Daphne Tay had spent three years building a language learning platform.
Legal terminology in French. Financial jargon in Mandarin. Compliance language across a dozen European dialects. Her team had invested thousands of hours researching domain-specific vocabulary, building databases of technical terms that didn’t exist in standard translation tools.
Then customers started asking for something different.
“Can your platform translate our documents with the formatting intact?” One client was expanding into Europe and needed product collateral translated into eight languages. Not vocabulary lessons. Actual PowerPoint decks with images and layouts preserved.
Another client needed training materials. Another needed compliance documents. The pattern was clear: they wanted visual translation, not language learning.
Daphne kept saying no. She’d built a learning platform, not a document translator. That wasn’t the vision.
But the requests kept coming.
Here’s what might surprise you: The hardest part wasn’t building the new technology. It was killing the original vision after three years of investment. Daphne had to decide whether to keep building what she thought the market needed or pivot to what customers were actually willing to pay for.
She chose the pivot. Today, Bluente translates documents for Fortune 500 manufacturers, Big Four law firms, and pharmaceutical companies expanding across Europe. They process translations 90% faster and 90% cheaper than human translators while preserving every image, format, and layout detail.
I sat down with Daphne to understand how she made this work. What I learned fundamentally changed how I think about market selection, trust-building in AI sales, and staying current on AI from 7,000 miles outside Silicon Valley.
Why Most AI Startups Choose the Wrong First Market
When Daphne decided to pivot from language learning to enterprise translation, she didn’t just chase the biggest opportunity. She ran a framework. That framework:
The Vertical Selection Scorecard: TAM × Volume × Existing Expertise
Most founders pick one of these factors. Daphne insisted on all three.
TAM (Total Addressable Market): Legal and finance weren’t just big markets. They were massive. The global language services industry hit $71.7B in 2024 and is projected to reach $75.7B in 2025. Within that, legal and financial translation represents one of the highest-value segments because the cost of errors is catastrophic.
Translation Volume: These industries need constant, high-volume translation work. Marketing collateral, training materials, compliance documents, product sheets. And when companies expand to Europe, that volume explodes. The EU has 24 official languages, which means 552 possible linguistic combinations (each language translates into 23 others).
Existing Expertise: This is where most founders miss the opportunity. Daphne’s legacy language learning platform had spent years building databases of legal and financial terminology across languages. That wasn’t wasted effort. It became training data for fine-tuning Bluente’s translation engine.
She didn’t abandon three years of work. She repositioned it as a competitive moat.
The Anchor Customer Strategy
Once Daphne selected legal and finance as her beachhead, she got tactical about customer sequencing.
She didn’t chase easy wins. She targeted the Big Four law firms in Singapore first.
Why? Because in high-stakes industries, your first customers matter more than your first revenue.
Landing the Big Four established immediate credibility. When Daphne pitched other enterprise clients, she could open with: “We work with [Big Four firm names].” That single line bypassed months of trust-building.
You need to close anchor customers before you have other anchor customers. It’s a classic cold-start problem. Daphne solved it by leaning into her differentiated positioning around domain expertise and security.
Steal This: The Vertical Selection Diagnostic
Before you commit to a market, answer these three questions:
- TAM Reality Check: Is this market large enough to support a venture-scale business? Be honest. Niche can work, but it needs to be a big niche.
- Volume Validation: Do customers in this vertical need your product once or repeatedly? Recurring, high-volume needs create defensible revenue.
- Unfair Advantage Audit: Do you have existing expertise, data, relationships, or technology that gives you a head start in this vertical? If not, why will you win?
If you can’t answer “yes” to all three, keep looking.
How Bluente Sells AI to the World’s Most Risk-Averse Buyers
Selling AI to legal and finance teams is like selling parachutes. The cost of failure is existential.
A mistranslation in a legal contract could void agreements worth millions. A translation error in financial disclosures could trigger regulatory violations. These buyers don’t care how fast or cheap your tool is if they can’t trust it.
Daphne built trust through a three-part stack: Domain Expertise + Zero Data Retention + Anchor Customers.
Part 1: The Domain-Specific Fine-Tuning Advantage
Remember that legacy language learning platform? Daphne had spent years researching legal and financial terminology across languages. Not surface-level vocabulary. Deep, nuanced, context-dependent terminology that changes meaning based on jurisdiction.
That research became a proprietary database. Bluente used it to fine-tune their translation engine specifically for legal and financial documents.
When Daphne pitched clients, she didn’t lead with “we use AI.” She led with “we’ve spent years building legal and financial language models that understand your domain.”
She positioned the AI as trained by domain experts, not as a generic tool.
Part 2: Zero Data Retention as a Feature
Enterprise legal and finance teams have a recurring nightmare: confidential documents leaking through third-party tools.
Bluente’s security positioning is simple. Zero data retention. Every document uploaded to Bluente’s portal is deleted from their servers within 24 hours. No storage. No training on customer data. No lingering copies.
This single policy eliminated 80% of security objections during the sales cycle. It gave legal and compliance teams the assurance they needed to pilot the tool.
However, this creates a tradeoff. Bluente can’t use customer data to continuously improve their models. They’ve accepted that constraint because the trust gain outweighs the optimization loss.
Part 3: The Netherlands Manufacturing Case Study
Here’s what trust-building looks like in practice.
Bluente’s client, an industrial manufacturing company based in the Netherlands with offices across Europe, needed training materials translated into French, Italian, Greek, and a dozen other European languages.
Before Bluente: Employees would copy and paste each line from PowerPoint decks into translation tools. Translate line by line. Manually reformat images. Fix broken layouts. Re-export. The process took days per document and required multiple people.
After Bluente: Upload the PowerPoint to Bluente’s portal. Select target languages. Wait 2-3 minutes. Download fully translated documents with images and formatting completely intact.
Result: 90% time savings. Documents were immediately usable without manual fixes.
The manufacturing company initially piloted Bluente for one training deck. Within three months, they’d migrated their entire European training library to the platform.
That’s the pattern Daphne sees across enterprise clients. Pilots convert to full deployments once teams realize the translation output genuinely requires no post-editing.
Why This Matters for AI Founders
McKinsey reports that 71% of organizations now regularly use generative AI in at least one business function (July 2024 survey). But adoption doesn’t mean trust.
Enterprise buyers are still terrified of AI failures. If you’re building for high-stakes industries, your positioning has to address that fear head-on. You can’t just talk about speed and cost. You need to demonstrate domain expertise, security protocols, and proof from comparable clients.
When to Fire Your Agencies and Hire In-House
By Q4 last year, Bluente had two agencies humming along nicely. One handled outbound. Another managed SEO and AEO (AI Engine Optimization).
Both were performing. Traffic was climbing. Deals were closing.
Then Daphne went to events and kept hearing the same feedback: “I wish I’d known about you earlier. I’d never heard of your brand before.”
That’s when she realized the gap. They needed to move faster on demand generation. But agencies, no matter how good, can’t iterate at startup speed.
Daphne started searching for two roles: a chief of staff/product manager to automate internal systems, and a founding growth lead to accelerate marketing. As she interviewed candidates, a pattern emerged.
Most of the systems she wanted to automate were actually go-to-market systems. Outbound sequences. SEO workflows. Content generation. Lead qualification.
Then, at a conference in San Francisco, an advisor mentioned a role she’d never heard of: GTM engineer. Someone who combines marketing strategy with technical implementation. Someone who can build and automate the entire demand generation stack.
That was exactly what Bluente needed.
The Agency vs. In-House Decision Framework
Daphne made the call to bring outbound and SEO in-house through a GTM engineer hire. Here’s why:
Speed of experimentation: Agencies operate on weekly or monthly cycles. Startups need to test hypotheses daily. When you want to try a new outbound sequence, a different ICP segment, or an experimental content angle, having someone internal means you can ship changes in hours instead of days.
Depth of capability: During interviews, Daphne realized candidates could do far more with outbound than her agency was delivering. Not because the agency was bad. Because agencies optimize for what works across multiple clients. In-house teams optimize for what works for your specific ICP.
System ownership: Agencies hand you dashboards and reports. Internal teams hand you the actual systems. When Daphne’s GTM engineer joins, they’ll own the entire stack: the workflows, the automations, the data pipelines. That ownership creates compounding advantage over time.
However, Daphne was careful about timing. She’d learned from her first product that hiring too early kills momentum. Her philosophy now: wait until you’re absolutely certain about the role before bringing someone on board.
What Makes a Great GTM Engineer
During Bluente’s hiring process, Daphne tested candidates in two ways:
- Depth of execution: The best candidates used AI to build entire systems during the interview process. They’d show up with working prototypes: “Here’s the outbound sequence I’d run. Here’s the automation I’d build. Here’s how I’d instrument it.”
That level of detail gave Daphne confidence they could execute, not just strategize.
- Tool fluency: Daphne asks every candidate: “What’s the latest tool you learned in the last week?”
Some candidates mention Claude or ChatGPT. The best candidates mention tools from three days ago. Tools Daphne hasn’t heard of yet.
Finding people who combine strategic marketing thinking with systems-building execution is challenging. Most marketers can’t code. Most engineers don’t understand go-to-market. The intersection is rare.
If you’re trying to hire a GTM engineer, look for evidence of both: strategic frameworks and technical implementations they’ve actually shipped.
Staying Current on AI from 7,000 Miles Away
Operating from Singapore, Daphne worried about being disconnected from Silicon Valley’s AI ecosystem.
She built a system to stay current that actually works better than living in San Francisco. Here’s how she does it:
System 1: Let Social Algorithms Work for You
Daphne follows AI tool creators on TikTok. Not LinkedIn thought leaders. Not Twitter theorists. People who actually demo tools.
Once she started engaging with that content (watching, liking, commenting), TikTok’s algorithm started surfacing more. Within weeks, her feed became a curated stream of emerging AI tools, builder demos, and technical breakdowns.
In other words, she turned TikTok into a personalized AI research engine without manually searching for content.
System 2: Follow Founders 2-3 Stages Ahead
Daphne tracks founders who are building in public on LinkedIn and X. Not peers at her stage. Founders running companies 2-3 stages more mature than Bluente.
When they share updates, she reads for tech stack mentions. When they post about new tools, she screenshots for later research. And when she’s curious, she comments directly: “Can you share what tech stack you use for this?”
Most founders respond. Those responses have led Daphne to discover tools like cursor, Windsurf, and deployment platforms she wouldn’t have found through traditional research.
System 3: Leverage Globally-Connected Investors
Bluente’s investors operate globally. They see best practices across portfolio companies in the US, China, and Europe.
Daphne structures monthly and quarterly check-ins to include a simple ask: “What are five tools or practices you’re seeing work across your portfolio?”
Her investors share insights from companies deploying AI at scale in different markets. That global perspective is actually more valuable than being physically in San Francisco because she gets pattern recognition across geographies instead of just Valley groupthink.
The Contrarian Take on Location
Think you need to be in San Francisco to build an AI startup? Here’s why that’s outdated.
Social media algorithms surface the same information globally. Founder communities share learnings publicly. Investors connect portfolios across regions. And 76% of consumers prefer buying products with information in their native language (study of 8,709 consumers across 29 countries), which means geographic diversity in founder location actually creates market insight advantages.
Daphne’s Singapore location gives her firsthand understanding of Asian and European enterprise buying behavior. For a global product, that’s strategic positioning.
Why YC Still Matters (But Not for the Reason You Think)
If Daphne were starting Bluente from zero today, she’d join Y Combinator immediately.
Not for the investment. Not for the prestige. For the first customer pool.
The YC network excels at pushing products to growing startups within the cohort ecosystem. When you’re looking for your first 10-20 customers, warm introductions from batch-mates matter infinitely more than cold outbound.
Daphne watched YC companies in her sector close their first enterprise deals through cohort connections. Those deals happened in weeks instead of months because trust was pre-established through the YC network.
Most accelerators are valuable at earliest stages. Bluente is now too late-stage to join. The window for maximum accelerator value is tight: post-product validation, pre-significant revenue.
However, Daphne’s point applies beyond YC. If you’re building B2B, warm introductions >> cold outbound for your first customer pool. Find the network that gives you access to your ICP, then optimize for relationship density over reach.
The Regional GTM Playbook: When Remote Sales Actually Lose Deals
Remote-first sales doesn’t work everywhere.
Bluente’s GTM strategy combines outbound with in-person events. Not because Daphne loves conferences. Because certain markets (Singapore, Middle East, parts of Europe) still require face-to-face meetings for enterprise deals.
Different markets have different expectations for establishing trust. In the US, you might close a deal on Zoom. In Singapore, buyers want to meet you in person first.
This isn’t about inefficiency. It’s about cultural buying behavior. Some regions equate in-person presence with commitment and legitimacy. A Zoom call signals you’re not serious about the market.
Bluente complements outbound sequences with strategic event attendance. They meet buyers face-to-face, let them see the team, and establish that they’re “out there” in the market.
Does your target market require in-person sales? Ask your first 5-10 prospects how they prefer to evaluate new vendors. If “in-person demo” or “meet the team” comes up repeatedly, adjust your GTM motion accordingly.
The Founder-Led Content Playbook for “Boring” Categories
Daphne started posting founder content on LinkedIn last year after watching successful founders build in public.
The results weren’t viral posts or massive engagement. But something interesting happened: people would come through other channels (referrals, outbound responses, event follow-ups) and mention they’d seen her on LinkedIn.
The content created indirect brand-building. It gave Bluente a personality beyond its product. It signaled that a real person was running the company, not a faceless corporate entity.
Daphne’s doubling down this year. She’s working to identify content pillars that make translation engaging despite it being a time-triggered category. Translation isn’t like sales tools where daily relevance makes content easy to create. People only think about translation when they urgently need it.
Her current hypothesis is that influencer partnerships will help uncover what resonates. By collaborating with industry voices, she can test different content angles and see what drives awareness in a “boring” B2B category.
The Attribution Challenge
Founder content impact is indirect and hard to measure. You won’t see direct conversions. You’ll see people say “I came across you somehow on LinkedIn” when they come through other channels.
That vague attribution frustrates data-driven founders. But the alternative (no founder presence) means you have zero personality in the market. And in crowded AI markets, personality is a tiebreaker.
If you’re building in a time-triggered or low-engagement category, founder content is about staying top-of-mind for the moment when your audience urgently needs you. Not about generating immediate pipeline.
What Daphne Would Do Differently
I asked Daphne: if you were starting Bluente from zero today, what would you change?
Three things:
- Rapid prototyping over custom builds: The AI tool ecosystem now lets you build functional prototypes in days. Daphne would test market demand with off-the-shelf tools first, then invest in custom development only if customer willingness-to-pay justified it.
- Join a strong-network accelerator early: Not for investment, but for access to the first customer pool. Warm introductions from batch-mates would have accelerated Bluente’s first 20 deals by months.
- Follow customer demand signals faster: The pivot from language learning to enterprise translation should have happened sooner. Daphne spent months trying to make the original vision work when customers were clearly asking for something different.
Customer pain beats founder vision every time. The faster you internalize that, the faster you’ll build something people actually pay for.
Resources
- Bluente – Try visual translation for your enterprise documents
- Nimdzi 100 Report – Global language services industry data
- EU Language Policy – European Union official languages
P.S. Here’s The Build vs. Buy Decision I’m Watching
I’ve been thinking about Daphne’s build-vs-buy framework a lot lately.
In my fractional and interim CMO work, I see founders making this decision constantly. Not just for product features, but for GTM systems, content engines, analytics stacks. The pattern is always the same: founders want to build custom because it feels like they’re creating a moat.
But most of the time, buying or licensing gets you to market validation faster. And market validation is the only thing that actually matters at early stages.
If you’re wrestling with a build-vs-buy decision right now (product, GTM, ops, anything), here’s my diagnostic:
- Can you validate customer willingness-to-pay without building this custom?
- Does building this custom create defensible differentiation customers will actually pay for?
- Will building this delay your ability to test with real customers?
If you answered yes to question 3 and no to questions 1-2, you should probably buy.
And if you need someone to help you think through GTM system design with that same engineering rigor Daphne applies to product decisions, reach out. I work with technical founders who want to apply systematic thinking to revenue operations. It’s basically GTM engineering, and it’s what I do.
– Lillian
