clear set of values.