How to Build a Successful Language AI Product

How to Build a Successful Language AI Product - Main

A new era of language productization is underway. SaaS (software-as-a-service) platforms now deliver a multitude of language capabilities — translation, dubbing, subtitles, transcription, voice cloning, and more — directly to individual users and businesses.

From product-led foundations, many SaaS platforms are graduating from appealing primarily to individual consumers to being deployed across large enterprise clients. At SlatorCon Zurich on October 4, 2023, Slator’s Managing Director Florian Faes moderated a panel discussion between three SaaS platforms, each at a different stage of this journey.

Céline Daley, VP of Customer Success at DeepL, André Bastié, CEO of Happy Scribe, and Jan Cizmar, CEO at Tolgee, shared with the SlatorCon audience what it takes to, firstly, create a product that people love and, secondly, scale a solution that wins over global enterprises.

Daley opened the discussion by explaining DeepL’s rationale for scaling from a freemium tool to an enterprise solution. “It comes from a desire to broaden the impact of what we do, to go from the individual, to the team, to the department, to the enterprise,” she explained.

“We see a lot of need for collaboration,” Daley said, adding. “How do you maintain language consistency, control what you’re translating, and ensure your brand voice is heard?”

The machine translation platform, which achieved unicorn status in January 2023, has “been scaling steadily double a year every year across most metrics,” according to the Customer Success VP. 

The company’s large base of individual users is both its foundation and an access point for broader enterprise engagement, in Daley’s view. “The people who love and use our product are in many ways a source of interest for us. Where are they working? What teams are they on? Could we sell a larger package to them?”

Creating Champions

A SaaS product at a much earlier stage of this journey is Tolgee. A developer-focused localization platform, Tolgee allows users to translate web applications and mobile apps inside a developer environment, without modifying the code.

Tolgee’s growth is product-led. “We’re targeting developers and trying to catch individuals at the beginning of their journey,” Cizmar told the SlatorCon audience. “We’re also open-source, which is important. It means there is no barrier to start using the platform. People can just start and then tell other people about it.”

One major benefit of this approach is, according to Cizmar, “You can just focus on building the product, instead of hiring a lot of salespeople.”

Transcription and subtitling platform Happy Scribe’s initial growth was similarly product-led. In fact, CEO Bastié told the SlatorCon audience, “We had no idea how to sell to enterprises at first. The only thing we were passionate about was building products that people really love.”

Bastié unpacked how the platform attracts users via landing pages and then drives growth via referrals. The CEO added that Happy Scribe is now adding new growth strategies to the mix. “We’re experimenting and experiencing adding a sales motion on top of a self-serve motion,” Bastié said.

“You can just focus on building the product, instead of hiring a lot of salespeople.” – Jan Cizmar, CEO, Tolgee

Explaining why the transcription platform is now gaining traction among enterprises, Bastié said, “This is a tool that doesn’t require expertise in languages to start leveraging the power of languages within an organization.”

How to Build a Successful Language AI Product - In Line

It’s Complicated

Enterprise needs are, of course, quite different from those of individual users. DeepL’s Daley cited permissions and usage analytics, as well as ISO certification and SOC II compliance, as some of the “very different set of issues to tackle” to create an enterprise solution.

Developing a pricing model can also be tricky. Enterprises “want to make sure they have predictability in their cost,” said Daley. However, “the breadth that [DeepL] can have across an enterprise can get quite complex.”

“There are going to be a lot of casual users who use DeepL once a week or once a month,” Daley elaborated. “Then there’s going to be the localization or legal or marketing department that use us hundreds of times a day, plus API use.”

The only thing we were passionate about was building products that people really love.” – André Bastié, CEO, Happy Scribe

DeepL’s current model mixes seat-based and consumption-based pricing. Happy Scribe has a similar model, mixing “pricing that is user-based paired with pricing that is more usage-based,” according to Bastié. Tolgee began with tiered subscription plans before switching to pay-as-you-go.  

“You have to experiment and you will fail a lot of times because it’s just complicated,” said Cizmar, adding, “It’s not easy because you have various use cases and different types of users and clients.”

“[Pricing] is a forever exercise to continue to iterate on,” agreed Daley.

Launch and Learn

As DeepL has scaled, the company has continued its investment in research but also “rounded out the company on the commercial side”, according to Daley, who added, “It’s about building out those go-to-market teams to ensure we can support our business buyer.”

German-headquartered DeepL currently sells in a limited number of countries, adding South Korea to the list in 2022. 

Daley outlined the launch process, explaining, “We tend to launch the language first, get feedback on the quality of it, see what usage is like, see what kind of interest there is in the business perspective, validate the market need, then go to market with PR and marketing.”

“Then there’s going to be the localization or legal or marketing department that use us hundreds of times a day, plus API use.” – Céline Daley, VP of Customer Success, DeepL

How to Build a Successful Language AI Product

HappyScribe, which CEO Bastié describes as a hybrid solution that combines AI with human capabilities, has taken a bold and exploratory approach to market expansion. “In June, we decided to launch globally to all languages. We started partnering with agencies all around the world.”

The result? “We learned a lot,” the CEO said. For example, the company learned, on the go, “how to make Chinese captions 12 characters per line.”

Bastié sees international expansion as a mechanism for deepening the company’s knowledge of markets and languages. “We don’t really launch for business purposes. It’s more about learning a lot,” the CEO said.

The wide-ranging SlatorCon SaaS in Language AI and Tech panel also explored topics such as quality evaluation, market perceptions of AI, and the impact of SaaS products on services demand.

For those who could not attend in person, SlatorCon Zurich 2023 recordings will be available in due course via our Pro and Enterprise plans.