Data platform vendor Databricks is using Grammarly’s generative AI-based writing and editing tool to improve its external communications and, in the process, saving about $1.4 million a year in productivity.

Databricks also is seeing a 1,994% return on investment, with 71% of its communications so far this year being improved.

The 10-year-old company brought Grammarly into the equation last year, when executives saw the business continuing to grow and the resulting pressure on a relatively small editorial group that was having to deal with rapidly expanding amounts of external communications demands.

Grammarly’s tool uses generative AI techniques to improve writing, not only catching spelling and grammatical errors but also by suggesting ways to make the writing more effective, engaging or clearer. There also are recommendations in such areas as tone and sentence variety, and it can review the text for plagiarism.

Grammarly also includes style guides for organizations. Databricks was able to integrate a custom company style guide that gives users real-time and on-brand suggestions as they write, according to a Grammarly case study.

Generative AI a Good Match for Writing

Writing generation and editing has been among the top use cases for generative AI, with its natural language processing capabilities, since the launch in late 2022 of the ChatGPT chatbot by OpenAI. A quick search for generative AI writing tools quickly brings up almost 20 results for creating text based on prompts and improving it, including ChatGPT.

For Databricks, Grammarly takes the content being created by the company’s writers and editors and offers way to improve it, saving the editorial team time and enabling it manage the content flow without having to add staff.

“I used to spend 20 to 30 hours a week reviewing, rewriting, and coaching,” Neil Hamilton, head of editorial at Databricks, said in a statement. “Grammarly has cut that by at least half, taking a lot of the grunt work off our plates so we can focus on what you could call the ‘magic.’ That’s allowed my team to scale our work in a way we couldn’t possibly have without it.”

A Symbiotic Relationship

The relationship between the two companies has been at least a year in the making. Grammarly in 2023 migrated from its in-house data lake to Databricks’ Data Intelligence Platform, which includes its Delta Lake lakehouse, Apache Spark for data processing, and Unity Catalog for data governance. According to Databricks, the move resulted in 110% faster querying – at a 10% drop in the cost to ingest – than a data warehouse and enabled Grammarly to make its 5 billion daily events available for analytics in less than 15 minutes rather than four hours.

Databricks is no stranger to generative AI, with the company boasting that its platform helps to unify the AI development lifecycle, from data collection and preparation to model development to monitoring. The platform includes MLflow for development tracking and evaluation of large-language models and DataBricks Model Serving for deploying LLMs.

The company’s Foundation Model APIs, now in public preview, are used for open LLMs while the platform also enables users to access third-party models hosted outside of Databricks.

According to a ROI analysis from Nucleus Research, Databricks turned to Grammarly after an email that was about to be sent to thousands of customers was flagged for containing grammatical errors, through which “the need for technology to streamline communications was brought into stark relief,” the researchers wrote.

“Databricks needed to standardize, track, and manage written external communications and documents,” they wrote. “Databricks could have expanded its editorial staff, but that carried high headcount expenses and was not a cost-effective method to scale operations with customer growth.”

The Turn to Grammarly

They opted instead for Grammarly Business, which not only led to significant savings but also enabled Databricks’ marketing teams to reduce editing times for thousands of documents by half and saved copywriters and editors 540 hours. The customer experience and support team reduced the time to resolution by 25% and the grew the average number of cases closed by 10% to 15%.

Another advantage is that most of its 100-plus technical support engineers are not native-English speakers, so Grammarly lets them better work with customers not only through language but also tone and empathy.

A range of Databricks departments use Grammarly, including marketing, sales, finance, IT, engineering and legal.

Databricks, which has more than 10,000 customers and 6,000 employees, saw payback of its investment in Grammarly in 1.2 months, according to Nucleus.

Publishing has long had safeguards for addressing problems in writing, Databricks’ Hamilton said in a Grammarly case study, adding that what Grammarly does is automate the process.

“It’s sort of invisible, and that’s a good thing,” he said. “People are using it automatically. It’s giving helpful suggestions and coaching without you needing to seek out its counsel. It’s always there to help people write better.”