Most small teams use AI like a vending machine. Ask a question, get an answer, walk away, start from scratch next time. Every conversation begins by re-explaining who the company is, who it sells to, and how it writes. That re-explaining is where the time goes and where the quality leaks.

Claude Projects fix that. A project is a workspace that holds shared context — your brand voice, your ICP, your positioning — that every conversation inside it draws on automatically. [Anthropic's documentation](https://www.anthropic.com/news/projects) covers the mechanics. The practical effect is that Claude stops being a stateless assistant and starts being closer to a team member who already knows the account.

Context is the bottleneck

The thing that makes AI output generic is missing context. Ask a plain chat to write a LinkedIn post for your company and it writes a LinkedIn post for a company. It does not know your voice, your buyer, or what you would never say. So it defaults to the average of everything it has seen, which is exactly the bland AI tone everyone can now spot.

A project loaded with your actual brand and ICP grounds the model in you. The output stops being the average and starts being on-brand. That is the entire value, and it is bigger than it sounds.

What goes in the project

Think about what a new hire would need to write on-brand on day one. That is roughly what goes in the project knowledge.

The brand voice guide. The ICP definition. The positioning and messaging. Three or four examples of strong existing content, because the model learns your voice better from examples than from description. The product or service details. And the do-not-do rules, the phrases and claims and tones you avoid.

Load that once. Now every conversation in the project inherits it. You ask for a blog draft and it comes back sounding like the company, not like generic AI, because the model is working from your context instead of its defaults.

The shared-brain part

Here is where it gets useful for a team. On the Team and Enterprise plans, projects can be shared, so the whole team works from the same context.

That solves a problem most small teams do not realize they have: consistency depending on who wrote the prompt. When the owner, a contractor, and a junior marketer all open separate chats, you get three different voices because each person prompts differently. When all three work inside the same project, the output is grounded in the same shared context regardless of who is typing. The brand brain lives in one place. Consistency stops being a function of who happened to write the best prompt that day.

Start with one project

Do not build ten projects. Build one, the content production project, because it is the highest-frequency, highest-leverage use.

Load the brand voice, the ICP, the positioning, and three strong example posts. Then route every first draft, every content brief, every repurposing task through it. The value becomes obvious fast, because the difference between generic-AI output and on-brand output is visible in the first draft you pull from it.

Once that one is working and the team trusts it, build a second project for analysis and reporting. But prove the concept on content first. It is the use that touches the team every day.

What it does not do

A project makes Claude a faster first-draft and analysis partner. It does not make strategic decisions. It does not own the brand. It does not replace the judgment of the person reviewing the output.

The failure mode is treating the project as the marketer instead of the marketer's leverage. A team that loads a project and then ships whatever it produces without review is producing fast, on-brand, strategically unsupervised content. The human is what makes the project worth having. The project is what makes the human faster.

Keep it current

A project is a living document, not a setup task. When the positioning shifts, update the project knowledge. When the ICP gets refined, update it. A project loaded with last year's context produces last year's output, confidently and on-brand and wrong.

Put a quarterly review on the calendar, the same cadence as an ICP refresh. The quarterly ICP refinement process is the natural moment to update the project too, since you are already revisiting who the company sells to.

If you only remember one thing

The reason AI output feels generic is missing context, and a Claude Project is the cleanest way to fix that for a small team. Load the brand brain once. Share it. Work from it.

Build one content project first. Keep it current. Treat it as leverage on human strategy, never as a replacement. Done right, it turns AI from a vending machine into a team member who already knows the account.

How to set up a Claude Project for a B2B marketing team: the knowledge base structure

The quality of a Claude Project depends almost entirely on what you put in it. A project loaded with vague, generic documents produces vague, generic output. A project loaded with specific, accurate documents about your actual company produces output that sounds like you and is grounded in your real situation. Here is what to include.

Brand voice document. This is the most important file in the project. It should describe the voice in concrete, testable terms: the formality level, the vocabulary the company uses and avoids, the sentence length and structure, the tone in different contexts (a blog post vs a LinkedIn update vs a proposal), and three to five example sentences that are on-brand alongside three to five that are off-brand. "Professional but approachable" is not useful. "Short declarative sentences, active voice, no jargon, no exclamation marks" is useful.

ICP definition. Who the company sells to, described specifically: the company profile (size, industry, stage, geography), the person's role and title, the problems they are trying to solve, the language they use when describing those problems, and what they are usually evaluating before talking to you. Update this every quarter — a stale ICP produces content written for buyers the company no longer prioritizes.

Positioning and messaging. The one-page positioning document: who the company serves, what it does for them, why it is different from the alternatives, and the core proof points. If this document does not exist yet, the exercise of writing it for the project will be useful in itself.

Do-not-do rules. A short list of things the company does not say, does not claim, and does not associate with. This is often the fastest way to prevent the most common AI errors — the generic confidence claims, the overclaiming, the competitor-baiting, the filler phrases that every AI defaults to.

Three to five strong example pieces. The best content the company has produced. Not average content — the pieces the team is proud of. Claude will learn from the texture and structure of these examples, not just the stated rules.

Real use cases: five things a marketing team should have a dedicated Project for

Most teams start with one project and discover the right structure as they use it. Based on how B2B marketing teams actually work, these are the five use cases that justify their own dedicated project.

  • Content production. The highest-frequency use. Blog posts, LinkedIn updates, email copy, content briefs, repurposing. This project gets the full brand voice, ICP, and example library. It is the one the team uses every day and the one that shows the fastest improvement in output quality.
  • Sales enablement. Proposal templates, one-pagers, case study drafts, objection-handling scripts, and email sequences for sales follow-up. This project needs the same ICP and positioning as content, plus product and pricing details, and the case study library. Keeping it separate from content production means the sales-facing materials stay distinct in tone and purpose.
  • Competitive analysis and research. A project loaded with competitor positioning, market context, and the company's own differentiation becomes a fast way to analyze competitor content, summarize industry reports, and draft competitive response messaging. Load new research as it comes in.
  • Reporting and analysis. If someone on the team spends time interpreting campaign data, writing monthly reports, or summarizing performance for leadership, a project with the reporting templates, metric definitions, and historical context speeds up that work. The model learns what good looks like for this company's reports.
  • Onboarding new contractors or team members. A project loaded with the brand guide, ICP, positioning, and examples is also the fastest way to brief a new freelance writer, a new agency, or a new hire. Instead of a two-hour onboarding document, you share the project and they start producing on-brand work from day one.