|
Welcome back! Salesforce and Microsoft may be worried about threats posed by AI upstarts like Anthropic to their core enterprise apps, but they’re taking a conciliatory approach. Even though the two incumbents sell proprietary AI agents that can be used inside their Slack and Teams messaging apps, for instance, they’ve decided the apps ought to be a “platform” or app store for any AI agents their customers might want, including Anthropic’s Claude. A partnership Anthropic and Salesforce announced last week lets Slack customers summon Claude in their group chats to review their communications and handle workplace tasks, as long as they connect it to the other apps or databases their business uses. And Anthropic recently told Microsoft it is planning to develop such a feature for Teams, according to someone briefed on the discussions. The upcoming Claude agent in Teams would likely compete directly with Microsoft’s own Copilot AI tool, which is a key to Microsoft’s revenue growth strategy. But Teams, like Slack, already allows customers to use external AI agents from coding app Cursor, product development app Linear and search app Perplexity, whose capabilities overlap with that of Copilot. (Teams has several times more users than Slack.) Similarly, the new Claude agent in Slack, Claude Tag, would seem to cut against Slack’s effort to make its competing AI agent, Slackbot, a big business in its own right. Yet Salesforce promoted Claude Tag heavily in social media. That prompted plenty of Salesforce employees to worry about whether Anthropic was akin to a fox entering their hen house, we reported over the weekend. Their concerns were understandable: As some of you may recall, an Anthropic official even said publicly last fall that the AI firm used its Claude to create a workplace chat app similar to Slack! It doesn’t help that Andrej Karpathy, an Anthropic staffer and AI celebrity, last week appeared to dismiss “Slack bot” in comparison to the Claude agent for Slack, which he referred to as a “new paradigm” in AI-human relations. Needless to say, other software executives had strong reactions to these developments. Salesforce opening up Slack to Anthropic seemed “crazy...like a real misstep that salesforce would allow this,” Parker Conrad, CEO of HR app Rippling, said in a post on X in response to our story over the weekend. A New ‘Like’ Button? Another major software CEO told us privately that the dynamic reminded them of when website publishers allowed Facebook to place a “Like” button on their sites. While the publishers did so to be more relevant to their visitors, the move gave Facebook tremendous intelligence about the web and its content as it pursued ways to bring such content into the “walled gardens” of its social media apps. Other onlookers viewed things differently. They said the Salesforce move seemed in line with the plan many enterprise apps have to allow their customers to use AI agents made by other companies—for a fee. But the external AI agent providers don’t need to pay Microsoft for the privilege of operating the agents in their messaging apps, according to a company spokesperson. The same is likely true of Salesforce, as well: sources familiar with two AI app developers, Viktor and Perplexity, told us they pay the software giant nothing to host their agents on Slack. And Microsoft’s public app store policies say the company only collects fees when customers spend money through third party apps, which isn’t the case for such Teams plugins. Salesforce declined to comment on the financial terms of its deals with other agent providers. We don’t know the details of Anthropic’s financial arrangement with Salesforce, which is probably fairly complex, given that Salesforce actually uses Anthropic’s models to power Slackbot and its other key AI agent tool, Agentforce. For all we know, Anthropic could be giving Salesforce a giant discount on that model access in exchange for promoting the Slack integration. Driving Bot Awareness The attention Slack got for the Claude Tag integration also appears to have alerted more Slack customers to the existence of AI agents in the messaging app, and that could be great news for Salesforce’s strategy. For instance, Poland-based AI startup Viktor told us it saw its “best week ever” following the Claude Tag launch, with more than 400 new customers signing up to use its generalized AI agent. The agent can work across Slack channels to help with tasks such as performing a cost analysis of a potential supplier by surfing the web for alternatives, for example. AI search firm Perplexity, which launched its Computer AI agent as a Slack app in April, said the number of enterprise users of its agent on Slack has risen 17% per week on average since then. Perplexity also offers Computer through Microsoft Teams but has existed longer in Slack, where customers can use it to handle a variety of administrative and coding tasks, such as creating a coding bug ticket in GitHub or tapping Snowflake databases to analyze a company’s performance and to produce slideshows or spreadsheets. Microsoft’s approach is even more strategic. It says it lets virtually any software provider build features for customers that connect to Microsoft applications. For example, Microsoft has long allowed Anthropic customers to use Claude to take over work they’re doing in Microsoft apps such as Excel, Powerpoint, and Word, despite selling Microsoft own competing agent, 365 Copilot. Microsoft executives have framed this approach as necessary to establish Microsoft as a “platform company” that promotes AI models and agents from a range of providers, giving customers more reasons to keep buying its apps. Without explicitly naming Anthropic, CEO Satya Nadella told shareholders in April that plugins like the Claude Office features would help entrench Microsoft among customers. “It’s fascinating that here we are in 2026, and the most exciting things [in AI] are plugins in Word or Excel,” Nadella said. “When you see that, that means we have a structural position in knowledge work.” Still, Microsoft doesn’t always approve such plugins. It blocked software built by competitor DataBricks that would have made it easier for customers to connect DataBricks AI agents to its business intelligence app PowerBI, for instance. Microsoft said at the time that the move was due to concerns over the accuracy and reliability of DataBricks’ tool, not competition. For its part, Anthropic has said it isn’t trying to hurt enterprise apps and it has partnered with plenty of them, not just Slack. Some software app makers are nevertheless scared. Design app Figma, for instance, dropped out of talks for a recent potential collaboration with Anthropic and fears the AI firm is trying to eat its lunch. Anthropic has been developing products and features that appeal to specific industries and workers, including designers. AWS Puts $1 Billion Toward FDE Efforts We told you last week that Amazon Web Services sales leaders recently began training some of the cloud firm’s “solution architects” to be forward deployed engineers that work on-site at customers’ offices to help them develop AI applications. Well, today AWS said it’s putting $1 billion into a new FDE organization for customers. You heard it here first!
| 
