|
Welcome back! At a Salesforce conference for its customers on Wednesday, Chief Marketing Officer Ariel Kelman showed a slide during a keynote speech about the cloud software giant’s use of artificial intelligence to supercharge its marketing automation products. “Your data is not our product,” it said, under the header “Our Values.” That may be the case. But Salesforce sure seems worried about customers using the critical corporate data they store in Salesforce’s apps to develop their own AI. In late May, Salesforce changed its data usage policies to prohibit customers from training AI models on the data they create in its messaging app, Slack, while also introducing other restrictions on how third-party software providers can use that data. In a blog post, Salesforce said the changes would “better protect customer data, uphold the integrity of Slack’s platform, and ensure its ecosystem continues to grow in a way that’s open and responsible.” Ironically, a little over a year ago, online sleuths on Hacker News pointed out that Slack’s own privacy policies permitted Slack to use its customers’ data to train certain AI models. It was the customers’ responsibility to contact Slack to opt out of that policy, which very few of them seemed to know about in the first place! For Salesforce’s Slack app, the data policy changes go even further. Customers typically want to let third-party applications access their Slack data so they can more easily search for messages and data, for example. But the Slack policy changes mean third-party providers can only access customers’ data temporarily, which limits the usefulness of these tools. This change got little attention until I reported on Tuesday how it was hobbling Salesforce AI rivals such as Glean, the fast-rising enterprise search firm that has been turning heads around the industry. Glean told its customers that Slack’s move to wall off data from Glean’s search product amounted to the company “locking your data within Slack.” The moves are throwing jet fuel on a long-simmering discussion about whether customers own the data they create when using enterprise applications or whether those rights belong to the providers of those applications, including Microsoft, Adobe, Oracle and SAP, to name a few. (These firms don’t seem to have the same restrictions as Salesforce when it comes to customers training AI on their own data.) Bill Gurley, general partner at Benchmark, is squarely in the former camp. “Enterprise customers need to declare if they support ‘open data’ or ‘closed data.’ And customers should move away from ‘closed data’ vendors as fast as possible,” Gurley said on X in response to our story about Salesforce blocking AI rivals from accessing Slack data. Gurley isn’t alone. Software executives from startups and established firms have publicly or privately (to me) criticized Salesforce over the moves. Some characterize them as anticompetitive and contrary to customers’ best interests. Several executives from smaller software providers who build products that use data from Salesforce CRM said they’re concerned about the moves. “I’m disappointed that Salesforce chose to play defense rather than offense,” said the CEO of one of these software providers. “It’s really a bad look if they start preventing interoperability and integrated systems.” Phil Fernandez, former chair and CEO of Marketo, a marketing automation company that is now part of Adobe Systems, shared an anecdote on LinkedIn about Salesforce asking his firm to pay “a huge amount” of money to let a Salesforce customer connect their CRM data to Marketo’s application several years ago. After Fernandez apparently rallied his company’s customers (who were also Salesforce customers) to oppose the change, Salesforce backed down, according to Fernandez. Salesforce spokespeople didn’t respond to a request for comment on the reaction to the Slack policy changes. AWS Scrambles to Catch Up in AI Agents Developing AI apps isn’t a walk in the park. That’s especially true for companies that run their businesses or apps on one cloud, such as Amazon Web Services, but want to use AI models hosted on competing clouds, such as OpenAI’s GPT-4o, to power some of those apps. Some cloud customers have told us AWS in particular has been a bit restrictive when customers want to develop sophisticated AI apps, such as agents that automatically patch websites after outages, on its main AI cloud app service, Bedrock. Customers have only been allowed to use certain models and agent-development software. That could soon change, as I reported this morning. AWS is planning to revamp Bedrock to give customers more flexibility on the tools and models they use to develop AI apps. AWS is far and away the biggest cloud provider, but in enabling AI app development, it seems to be chasing after Google and Microsoft.
| 
