Revenue platforms spent the last two years building AI copilots inside their own dashboards, each one walled off from the others. Clari and Salesloft are now doing the opposite: pushing their data out to every major AI assistant a sales team already uses, from Claude to ChatGPT to Microsoft Copilot to Google Gemini. The shift signals that revenue intelligence is moving from a locked-in feature that lives inside one vendor’s screen to shared infrastructure that any AI agent can query, a change that matters more than any single feature announcement.
From dashboard to data layer
Clari and Salesloft, merged into a single revenue orchestration platform, built a Model Context Protocol (MCP) Server that lets outside AI tools read live pipeline, deal, and conversation data without a custom integration. MCP is the open connector standard, championed by Anthropic, that gives AI assistants a governed way to pull from and act on business systems instead of relying on stale exports or one-off API work.
The company’s own newsroom describes the goal as closing “the gap between insight and action,” in the words of Clari + Salesloft CEO Steve Cox, so that a forecast risk flagged in a deal record can trigger an action in whatever AI tool a rep or manager already has open, rather than sitting in a dashboard nobody checks until the pipeline review.
What is rolling out
The latest phase puts Salesloft natively inside Claude’s connector directory, meaning reps can pull cadence and activity data into Claude without configuring a custom connector first. That is the first release in a run that continues through late summer: Clari Copilot call intelligence, forecasting and deal-inspection data, write-back capabilities, and dedicated connectors for ChatGPT and Microsoft Copilot are all queued to follow, with Google Gemini support rounding out coverage of the major assistants sales teams are already standardizing on.
Clari + Salesloft chief product officer Kylie Fuentes framed the goal as removing a bottleneck that has nothing to do with the underlying AI models: “The limiting factor in revenue AI is no longer vendor technology. It’s your team’s imagination.” That framing, extending revenue context into whatever agentic workflow a team builds rather than shipping one more built-in assistant, is the real story: the vendor is betting that openness, not another proprietary copilot, is what wins the next round of the AI sales tools race.
Why open up the data at all
The company points to a gap it says its own Clari Labs research uncovered: 87% of enterprises missed revenue targets despite record AI investment. That statistic is doing a lot of work in the announcement, and it should be read as the vendor’s justification for the strategy rather than an independent audit. But it lines up with a pattern showing up across the sales tech landscape this year: buying more AI tools has not, on its own, closed more deals. Several vendors are now betting that the fix is connective tissue between systems rather than another isolated feature.
ZoomInfo made a similar bet last week with the launch of its GTM.AI command-line interface, exposing verified go-to-market data directly to AI agents and developer tooling rather than only through a web dashboard. Read together, the two moves suggest sales intelligence and revenue orchestration vendors are converging on the same conclusion: the customer is no longer only the human rep staring at a screen. It is also the AI agent acting on that rep’s behalf, and it needs a data interface built for machines, not just for people.
Not every vendor is choosing the same route to reach AI agents. Some are distributing through a single hyperscaler’s marketplace rather than opening a protocol to every assistant at once, a narrower bet that trades broad reach for a deeper relationship with one platform. Clari + Salesloft’s approach is the opposite wager: cover every major assistant a buyer might already be using, on the theory that revenue teams will not standardize on one AI ecosystem any time soon, and the vendor that shows up everywhere wins the workflow regardless of which assistant the rep picked.
The skeptic’s view
Opening governed data access to third-party AI assistants raises the obvious question of control. Every additional AI ecosystem with read (and eventually write) access to pipeline and call data is another surface for data governance, permissioning, and compliance teams to police. Vendors that move fast on MCP exposure will need equally fast investment in audit logging and access controls, or the openness that is meant to be a selling point becomes a security review that stalls procurement instead.
What it means for the sales leader
For a revenue leader evaluating this shift, the question is not whether your reps will eventually touch an AI assistant during their workflow; most already do. The question is whether your revenue platform’s data can follow them there, or whether it stays trapped behind a login screen only your CRM vendor controls. A platform that exposes governed, live data to the AI tools your team has already adopted removes a rebuilding step that costly custom integrations used to require.
Before adopting any MCP-based connector, revenue operations teams should confirm three things: what data scope the connector actually exposes (read-only versus write-back), what audit trail exists for AI-initiated actions on deal records, and whether the licensing tier required (Salesloft’s rollout is tied to its Agentic add-on) matches what the team already pays for. Vendors are moving quickly to claim “open” as a selling point; the operational discipline to make that openness safe is the part still being built in real time.
The practical test will come at renewal time, when revenue operations teams compare what each vendor’s MCP exposure actually unlocked against what it cost in governance overhead. A connector that saves reps a login is a convenience; one that lets an AI agent quietly rewrite a forecast without a clear audit trail is a liability wearing a feature’s name. Leaders who get ahead of that distinction now, rather than after an AI agent touches a deal record it should not have, will be the ones setting policy instead of reacting to an incident.
Source: Salesloft
