SAP's data double-down; Glean stretches out

Welcome to Runtime! Today on Product Saturday: SAP strikes a big partnership with Databricks, Glean introduces its agent builder, and the quote of the week.

(Was this email forwarded to you? Sign up here to get Runtime each week.)

Ship it

Lake effect: SAP customers who are looking for something a little more modern in a data service got their wish this week with the launch of SAP Business Data Cloud, which is essentially a sweeping partnership with Databricks. Customers will now be able to mix their SAP data with third-party data stored in a data lake using Databricks' tools.

"With SAP Business Data Cloud, we are offering data products that bring together financial data like general ledger or cash flow data, with external data such as inflation rates, allowing finance to run simulations, for example, of the impact of inflation on the company's financial performance," SAP CEO Christian Klein said in a press conference. Constellation Research analyst Doug Henschen noted that the partnership does not extend to Databricks' Lakeflow and Lakehouse Federation products, which gives Databricks an opportunity to bring on a bunch of new SAP customers.

Reason season: Over the last several months, the launch of so-called "reasoning" generative AI models designed for deeper analytical tasks has changed the way the market evaluates top-tier models. IBM will be a little late to the starting line, introducing a "preview release" of reasoning capabilities this week that will be included in the upcoming launch of Granite 3.2.

"One of the most exciting new frontiers in generative AI is improving model performance by using more inference-time compute so that the model can reason about a task before answering," IBM said in a blog post. It compared its "preview release" quite favorably to the competition in a bunch of benchmarks, but also made it available under the same Apache 2.0 license that governs the Granite family of models.

Distance vision: Glean has raised over $600 million in recent years to build out its enterprise search service, and this week it entered its agentic period. Glean Agents is a new "horizontal agent environment" designed to help companies build AI agents that can pull data from a wide variety of internal and external sources.

"With Glean’s natural language agent builder, employees can build agents by describing the agent’s desired output in simple, natural language," the company said in a press release. Those agents will be able to browse corporate data stored in Salesforce, Atlassian, and Databricks as well as real-time data from the internet, Glean said.

Snow day: Snowflake also jumped into the AI agent fray this week, because at this point can you really call yourself an enterprise software company if you haven't announced an agentic AI strategy? Cortex Agents, available as a preview, combines its Cortex Analyst and Cortex Search features with Anthropic's Claude large-language model.

"Snowflake customers now have a unified platform for processing and retrieval of both structured and unstructured data with high accuracy out-of-the-box," the company said in a blog post. As covered in Thursday's edition of Runtime, concerns about accuracy are keeping a lot of businesses on the sidelines when it comes to putting AI agents into production.

Record store: Meanwhile, Workday launched a new service that will allow companies to "manage their entire fleet of AI agents" — that's right, all of them — through its platform. The Workday Agent System of Record was designed to help companies introduce new AI agents while keeping tabs on how much they cost to run, the company said in a press release.

There are lots of companies vying for that central role in our glorious agentic AI future, but Workday's plan is to convince customers that AI agents are actually part of the workforce you're already managing with Workday, as opposed to, you know, software. Runtime has to agree with Phil Wainewright at Digimonica on this one: "Yes, AI agents need careful management and governance. But that's a job for IT, not HR."

Stat of the week

Cybersecurity professionals have to always have one eye on the ever-expanding number of new threats to their organizations, but sometimes the biggest threats are lying in plain sight. More than 62% of security incidents can be traced back to issues that companies already knew they had to deal with, according to a new report from Zest Security.

Quote of the week

"We want people to be intrinsically motivated. We don't think, for this particular technology change, that extrinsic motivation works." — Canva CTO Brendan Humphreys, explaining how the company got its developers to experiment with generative AI tools by giving them access to lots of options.

The Runtime roundup

The Trump administration asked TSMC if it would be interested in taking over Intel's chip-making factories, according to Bloomberg, and the New York Times said Intel is potentially on board.

It was another banner week for Elon Musk's Nerd Reich; 404 Media reported that DOGE left its signature website unsecured to anyone who wanted to push their own updates live, and according to HuffPost they also posted classified information on that site.

Thanks for reading — see you Tuesday!