Revolut's Creative Team: Using AI for On-Brand Copy Prototyping.
How Revolut's Designers Use AI to Prototype Copy Without Waiting for Copywriters
A designer at Revolut finishes a marketing mockup at 4 p.m. on Thursday. The layout is strong. The visual hierarchy works. But there's one problem: the headline reads "Lorem ipsum dolor sit amet."
She needs copy to know if the design actually works. Is the headline too long for this layout? Does the message fit the visual? Would this resonate with customers?
She requests copy from the writing team. They're two weeks out. The project ships in ten days.
This standoff happens at creative teams everywhere, dozens of times a week. Designers can't move forward. Copywriters are overwhelmed. Projects stall.
Gareth Morgan, Head of Global Brand Design at Revolut, built a solution that breaks this cycle without compromising quality.
The Risk Spectrum: Where AI Fits (and Doesn't)
Before explaining Revolut's workflow, it's essential to understand where AI-generated copy belongs on the risk spectrum.
Internal prototyping and iteration (AI-appropriate) Using AI to generate placeholder copy for design mockups, internal stakeholder reviews, and early-stage concepts. Human copywriters see the context before writing final copy.
Internal drafts and strategy documents (AI-assisted, human-finalized) First drafts of strategy docs, internal presentations, or planning documents. Always reviewed and edited by a human before sharing.
Customer-facing marketing copy (AI-assisted, human-written) Headlines, body copy, CTAs, and social captions. AI may help with ideation or structure, but a professional copywriter writes and approves the final version.
Regulated and high-stakes content (No AI) Legal disclaimers, terms of service, financial disclosures, medical information, or any copy with compliance requirements. Only human experts touch these.
Revolut's workflow lives firmly in the first category. AI generates realistic placeholder copy so designers can iterate, but every word customers see has been written or approved by a professional copywriter.
The Solution: AI as Prototyping Partner
Designers at Revolut now use AI tools to generate on-brand placeholder copy when starting new designs. Instead of lorem ipsum, they prompt ChatGPT or Claude with brand guidelines to create realistic text.
This transforms how quickly they can iterate. Because the copy is on-brand and contextually relevant, even if not final, stakeholders can evaluate whether the design actually works. Is the headline too long? Does the message fit the visual? Would this resonate with customers? These questions are impossible to assess with lorem ipsum.
When copywriters enter the process, they see designs with realistic copy. They understand the direction immediately and can refine from there. Often, they'll keep the structure and tone while tightening the language or adjusting for brand voice nuances.
"AI doesn't replace designers or copywriters," says Gareth. "But on our team, it's a valuable tool to help design teams move forward without resorting to generic placeholders when copywriters are overloaded or unavailable."
The Results
Design iterations moved 40% faster. Stakeholder feedback became more specific because people were evaluating realistic designs instead of abstract layouts. Copywriters spent less time on basic placeholder work and more time on high-impact messaging.
The cultural shift mattered most. Designers felt less blocked and could maintain momentum. Copywriters appreciated seeing designs with context instead of blank layouts requesting "TK headline here." Cross-functional collaboration improved because everyone was looking at realistic prototypes earlier.
Perhaps most surprising: introducing AI into the process actually strengthened human collaboration, not weakened it.
Implementation: The Revolut Starter Kit
Five Prompt Templates to Start With
Product launch headline "Write three headline options for [product/feature name]. Tone: [your brand tone]. Target audience: [specific persona]. Highlight this benefit: [core value prop]. Length: 6-8 words."
Feature explanation body copy "Write 2-3 sentences explaining [feature name] for [target audience]. Tone: [your brand tone]. Focus on the customer benefit, not technical details. Reading level: 8th grade."
Conversion-focused CTA "Write five CTA button text options for [desired action]. Tone: [your brand tone]. Context: User is [describe user state]. Length: 2-4 words max."
Social media caption "Write an Instagram caption for [campaign/announcement]. Tone: [your brand tone]. Include: [key message]. Length: 150 characters max. No hashtags."
Email subject line "Write five subject line options for an email about [topic]. Audience: [persona]. Tone: [your brand tone]. Goal: [open rate / click rate]. Length: 6-10 words."
Three Brand Voice Components to Define First
Before designers start prompting AI, document these clearly:
Tone attributes with examples Not just "friendly and professional." Be specific: "Confident but not arrogant. Example: 'Send money abroad in seconds' not 'We're the world's best money transfer app.'"
Vocabulary: Use and avoid lists Use: "fast, simple, secure, transparent." Avoid: "revolutionary, game-changing, disruptive, leverage." Include 15-20 terms in each list.
Sentence structure preferences "Active voice. Short sentences (15 words average). Lead with benefits, not features. Never use jargon without explaining it first."
Four Times Designers Should Skip AI and Ask a Copywriter Directly
Anything customer-facing in its final form If this copy will go live without copywriter review, don't use AI. No exceptions.
Messaging that requires nuance or sensitivity Financial hardship, security breaches, service outages, pricing changes, these need human judgment.
Copy that must comply with regulations Legal disclaimers, financial disclosures, data privacy notices, terms of service.
Brand-defining moments Taglines, mission statements, major campaign headlines, these shape your brand and deserve dedicated creative thinking.
When this Approach Works (and When it Doesn't)
This workflow makes sense when your brand voice is clearly documented. If you don't have guidelines, AI will produce generic output. Define your voice first.
It requires designers trained to prompt effectively. This isn't typing random questions into ChatGPT. Designers need to understand how to craft prompts that generate on-brand results.
It works when copywriters are genuinely overloaded, not when they have capacity. If your copywriting team isn't stretched thin, use them directly. This workflow solves a bottleneck problem.
It absolutely requires quality control checkpoints. Someone with copywriting expertise must review AI-generated text before it's used, even internally for stakeholder presentations.
And it only works with copywriter buy-in. If copywriters see this as threatening rather than helpful, it fails. Position it clearly: "This frees you from basic placeholder work so you can focus on high-impact messaging."
Common Mistakes to Avoid
Don't skip the human review, even when AI output looks polished. "Sometimes, people are so impressed by what AI can do that they don't think critically," says Gareth. "The AI 'wow factor' can fool people into thinking they've created a good end product. We need to move past that, question the output, push it, and squeeze it to get to the really good stuff."
Don't let AI-generated copy drift into customer-facing contexts without explicit copywriter approval. The boundaries must be crystal clear: prototyping yes, publishing no.
Don't forget to update your documentation as you learn. Include examples of effective prompts, common AI failure modes your team has discovered, and evolving guidelines on when to use AI versus when to go straight to a copywriter.
Good Prompt vs. Bad Prompt
Bad prompt: "Write a headline about sending money abroad"
Why it fails: Too vague. No brand context, no audience, no constraints. You'll get generic output.
Good prompt: "Write three headline options for Revolut's international transfer feature. Tone: confident, straightforward, benefit-focused. Target: small business owners sending supplier payments abroad. Highlight speed and transparency. Length: 6-8 words."
Why it works: Specific context, clear constraints, brand tone guidance. The AI has enough information to generate on-brand options.
——————————————
Bad prompt: "Make this sound better: 'With Revolut you can send money to other countries very easily and quickly with low fees.'"
Why it fails: AI will polish the sentence but won't restructure for impact. You'll get incremental improvement, not strategic thinking.
Good prompt: "Rewrite this value proposition with active voice, leading with the customer benefit. Original: 'With Revolut you can send money to other countries very easily and quickly with low fees.' Make it punchy and confident. Three options, 8 words max each."
Why it works: Clear direction on structure, specific constraints, multiple options to choose from.
The Broader Principle
Revolut's approach illustrates something bigger than copywriting: AI excels at maintaining creative momentum, not replacing human expertise.
"Brands have always needed guidelines to keep their content cohesive. I don't think AI is a threat to that, it's just one more tool to use to a high standard," says Gareth.
The pattern applies everywhere: designers can use AI to generate layout variations, but a senior designer still art directs. Marketers can use AI for first-draft strategy, but leadership still makes final decisions. Researchers can use AI to synthesize findings, but experts still validate conclusions.
AI accelerates the early stages so humans can focus on the high-value final stages where judgment, nuance, and craft matter most.
Bottom Line
AI won't replace your copywriters. But used within clear boundaries (prototyping yes, publishing no) it eliminates the waiting that slows creative teams down.
The key is knowing exactly where AI belongs on the risk spectrum. Keep it in the prototyping zone. Give designers the tools to maintain momentum. And keep humans firmly in charge of every word customers see.
