Claude becomes the CRM front door

Salesforce made AI the earnings story

Salesforce entered the fortnight with scrutiny on Agentforce. On 22 Aug, The Register reported that Salesforce partners were not seeing significant revenue generation from the Agentforce AI platform. On 25 Aug, further press coverage said the AI agent platform had not delivered meaningful growth two years after launch, according to customers cited in that article.

Then the company changed the frame. On 26 Aug, Salesforce announced an expanded Anthropic partnership and a product called Claudeforce. On 29 Aug, it integrated its entire CRM into Claude and launched the Claudeforce AI sales plugin with Anthropic. On 27 Aug, its homepage hero moved from “Get started with the #1 Agentic CRM” to “The #1 AI meets the #1 CRM”. That small wording change matters because it moved the promise from a Salesforce-native agent category to a broader AI plus CRM claim.

The market response was unusually loud. On 27 Aug, Salesforce reported record second quarter fiscal 2027 results and raised its full-year outlook. On 28 Aug, it said Agentforce ARR had passed $1.5bn and again raised its fiscal 2027 outlook. Press coverage across 27 to 30 Aug reported stock rises of 20% to 22%, with other reports citing a 14% jump after Q2 results and strong AI momentum. The week also brought sales and marketing leadership motion: Kaylin Voss returned from OpenAI on 22 Aug, Salesforce hired a second sales leader from OpenAI on 25 Aug, Erin Oles was promoted to President and Chief Marketing Officer on 26 Aug, and Jacqueline Cone became Senior Vice President, Global Field Marketing on 27 Aug.

Claude is becoming distribution, not plumbing

The most important competitive line this fortnight was not simply “who has AI”. It was where the CRM shows up when the seller asks for help. Salesforce put the whole CRM into Claude on 29 Aug. Pipedrive’s MCP Connector was made available in Claude’s Official Marketplace on 22 Aug, with three press articles covering the launch. HubSpot introduced an AEO tool for AI visibility in search, reported on 23 Aug and again on 27 Aug. These are not identical moves, but they share a direction: revenue software is being pulled into AI workspaces and AI discovery surfaces.

Pipedrive paired that distribution move with capability expansion. On 22 Aug, it completed the acquisition of Outfunnel to integrate sales and marketing data sync into its CRM and expand native sales and marketing automation features. The same event set reported $189mn ARR. The deal was again observed on 23 Aug and 25 Aug, with no disclosed deal value. On 22 Aug, Pipedrive also had 16 open roles, all in engineering and machine learning functions, including Principal AI/ML Scientist & Engineer, Principal Decision Scientist, Data Protection Officer and Principal Security Engineer.

The implication is direct. Claude is no longer just a model supplier in this market. It is becoming a place where CRM vendors seek placement, workflow and credibility. Salesforce took the flagship route with Anthropic and Claudeforce. Pipedrive took the connector route. HubSpot attacked the discoverability layer with AEO. For operators, the question is no longer whether AI sits in the roadmap. It is whether the product can be reached from the new places sellers are beginning to work.

The quiet rewrites were more revealing than the launches

HubSpot’s public surface shifted in several small steps. On 24 Aug, it created two product pages tied to “poppy sandbox”. On 27 Aug, it added a “Smart Deal Progression Series A” blog page aimed at startups and tech stacks. On 28 Aug, it created a user persona template product page and its GitHub organisation added a new “hubspot-cli” repository, after adding “agent-cli-skills” and “boomslang” on 25 Aug and “hbase” and “slack-client” on 26 Aug. On 29 Aug, HubSpot removed its data quality software product page.

None of those items alone would define HubSpot’s strategy. Together, the timing is harder to ignore. In the same week that HubSpot press coverage centred on Breeze AI, Agent Hub, the AEO tool and an earnings rebound of 7.9% on 28 Aug, its website moved away from one named data quality page and towards templates, startup content, sandbox pages and developer-facing repositories. That is the kind of packaging drift that shows up before it becomes sales enablement language.

Attio’s changes were similarly concentrated. On 26 Aug, its changelog introduced new Workflows apps, new Workflows blocks, renamed system objects, a refreshed sidebar and developer platform updates. The same day, its sitemap added seven product pages covering those areas, and the website swapped older headings such as “Formula attributes are here” and “The mobile app just got faster” for “New Workflows apps”, “New Workflows blocks” and “Rename system objects”. On 27 Aug, Attio launched a ClickUp product page and removed an AndOpen product page. On 28 Aug, it added ClickUp help documentation and listed three Senior Platform Engineer roles in London and Poland. The product story is extensibility, customisation and integration density, not a single headline launch.

Close chose proof over platform theatre

Close had the least market noise among the five, but the clearest proof motion. On 28 Aug, it added a Devvo customers page and a case study titled “How Devvo Turns Dormant Leads Into an Extra 85 Car Sales a Month”. On 29 Aug, its case studies page added Giveffect and Flatiron School logos, with new stories for both. The Flatiron School case study reported a 75% increase in conversion over 60 days after rebuilding its sales motion in Close.

The product surface moved too. On 26 Aug, Close launched a “Playbook to Scaling Sales Teams” blog page and added product pages covering an event log, improved navigation for MacOS and Windows, and integrations with Callpal and Dedupely. On 22 Aug, its YouTube output covered scalable sales process design, coaching in sales leadership, the human side of CRM selection, AI in sales and why 6P Marketing chose Close over competitors. The same day, Close listed five remote US engineering openings, including backend, frontend growth and communications roles.

This is a different operating posture from Salesforce’s Anthropic week or Pipedrive’s acquisition week. Close is not trying to own the AI interface in the observed events. It is tightening the practical buyer case: conversion, dormant leads, coaching, integrations and workflow polish. In a market crowded with AI claims, that evidence stack has value.