Skip to content
Vol. 1 · Weekly editionWeekly · in your inbox
inklaretaal
Vol. 1 · No. 202638

Sales edition

in·klare·taal
Edition 202638 · Monday 14 September 2026 · Clarity since 2026

Salesforce is putting ready-made sales agents on the market, the bosses of AI are hitting the brakes themselves, and a judge says: what you type into ChatGPT can lose its confidentiality.

What changes for you
Een stapel gegevens: data

Salesforce delivers seven AI colleagues, two do your groundwork

Hunter hunts for new leads on its own, Piper catches incoming requests: ready to use, so no build project needed.

On 11 September, Salesforce introduced seven AI agents, each with a name and a job: Casey, Paige, Carter, Hunter, Marshall, Piper and Fin. They take over chunks of work, from customer service to incoming sales and the supply chain. Six are generally available right away. Hunter, the agent for cold outreach, is still in pilot. It will only become generally available in November 2026. For salespeople, two names matter most. Hunter works through a pipeline, from research to first contact, and keeps working with salespeople on one goal for weeks at a time. Piper works through your website and your inbox: it qualifies incoming leads and prepares them for a conversation. Both agents connect to Customer 360, the pool of customer data you already have. Because of this, little setup is needed before they can start using your customer context. The new part is not "AI in the CRM", we already had that. The new part is that you no longer build an agent, you switch one on that already has a job description. Salesforce claims it works: Agentforce's annual revenue is above 1.5 billion dollars, more than 240 percent higher than a year earlier, with 3.2 billion "agentic work units" processed last quarter. Next week, on 15 and 16 September, Salesforce will speak about this at Dreamforce, with a keynote from Marc Benioff on the "Agentic Enterprise". The agents draw their knowledge from Customer 360, where your customer data already sits: contacts, deals, conversations, purchases. Because the agent can access that directly, nobody needs to build a link first or fill in questionnaires. Hunter works out on its own who is worth approaching, drafts an approach and keeps that line going. Piper does the opposite: it catches what comes in through forms and email, asks follow-up questions and separates serious requests from the rest. So what comes out is only as good as what is in your CRM. Messy fields and old contacts give you messy outreach. There are two limits. Hunter is not generally available yet, so what you are reading now is a promise with a date in November. And an agent that makes contact on its own is making contact with your customers, under your name. A generic opening email to an account you have held for years will cost you more than it earns. So agree which accounts the agent should not touch, and let every message pass by a human for the first few weeks. Ask your admin this week which of the seven agents are in your licence, and who decides what they are allowed to send.

Source

Background for subscribers
What this means for you
Ready-made agents are changing how you approach customers

This edition is about one shift: AI agents that do the work themselves, not just give advice. Salesforce delivers Hunter and Piper as ready-made sales colleagues, LinkedIn and HubSpot are building similar help for prospecting, and this only works well if your CRM is in order. At the same time, the builders of AI are themselves warning against too much trust in ever-faster models, and a court case shows that what you type into a chatbot doesn't always stay confidential. For you, this means: use the tools to find the right customer faster, but keep a close eye yourself on what an agent sends on your behalf, what you promise about future features, and which customer information you share where.

Het programma dat de AI laat werken: een model

The bosses of AI hit the brakes themselves: don't count on ever-faster tools

This weekend, Sam Altman (OpenAI) joined Dario Amodei (Anthropic) and Elon Musk in calling for the deliberate slowing down of development of powerful AI models. Altman wrote on X that no American competitive pressure justifies recklessness. Amodei presented a three-step plan, including independent overseers who get almost employee-level access inside the big AI labs. Earlier that week, Altman told his own staff that OpenAI is open to slowing down. Markets reacted with shock: SoftBank fell 13 percent in Tokyo on Monday, Samsung Electronics and SK Hynix dropped more than 4 percent in Seoul. For salespeople, this is not an abstract debate. It says something about the assumptions in your pitch to the customer. If you sell software or services with an AI promise, there is often a silent assumption inside it: next year the model will be able to do more. If the builders themselves want to slow the pace, that assumption becomes shaky. So sell what the tool does today, with today's price and today's result. Customers who later feel misled by a roadmap promise cost you twice: the deal and the relationship. Cut the line 'that will be possible soon' from your pitch, or state clearly that it does not exist yet.

Source

Background for subscribers

This is a taster.

Subscribers read the whole Sales edition: every story, each with its background in plain language.

Subscribe →

Not ready for that? Read the free letter first →

inklaretaal

AI and tech, made simple.

This is the Sales edition. Subscribers get the background to every story, in plain language.
Subscribe
inklaretaal · Amsterdam · © 2026