Blackbaud unveils Lantern, a fundraising-specific language model with Databricks
Blackbaud said Tuesday at bbcon 2026 that it launched Lantern, a domain-specific language model built with Databricks and post-trained on fundraising workflows so social-impact teams get intelligence tuned to giving behavior rather than consumer shopping patterns.
Nonprofits have been running general chatbots on donor data. Lantern is Blackbaud’s attempt to put a model that speaks fundraising — who gives again, who gives more, and how long the relationship lasts — inside Raiser’s Edge NXT and Agents for Good, on the Databricks platform where Blackbaud says the model is built.
On Tuesday, 29 September 2026, Blackbaud unveiled Lantern at bbcon 2026. The company trades as BLKB on the Nasdaq. Blackbaud calls Lantern the first domain-specific language model purpose-built to optimize fundraising intelligence. A language model is software that reads and writes text. Domain-specific means this one is aimed at fundraising, one field, instead of every topic on the internet. The newsroom post is dated September 29, 2026, and does not print an hour. PR Newswire carried the same release, datelined Columbus, Ohio. That page stamps it Sep 29, 2026, 10:15 ET, which is 10:15 a.m. Eastern. Those lines are Blackbaud’s.
Blackbaud says large language models are underserving social impact. It says 85 percent of social impact professionals use AI in their daily jobs, and only about 10 percent of organizations are seeing significant returns. Eighty-five percent is about seventeen people in twenty. About 10 percent is about one organization in ten. Blackbaud’s phrase for the return is “significant dividends.” It says Lantern is built to close that gap, through a strategic collaboration with Databricks. The model, Blackbaud says, rests on more than four decades of fundraising workflows, decisions, and results, so it understands giving behavior. Giving is a donation. The contrast Blackbaud draws is buying behavior, a purchase. Those figures and that contrast are Blackbaud’s. The release does not say how the 85 percent or the 10 percent were counted.
How Blackbaud says the model was built. With Databricks, it is combining leading open-weight models on the Databricks Data + AI Platform with more than 40 years of social impact intelligence. An open-weight model is one whose learned settings are published, so another company can keep training it. The Data + AI Platform is Databricks’ system for holding that data and running the training. Post-training is the extra teaching after the base model already exists. Blackbaud says Lantern is post-trained on synthetic scenarios modeled on its Social Impact Signal Graph. Synthetic scenarios are examples built to stand in for real fundraising situations. Blackbaud calls the graph comprehensive, and says the scenarios are there so the model reflects real-world fundraising patterns and the complexity of that work. It says this setup lets Lantern change as the technology moves. Those lines are Blackbaud’s.
What Blackbaud says the model is for. It says Lantern understands how a fundraising relationship is built, looked after, and turned into a gift. It says the model reflects retention, upgrade paths, and lifetime relationships. Retention is whether a donor gives again. An upgrade path is a donor who gives more over time. A lifetime relationship is the whole stretch of that giving, one gift after another. Blackbaud says those ideas sit at the center of fundraising and are not built into a general-purpose model. Mike Gianoni, president, chief executive, and vice chairman of the board, said frontier models learn from the internet, while Lantern learns from philanthropy. He said more than $100 billion is raised, granted, or managed through Blackbaud’s platforms every year, and that the company knows why, when, and how people give. More than $100 billion is his figure for money already moving through those platforms. He said Lantern understands fundraising goals, donor dynamics, and nonprofit priorities in ways general-purpose AI does not, and that it puts that intelligence in a fundraiser’s hands. Those sentences are his, in the release.
Andy Kofoid, president of global field operations at Databricks, said philanthropy has an enormous opportunity to benefit from AI, and that the returns depend on the right data, the right context, and governance. Governance, here, means the rules for who can use the data and how. He said Blackbaud and Databricks together are giving nonprofits the data foundation they need to scale fundraising and increase real-world impact. That quotation is his, in the release. Blackbaud also says Lantern is developed under its Responsible AI principles, with privacy, transparency, and governance fitted to the sector. Those lines are Blackbaud’s. The release names the principles. It does not print their text.
Carrie Cobb, Blackbaud’s chief data and AI officer, said the future of AI is about who best reflects the world their customers work in. She said Lantern was purpose-built to reflect fundraising, using decades of sector expertise and the intelligence in the Social Impact Signal Graph. She said the result is more relevant, more transparent, and more actionable because it understands the relationships, behaviors, and opportunities that drive impact. She said the sector needs AI designed to support social impact. That quotation is hers, in the release. Blackbaud says it is building Lantern with input from customers including Boston University, Hesed House, and YMCA of the North, aimed at what frontline fundraisers want to get done. Tim Cerato, assistant vice president for constituent relationship management at Boston University, said every industry has its own language, expertise, decision-making, rules, and guardrails. He said one of the biggest unlocks for AI is understanding those nuances and amplifying the people doing the work, so a team can spend its energy on the outcomes that matter. That quotation is his. Hesed House and YMCA of the North are named as customers giving input. The release does not quote them.
Where a fundraiser would meet it. Blackbaud says Lantern will be embedded in its Platform for Good, so the fundraising intelligence sits inside the products organizations already use to reach supporters and raise money. It says Lantern will begin powering select Raiser’s Edge NXT and Agents for Good capabilities in 2027, with broader availability planned later across that same platform. Raiser’s Edge NXT is Blackbaud’s system for tracking donors and gifts. Agents for Good is the other product named on that line. Select means some of those capabilities. The release does not list which ones. 2027 is the year it names. The page does not name a month, and it does not say the capabilities are on today. It does not print a price. Blackbaud says statements about expected product benefits are forward-looking. Those lines are Blackbaud’s.
In plain terms, Blackbaud told its bbcon audience that Lantern is a language model for fundraising, built with Databricks and taught further on giving patterns, and that pieces of it are planned for Raiser’s Edge NXT and Agents for Good in 2027. The 85 percent and the about 10 percent are Blackbaud’s. The more than $100 billion is the chief executive’s figure for money already moving through the company’s platforms.
The picture is Blackbaud’s Introducing Lantern launch graphic from the newsroom announcement. White type reads “INTRODUCING” over the Lantern name, with “Blackbaud AI” underneath, on a dark purple field. It is the company’s card for this model.