“These agents are not tools. They are employees. They work 24/7 to analyse data, make decisions, take action, and we can all start to imagine this company managing millions of customer interactions every day with Agentforce, seamlessly solving problems, processing transactions, anticipating customer needs, freeing up humans to focus on strategic initiatives and building meaningful relationships,” enthused Salesforce founder, chairman and CEO Marc Benioff at his third quarter results presentation, in a call during which he cheerfully mocked Microsoft’s Copilot. “And that’s going to evolve into the current customers, whether it’s a large hospital or a large hotel where not only are agents working 24/7, but robots are also working side-by-side with humans. It’s all happening before our eyes and it's not just a distant future. It’s happening right now.””

“Companies like Fedex, Adecco, Accenture, Ace Hardware, IBM, RBC Wealth Management and many others are now building their digital workforce on the Salesforce platform with Agentforce. So the world;s biggest and most important companies, in every geography and every sector, are now creating and supplying agents,” he added.

Slack: the starting point for software integration

On Wednesday, the company went one step further by deploying its AI agents in Slack, the collaborative enterprise tool. Over the years, businesses have accumulated software to meet their needs and employees spend their time navigating from one to another, transferring data or looking for solutions. With the integration (plugging in) of all or some of this enterprise software into Slack and with the addition of a layer of AI agents, an employee could increasingly easily find the information they need in Slack. The AI agents themselves could automate certain tasks depending on certain situations, and what’s special about this feature is that the AI agents learn as they go along, which improves their results. They will also be able to personalise their responses for each user.

“These agents can, for example, track leads, conduct customer interviews, give feedback, schedule service appointments, provide forecasts and data visualisations, send messages directly into Slack, share project overviews or update a canvas,” says Salesforce. Third-party partners (Appexchange) can enhance these skills with their own offerings (e.g., Asymbl, DocuSign, Neuron 7), broadening the scope of agents’ activities.

A new set of turnkey skills is available within platforms such as Salesforce CRM, Tableau and Slack. These pre-built skills enable the rapid creation of agents capable of performing specific tasks without the need for complex development.

Before its acquisition by Salesforce in 2021, Slack had 600,000 organisations using it, including 110,000 with paid accounts.

“Indeed is the world’s largest job search site. Every second, three people find a job through our platform. Indeed’s success is based on our unique ability to match 580m jobseeker profiles with more than 3m employers looking for talent,” explains Indeed CIO Anthony Moisant, in Salesforce’s official communication around the deployment of its AI agents in enterprise collaboration tool Slack. “Thanks to the power of Salesforce Data Cloud, we have optimised our data architecture and now have a solid foundation to refine our go-to-market strategies and strengthen connections with the rich profiles of our job seekers. Agentforce allows Indeed to stay focused on its mission of helping people find jobs, while achieving its goals of reducing the average time to hire by 50 percent and helping 30m people facing barriers find jobs by the end of fiscal year 2030.”

Why is this interesting?

Saves time: instead of switching from one tool to another, users stay in Slack and get what they need straight away. No more fumbling around in different systems or asking a colleague.

Fluidity and simplicity: the Slack interface is already familiar. Users ask a question as if they were talking to a colleague. The AI agent responds clearly, without unnecessary technicalities.

Improved productivity: recurring tasks (consulting customer data, generating a report, checking the status of a business opportunity, etc.) can be automated. This allows staff to concentrate on activities with higher added value.

Adaptation to the organisation: the AI agent is not just a generic chatbot. It can be personalised, trained with internal data, business processes and specific jargon, to become a truly specialised assistant.

Why is this problematic?

Cost and return on investment: companies will be wondering how cost-effective it is to implement autonomous AI agents, given the effort involved in configuration, training and maintenance.

Deployment complexity: although Salesforce promises simplified integration with Mulesoft and Slack, some critics may point to the difficulty of customising and scaling agents to very specific needs.

Accuracy and quality of responses: it is possible that some commentators will test the robustness of the Atlas reasoning engine and report cases where the agent does not provide the most appropriate response, or where the quality of reasoning falls short of expectations.

Confidentiality and data governance issues: in the current context, concerns about security, privacy and regulatory compliance (GDPR, etc.) are common.

Others will add considerations about the replacement of humans by these agents, to which the company responds that this mainly makes it possible to improve human productivity. Not replace it.

This article was originally published in .