Salesforce buys the meter, Anthropic goes dark

Salesforce opened the week by agreeing to acquire m3ter, a metering and rating engine built for consumption-based billing. After a year of selling consumption-priced agents, it bought the company that counts the consumption. Full treatment below.

The rest of the week ran in two other directions. Salesloft folded Clari Copilot into its platform and added a Claude connector, more evidence of the engagement layer turning into an agentic hub. And in the most consequential model-layer event of the year, Anthropic was ordered to pull two of its frontier models offline for every customer, a reminder that the AI underneath your stack now answers to more than its vendor.

Here’s what moved.

Salesforce — Signs Definitive Agreement to Acquire m3ter

Salesforce announced on June 8th that it has signed a definitive agreement to acquire m3ter, described in the announcement as “a leading metering and rating platform purpose-built for consumption-based monetization” which brings metering and rating capabilities natively to Agentforce Revenue Management. The deal is expected to close in the second quarter of Salesforce’s fiscal year 2027. Terms were not disclosed.

Salesforce’s last earnings call showed its consumption pricing can work at scale. The underlying architecture is the hidden key that makes the motion successful. m3ter is the infrastructure that measures and rates usage at high volume, which is the unglamorous prerequisite for billing anyone on what they actually consumed. Buying it means consumption billing moves from a model that Salesforce talks about to a capability built into the platform, and it does so natively rather than through a bolted-on third party.

This acquisition is a win for Salesforce on two fronts. The first is you: a CRM that can meter usage natively can bill you on consumption with far more granularity than a seat count. This bolsters Salesforce’s growing consumption motion. The second is less obvious but matters more for revenue teams at companies that sell usage-priced products themselves. Agentforce Revenue Management allows companies to run their own quote-to-cash. Salesforce will now fold m3ter’s capabilities directly into that product. So, if your company sells anything on a usage basis (an API, a platform charged per transaction, a product with overage fees), you can use m3ter’s capabilities yourself, inside Salesforce, to bill your own customers. If a transition to consumption billing is anywhere on your roadmap, this belongs on the watch list.

Source: Salesforce Newsroom — m3ter acquisition, June 8, 2026

Salesloft — Clari Copilot Replaces Native Conversations; a Claude Connector Lands

Salesloft’s June 9th release is the first real product evidence of the Clari and Salesloft merger reaching the stack. Clari Copilot is now integrated directly into Salesloft without requiring swivel-chairing and brings Clari’s conversational intelligence into deal summary agents, research agents, activity feed, and meeting prep. Access requires an Advanced or higher license tier. The release also adds a connector to link Salesloft to Claude through the MCP Connectors Directory, available with the Salesloft Agentic add-on, plus Cadence Collections and three new AI adoption metrics.

If you run both tools, Clari Copilot is taking over from native Salesloft Conversations. That migration happens in phases, so it won’t hit you all at once, but current Conversations customers will get moved eventually. Find out which phase you’re in and what actually changes in the handoff, and budget for the fact that Copilot now requires an Advanced license or higher. The Claude connector in the same release is the smaller story, but it points the same way the rest of the engagement layer is headed: the assistant is becoming the front door to your other tools, not a walled garden.

Source: Salesloft Champions — June 2026 release notes

Anthropic — Government Directive Pulls Fable 5 and Mythos 5 Offline for Everyone

On June 12th, Anthropic abruptly disabled its Fable 5 and Mythos 5 models for all customers after the US government issued an export-control directive suspending access for any foreign national, whether inside or outside the United States, including Anthropic’s own foreign-national employees. A nationality-based restriction at that scope left no partial option, so both models went dark for everyone, though Anthropic said access to its other models was unaffected. The government cited national security and, per Anthropic’s account, a method of jailbreaking Fable 5. Anthropic disputes that a narrow jailbreak justifies recalling a model deployed to hundreds of millions of people and says it is working to restore access.

Set aside the question of whether the government’s decision was justified or not, because the impact to your business is simple: a frontier AI lab that GTM tools depend on switched off two of its models in a matter of hours, for reasons that had nothing to do with the product.

Still, the story is not about the model, it is an InfoSec and supply-chain story. The AI features your team builds, or your vendors sell, run on a short list of frontier models, and this week one of them disappeared for every customer with effectively no notice. Most were not yet running on the newly released Fable 5 model, but the point remains. Whatever uptime you assume for an AI feature in your stack, the real dependency sits one layer below your vendor, at a model that can now be pulled by a jailbreak disclosure, a security incident, or a government order. If a tool you run quietly routes to a specific model, you inherit that continuity risk and the vendor’s status page will not warn you before it bites.

The governance work this creates is concrete, and it belongs in a vendor review next to SOC 2 and data residency. Capability resiliency should be a key element in your stack architecture, with even more rigor for any workflow embedded with a specific frontier model that touches customers and has an SLA. Begin outlining which model backs each feature and whether there is an automatic fallback if that model goes offline. None of this was on most RevOps risk registers a week ago. The lesson is not to back away from AI features, it is that model-layer dependency has become an InfoSec and business-continuity concern, and it now needs an owner.

Sources: Anthropic statement, June 12, 2026 · Bloomberg, June 13, 2026

A closer look at one smaller stack vendor each issue.

Unify — Opening the Data Layer on Signal-Based Outbound

Unify is an agentic, signal-based outbound platform: it combines intent signals, B2B contact data, AI research agents, and automated multi-step “Plays” into one warm-outbound workflow, with customers listed on its site including Perplexity, Together AI, Cursor, and Flock Safety. It sits in the crowded category of tools trying to turn buying signals into sequenced outreach without a human stitching the steps together. What makes it worth a closer look in this issue is not the signals, which everyone now claims, but what it shipped over the last few weeks: a deliberate opening of its own data layer.

Webhooks in Plays, on May 26th, lets Unify “push prospect and company data to any tool in your stack with a Webhook Action in Plays that fire the moment a signal triggers,” naming destinations like n8n, Zapier, dialers, LinkedIn, and marketing automation tools, “without a single CSV export or CRM workaround.” The Unify API, on May 21st, runs the other direction, letting teams “connect your data warehouse, third-party tools, or internal systems directly to Unify” and use that data “to power targeting, plays, exclusions, and more.”

The design plays right into Enriched’s framework for the future of GTM. Most signal tools assume they are the destination: signals come in, outreach goes out, and the data lives inside the tool. Unify is betting its users would rather treat it as a connectable signal engine, with the warehouse on one side and the rest of the stack on the other. For a RevOps team, that reframes the buying question. The question isn’t whether Unify’s signals are better than the next vendor’s, it is whether you want signal detection to be a service that routes into your existing stack or another walled tool you check separately. Outbound tools and engagement-layer incumbents are all betting that the GTM ecosystem is just that, an ecosystem. Interoperability has become table stakes rather than a differentiator.

Source: Unify changelog, May 2026

The rest of the week, in brief: