MCP & Agent Interfaces28 Sep 2026
Shopify gives Meta's Muse agent full checkout access via Shop Pay
Meta's September 24 recap of Connect 2026 confirmed that Muse, its personal AI shopping agent, now has access to Shopify's full product catalog and can check out directly using Shop Pay or PayPal — letting the agent complete a purchase itself rather than just surface product recommendations, with Shopify and Meta executives describing the rollout as agentic checkout across all Shopify stores.
Agent commerce just moved from pilot to default distribution on a major platform — GTM and commerce leaders should treat "is our storefront agent-checkout ready" as a live operational question, not a 2027 roadmap item.
Meta newsroom →
MCP & Agent Interfaces28 Sep 2026
Microsoft's Work IQ lets Copilot and agents write directly into Dynamics 365 and Power Platform
Microsoft's September 25 announcement of Work IQ describes an intelligence layer that grounds Copilot and agents in Dynamics 365 and Power Platform data and adds governed write actions — reassigning ownership, escalating service issues, updating contract records — with public preview opening September 30 across Dynamics 365 Sales, Service and Power Apps, billed on a usage-based model kept separate from standard Copilot licensing.
Splitting a usage-based meter for the agent layer off from seat-based Copilot licensing is itself a pricing signal — the governance and grounding layer is becoming its own billable product, not a bundled feature.
Microsoft Dynamics 365 blog →
MCP & Agent Interfaces21 Sep 2026
Salesforce launches AIforce, letting external AI agents update records and trigger workflows directly
Salesforce introduced AIforce, an interface layer that exposes its data, workflows, permissions and governance to external AI agents and platforms — including Claude, which launches with 37 prebuilt sales skills — so agents can reason across Salesforce data, update records and trigger workflows outside the traditional UI, with actions still routed back through existing permissions.
This is Salesforce betting the next interface is agent-native rather than screen-native — worth watching whether your own CRM data becomes something agents act on directly, not just something reps look at.
Salesforce newsroom →
MCP & Agent Interfaces21 Sep 2026
GitLab 19.4 expands its MCP server with real write actions across code, issues and vulnerabilities
GitLab's 19.4 release adds new MCP server tools that let connected agents open and merge merge requests, run or cancel pipelines, commit and fork repositories, and create or update issues, epics and vulnerabilities — not just read them — alongside a new credit-cap page for setting default and per-user AI usage limits.
Pairing real write access with per-user spend caps in the same release is the template other vendors will need: agents that can actually merge code or close a ticket also need a governor on what they're allowed to spend doing it.
GitLab →
MCP & Agent Interfaces10 Sep 2026
OpenAI ships a managed Agents API that runs and orchestrates agents in production
OpenAI moved its Agents API into public beta: a hosted service that lets agents run code, manage files, coordinate parallel subagents and keep long running sessions alive across extended workflows, with support for MCP, custom functions and built in tools, and no fee beyond standard token and tool usage.
For GTM teams building agent-driven workflows, hosted orchestration removes months of infrastructure work — the competitive question shifts from "can we run an agent reliably" to "whose agent understands our business best."
OpenAI →
MCP & Agent Interfaces10 Sep 2026
Salesforce's new AI Control Plane lets agents reserve inventory and trigger fulfillment directly
Salesforce introduced the Trusted Enterprise AI Harness, a six-part architecture anchored by an AI Control Plane that registers agents and governs real write actions — in Salesforce's own example, reserving inventory, updating an order and triggering fulfillment — across Salesforce, third-party and customer systems. Pricing is still to be announced ahead of a planned early FY28 rollout.
This is a governance layer for agent write access, not another chat interface: worth tracking as the template other CRM and ERP vendors are likely to copy for policing what agents are actually allowed to do.
Salesforce newsroom →
Earnings & AI Signals10 Sep 2026
Adobe: AI-first ARR grew more than 150% year over year as it leans into agentic experiences
On its Q3 FY2026 earnings release, Adobe reported AI-first annualized recurring revenue up more than 150% year over year and crossed one billion monthly active users across its creativity and productivity products, with incoming CEO Anil Chakravarthy framing the ambition as making Adobe "the leader in agentic software for creativity, productivity and customer experience."
A 150%+ AI-ARR growth rate from a company Adobe's size is a concrete data point for anyone building the board-deck case that agentic pricing lines can scale past pilot revenue.
Adobe Q3 FY2026 earnings release →
Voices13 Sep 2026
Harvey builds its GTM org around ~180 embedded ex-lawyers, not typical CS hires
Jason Lemkin details how legal AI company Harvey staffs pre-sales, post-sales and custom-build roles with roughly 180 former practicing attorneys rather than conventional implementation staff, paying them $220K–$320K OTE on a 75/25 split plus equity — a variable-comp structure usually reserved for quota-carrying sales roles — funded off its $11B raise.
When the product needs deep domain trust to get adopted, "customer success" starts to look like a second sales motion with its own comp plan: worth checking whether your own post-sales AI roles are priced like overhead or like revenue drivers.
Jason Lemkin, SaaStr →
Voices11 Sep 2026
PricingSaaS: vendors need a four-layer pricing stack for agentic AI, not one model
Surveying recent packaging moves from OpenAI, Notion, Canva, Clay, Replit, Suno and Runway, PricingSaaS argues that because agentic tasks now span costs from fractions of a cent to thousands of dollars a day, a single AI price point breaks down; they propose bundling near-free routine work, metering mid-cost tasks to the cheapest capable model, gating frontier-model access behind permissions and spend caps, and selling the most capable agents as flat-fee "digital labor."
A usable mental model for Layer 3: if your pricing page still has one AI add-on SKU, it's probably conflating four cost tiers that should be priced and gated separately.
PricingSaaS →
Voices09 Sep 2026
ElevenLabs still pays humans commission when its AI agents close the deal
Jason Lemkin's interview with ElevenLabs VP of Revenue Carles Reina describes a GTM org where AI SDRs, AEs and CSMs work deep into the pipeline, yet human reps keep a 20x-of-base quota (e.g. $100K base = $2M ARR target) and earn commission on any deal an agent originates or expands — with zero commission on POCs, so payouts track only recurring revenue.
The comp lesson isn't "replace the rep," it's "keep the rep economically rooting for the agent" — a structure other orgs deploying AI SDRs will need to copy or reinvent.
Jason Lemkin, SaaStr →
MCP & Agent Interfaces07 Sep 2026
Genesys adds an AI Control Plane and Orchestrator that route work between agents and humans
Genesys announced new agentic orchestration tools for Genesys Cloud, including an Orchestrator that assigns and moves work between AI agents, human staff and systems, and an AI Control Plane for centralized policy and observability over autonomous actions.
A dedicated governance layer for agent actions, not just a chat interface, is the packaging pattern to watch: expect other CX and GTM platforms to ship their own version of this control plane.
Genesys newsroom →
Pricing & Packaging07 Sep 2026
Atlassian adds a per resolution meter for autonomous AI agent outcomes
Atlassian is expanding usage based pricing across Rovo credits, automation steps and a new "AI agent resolutions" meter that bills per outcome when an agent autonomously closes a request without human escalation, effective December 3.
This is one of the cleanest outcome based pricing lines from a major platform vendor to date: a concrete reference point if you're building or negotiating agent driven service pricing yourself.
Atlassian blog →
Earnings & AI Signals07 Sep 2026
Asana: AI products now drive 25% of net new ARR, up from 17% last quarter
On its Q2 FY2027 earnings call, Asana said AI Studio and AI Teammates drove about 25% of net new ARR, up from 17% the prior quarter, and confirmed its new "Agentic Work Management" product launches self serve in mid September.
A jump like that in a single quarter is a real, board level signal that agentic features are already moving expansion revenue: a useful benchmark to hold your own AI attach numbers against.
Asana Q2 FY2027 prepared remarks →
MCP & Agent Interfaces21 Aug 2026
HubSpot ships a first party MCP server with scope gated CRM read/write access
A remote MCP server, separate from the in product Breeze copilot, with admin authorised scopes covering CRM objects and engagements.
This is the shape most categories will eventually converge on: a permissioned agent interface that sits apart from the chat copilot. If your CRM vendor doesn't have an answer to this yet, that's worth asking about directly.
HubSpot developer documentation → · full audit: Signal #001 →
MCP & Agent Interfaces21 Aug 2026
Metronome publishes @metronome/mcp with write access to contracts and usage ingestion
Bearer token authenticated, bundled in Metronome's official Node SDK, exposing writes including contract creation, not just usage queries.
Billing is usually the last system anyone lets an agent touch. A metering vendor shipping write scopes here, not just read, is a signal the rest of the category will have to match it.
Metronome documentation → · full audit: Signal #001 →
Pricing & Packaging21 Aug 2026
PROS divests its B2B CPQ and pricing line to Conga, repositions around airline retailing
PROS is no longer the B2B pricing science vendor it was categorised as even a year ago: that product line now sits with Conga.
A reminder to re check vendor category maps on a cycle, not just build them once. "Best in class B2B pricing software" lists compiled a year ago may be citing a product that no longer exists in that market.
PROS / Conga announcement → · full audit: Signal #001 →
MCP & Agent Interfaces21 Aug 2026
ActiveCampaign documents a first party MCP server with real write actions
A per account remote MCP server, generated under Settings → Developer, supporting writes across contacts, lists, tags, automations and campaign sends.
More than the rule based automation layer this category is usually credited with. Worth revisiting any "automation only, not agentic" assumptions about marketing automation vendors generally.
ActiveCampaign developer documentation → · full audit: Signal #001 →