It's the fifth time this week you've pasted your brand guidelines into a blank Claude chat. By the time you've explained who you are, who you sell to, and how you like things written, the actual question feels like an afterthought. That re-explaining is the whole problem, and it's why most marketers think Claude is fine but not transformative.
To use Claude well for B2B marketing, stop treating it as a chatbot and start treating it as a system. Create a Project, load it once with your brand voice, your real audience language, and your editing standards, then run repeatable workflows on top: briefs, first drafts, research synthesis, repurposing. The context you give it, not the cleverness of the prompt you type, is what decides the quality of what comes back.
Because most people use it like a search box. You open a fresh chat, type a one-line request, get an answer that's competent but generic, and quietly conclude the tool is overrated. The problem isn't the model. It's that a blank chat knows nothing about your company, so it gives you the average of everything it has read on the internet. Average input, average output.
This isn't a niche observation. Salesforce's 2026 State of Marketing report, which surveyed nearly 4,500 marketers, found that 75% have adopted AI yet 84% admit their campaigns still feel generic. Their own marketing CMO, Bobby Jania, put it bluntly: "We are using the most powerful technology in history to send more one-way spam, faster." The diagnosis there is the same as the one at your desk. The tool is everywhere; the context that would make it produce something specific is missing.
The marketers getting real value have noticed something the rest haven't: with Claude, the quality shows up when context compounds. The first answer is rarely the good one. The value lives in a properly briefed system, not a clever opening line.
The mechanism is a Claude Project. A Project is a persistent workspace with a knowledge base and a set of standing instructions that every conversation inside it inherits automatically. You load your context once, and from then on Claude already knows who you are before you've typed a word. No re-pasting the brand guide. No re-explaining the audience. You go straight to the question that matters.
That sounds small. It's the single biggest unlock there is. The blank-page tax disappears, and output quality stops depending on what you happened to remember to include that day. It becomes consistent because the foundation is consistent.
You don't have to take our word for the scale of it. HubSpot, a company with a fairly demanding bar for marketing quality, built hundreds of Claude Projects across its teams specifically to centralise context, and reports a 40% productivity increase across content and web development work. (That's HubSpot's own figure, so weight it accordingly, but the mechanism behind it is the point: shared, persistent context.) Their CMO Kipp Bodnar said the thing most marketers feel but rarely say out loud: "Claude has really good taste, and marketing's all about taste." Norway's largest communications group, TRY, reported developing proposals 40% faster after rolling Claude out across 400-plus creatives. Different companies, same pattern: the gains come from context that persists, not prompts that impress.
The honest answer: anything repeatable, context-heavy, and writing-centred. That covers more of the job than most people expect.
Where it earns its keep most reliably: content briefs, first drafts of long-form content, and research synthesis. You can paste an analyst report, five competitor pages, and three call transcripts into a single conversation and ask for a structured competitive picture, and Claude will hold all of it at once. Add repurposing a pillar piece into channel-ready variants, pulling real customer language out of sales-call notes, and quality-checking copy against your house style, and you've described most of a marketing manager's week.
The pattern worth internalising is this: Claude is strongest on the grind, not the final polish. It gets you past the blank page, does the first structure, and synthesises the messy inputs, which is exactly the part that eats your afternoon. The last 20%, the sharp claim and the bit of taste that makes the piece yours, is still your job. That's not a weakness. That's the trade you want.
Context, not prompt tricks. This is why "best Claude prompts for marketing" lists tend to disappoint: a brilliant prompt fired at an empty Project still gives you the average. Three things do most of the work, and they're the three most teams skip.
Brand voice, shown not described. Don't tell Claude you're "bold and friendly." Adjectives mean nothing to it. Paste in three or four real pieces of writing you'd happily put your name to, labelled by context. Samples beat rules every time.
Audience, in their own words. Lift the actual phrases your customers use on calls. "We're stuck competing on price" lands very differently from a persona caricature like "Sarah, 42, frustrated with manual processes." Real language starts Claude in the customer's world rather than the consultant's.
Standards, written down. The patterns you always end up editing out: the AI tells, the words you've banned, the structures that creep in. Once those live in the Project, Claude stops making the same mistakes and you stop fixing them.
Get those three right and the difference is night and day. Skip them and no amount of prompt engineering rescues the output.
Keep it deliberately small. One Project, three short files, one workflow. It's about 90 minutes of focused setup, and you can do all of it inside the standard Claude interface with no technical skill.
Set up a single Project for your marketing work. In the custom instructions, cover who you are, who you're writing for, the register you use, and that you write in British English. Then add three plain documents to the knowledge base: a brand-voice file (your samples), an audience file (real customer phrases), and a standards file (what to cut). Finally, pick one high-frequency task to run through it. Content briefs are the best first choice, because you do them constantly, the output is easy to judge, and a good brief lifts the quality of everything downstream.
Resist the urge to perfect the files on day one. They get sharper through use. Every time Claude produces something almost right but not quite, that gap tells you exactly which rule to add. Five rules you actually use beat fifty you wrote pre-emptively.
Once that workflow is reliable, two upgrades are worth knowing about. You can turn a proven workflow into a reusable Skill the whole team draws on, so everyone produces to the same standard without memorising it. And connectors can now pull live context, such as your CRM data, straight into Claude, so it can answer questions grounded in your actual pipeline rather than the abstract. Both are next steps, not starting points. Get one workflow working first.
It won't originate your positioning. Claude is excellent at analysing existing positioning, spotting gaps, and documenting strategy in a tidy framework. It cannot produce the outlier insight that makes positioning distinctive, because that comes from judgement and taste you hold and it doesn't. It accelerates the work around the thinking, not the thinking itself.
It can't interview your customers. It will synthesise a transcript beautifully, but it can't read the room, hear what's underneath an answer, or push harder on the interesting thing someone just said. That stays human.
And while it's now common to hear that Claude "can't see anything current," that's out of date. Claude.ai has had web search as a toggle across all plans since 2025, and connectors can bring in live data. What's still true is narrower but important: it doesn't know your numbers unless you give it access, and it will confidently fill any gap you leave. Treat what comes back as a strong first draft to verify, never a finished artefact.
The other honest limit is a team one. Individual adoption is easy. Consistency across a team is where most marketing functions stall, because without a shared Project and shared context files, three people produce work in three different registers. The fix is the same as the original lesson: put the context somewhere shared, and let everyone draw from it.
The shift isn't a better prompt. It's a system that compounds. A team six months into running Claude properly has dramatically better infrastructure than a team in month one, even though they're using the identical tool. The marketers who pull ahead aren't the ones who found a magic prompt. They're the ones who stopped starting from scratch.
That's the same instinct behind becoming genuinely AI-native rather than just AI-curious: treating AI as part of how the work runs, not a gadget you reach for occasionally. If you want the wider view of where AI fits across the marketing function, our pillar guide AI in Marketing: A B2B SaaS Marketer's Field Guide maps the full picture, and How to Use AI for Marketing: A 90-Day Playbook takes the phased route across every tool, not just Claude.
If you'd rather learn this with a room full of marketers doing the same thing, our Claude AI for Marketers Sprint inside SaaStrix walks through exactly this: setting Claude up as a system, building the context files, and turning real B2B marketing tasks into workflows you reuse. It's taught through actual marketing work, not generic prompt theory. You can join SaaStrix here.
What is the best way to use Claude for B2B marketing?
Treat it as a system, not a chatbot. Create a Claude Project, load it once with your brand voice, your real audience language, and your editing standards, then run repeatable workflows on top: briefs, first drafts, research synthesis, repurposing. The context you give it, not the cleverness of the prompt, decides the quality of what comes back.
Why does Claude output feel generic?
Because a blank chat knows nothing about your company, so it gives you the average of everything it has read. Average input, average output. The fix is not a better prompt; it is persistent context, which is why prompt lists fired at an empty Project tend to disappoint.
What is a Claude Project and why does it matter for marketers?
A Project is a persistent workspace with a knowledge base and standing instructions that every conversation inside it inherits automatically. You load your context once, and from then on Claude already knows who you are before you have typed a word. The blank-page tax disappears and output quality stops depending on what you remembered to paste in that day.
Which marketing tasks is Claude genuinely good at?
Anything repeatable, context-heavy and writing-centred: content briefs, first drafts of long-form content, research synthesis across messy inputs, repurposing a pillar piece into channel-ready variants, and quality-checking copy against your house style. It is strongest on the grind, not the final polish; the sharp claim and the taste are still your job.
What can Claude not do for marketers?
It will not originate your positioning, because the outlier insight comes from judgement it does not hold, and it cannot interview your customers. It also does not know your numbers unless you give it access, and it will confidently fill any gap you leave, so treat what comes back as a strong first draft to verify, never a finished artefact.
SaaStrix is where B2B marketers become agentic marketing leaders, and you can try it free for five days.
.svg){kind=icon}