When Salesforce introduced Headless 360 at its TDX developer convention final month, the response from some quarters was underwhelmed. Pricing complaints. Questions on roadmap readability. A number of pointed options that Salesforce was retreating from the CRM market.
Simon Harrison watched all of this with barely hid frustration.
“You guys, are you kidding me?” says Harrison, founding father of Actionary, a know-how advisory agency. “What are you speaking about, the pricing wants enhancing? Do you realise what’s simply landed in your laps?”
Harrison’s argument will not be that Salesforce has finished one thing intelligent. It’s that Salesforce has finished one thing that can set the bar for each enterprise utility from this level ahead, and that the UC and collaboration business specifically wants to take a seat up and concentrate.
What “Headless” Truly Means
The idea is simple sufficient: strip away the consumer interface as the first manner of interacting with software program, and expose the underlying knowledge and performance via an API floor that brokers, command line instruments, or any suitable shopper can name instantly. Salesforce’s personal framing is blunt concerning the path of journey: “Why must you ever log into Salesforce once more?” requested Co-Founder Parker Harris at TDX.
What makes Headless 360 vital, Harrison argues, will not be the idea however the engineering behind it. Salesforce has had APIs since 2006. Redesigning these APIs particularly for non-human callers is a distinct endeavor completely.
“Designing these APIs as a floor for non-human callers was an enormous endeavor,” he says. “I’d not be stunned if this has been occurring for two-plus years.”
The results of that work is a platform with greater than 60 new MCP instruments and 30-plus preconfigured coding expertise, enabling an agent to work together with Salesforce knowledge exactly and repeatedly, with out the anomaly that makes most integrations unreliable in apply.
The Downside Most Distributors Haven’t Solved
Harrison constructed his personal CRM utility in Python to grasp these dynamics first-hand, and the expertise taught him one thing he says most distributors are nonetheless getting fallacious.
Giant language fashions exist in a state of fixed pressure between two modes: open-ended reasoning, the place they’re highly effective and versatile, and exact motion, the place they have to be deterministic and constant. Getting each from the identical mannequin, with out sacrificing one for the opposite, is the engineering problem that the majority platforms haven’t cracked.
“You go from a mannequin the place the corpus of information and the flexibility to choose one instrument, net search, allows it to reply questions on all the pieces it’s been skilled on,” he says. “To: overlook all that. Simply perceive intent, and use these instruments.”
The mechanism that resolves this pressure is Mannequin Context Protocol (MCP), an open commonplace initially developed by Anthropic. It acts as a structured language between the mannequin and the applying, permitting the mannequin to motive freely however act solely via well-defined, scoped instruments with clear enter and output schemas.
Distributors that lack the API depth to help this are, in Harrison’s view, in deep trouble.
“There are monoliths on the market that also don’t have a good API floor. They’re going to battle.”
What This Means for UC Platforms
For UC patrons, probably the most urgent query is what this shift means for the collaboration and communications instruments they already depend on. The platforms sitting on the richest concentrations of office knowledge, calendars, emails, conversations, and assembly recordings, will not be but doing sufficient with it.
Harrison’s reply is anchored in a easy every day behavior he has constructed along with his personal agent-connected CRM. “Inform me what I ought to fear about in the present day,” he says, describing what he asks every morning. “And my checklist is prioritised to most impact. I don’t should traverse these numerous screens to determine what I’m doing, why, and the way.”
That form of interplay ought to be doable inside a UC platform. But it surely requires the platform to show its knowledge via an agent-ready API floor.
Microsoft Copilot in Groups is the obvious instance. MCP is now usually accessible in Copilot Studio, and brokers in Groups channels can connect with third-party instruments by way of MCP servers. Harrison sees the potential clearly, but in addition the hole between the place Copilot is and the place it must go.
“What I want from Copilot is an MCP shopper functionality. If I may connect with my MCP server inside Copilot, I’m now triaging my calendar, my emails, my chats, with the information in my CRM utility, with out having to take a look at both. And that’s enormous.”
Of the main UC gamers, Harrison reserves his strongest present reward for Zoom. At Zoomtopia 2025, the corporate introduced MCP help inside AI Companion 3.0, permitting admins to configure instruments for customized and digital brokers throughout the platform. Extra lately, Zoom launched an MCP connector for Claude, pulling assembly summaries, transcripts, and scheduling knowledge into exterior agent workflows. “Zoom is probably doing the most effective job when it comes to the office agent strategy,” says Harrison.
The Bar Has Been Set
Harrison will not be a Salesforce accomplice or paid analyst. His conviction comes from a yr of constructing agent-connected purposes and observing what occurs when the structure is completed correctly.
His dwell demonstration throughout our dialog made the purpose cleanly. Inside minutes, utilizing solely pure language directions to an agent linked by way of MCP to his CRM, he created a totally enriched firm account, pulled in a management workforce from a public net web page, and obtained a prioritised breakdown of his most urgent duties. No display screen navigation. No handbook knowledge entry.
“I obtain issues that may take days, weeks. It takes me minutes. It’s extraordinary.”
Some analysts have urged warning, noting that Salesforce has but to reveal pricing for the headless capabilities, and that questions stay round SLAs for MCP instrument calls in manufacturing environments. These are honest factors for patrons to press on. However Harrison’s argument is that the path of journey is what issues right here, not the launch-day pricing sheet.
Each utility vendor, UC platform included, will finally should reply the identical query Salesforce has spent two and a half years quietly resolving: is your API floor adequate for an agent to work with?
“The headline about Salesforce,” Harrison says, “is that that is the way forward for purposes. It’s, fingers down, the one manner we are going to determine to make use of purposes sooner or later.”
Whether or not the UC business strikes quick sufficient to satisfy that bar is, for now, an open query.
