Agentforce: a new category of work or a new narrative for enterprise AI?

Since the launch of Agentforce, Salesforce has been trying to expand the definition of what an enterprise platform can do. The proposition moves beyond people using features and toward agents capable of interpreting context, making decisions within limits, and executing tasks.

View the full Salesforce analysis

This article complements the historical analysis of the brand.

A new category needs to change behavior, not just terminology

Enterprise technology goes through cycles of terminology. Some terms describe real changes. Others reorganize existing capabilities under more attractive language.

The difference appears when technology changes how companies buy, operate, measure, and organize work.

That is what happened with SaaS. The term gained strength because access, infrastructure, billing, updates, and the relationship with the supplier changed significantly.

Agentforce needs to pass a similar test.

First question: what is the unit of value?

In traditional software, value is often measured by access to features, users, capacity, data, and supported processes.

In an agent system, the promise shifts toward work completed. The agent receives an objective, consults information, decides on a sequence of actions, and executes tasks within a set of permissions.

If companies begin measuring value mainly by completed tasks, time returned to teams, additional capacity, and economic results, there is a meaningful change in the unit of value.

Second question: was the workflow actually redesigned?

Adding an assistant to an existing interface can improve productivity without creating a new category.

The change becomes deeper when the process is designed on the assumption that agents will be present from the start. In this scenario, people define objectives, exceptions, and limits, while systems execute operational steps and escalate cases that require human judgment.

This distinction connects to the analysis of how to use AI without making your text generic. The principle is similar: AI creates more value when it takes on a clear function within a process, and less when it is added superficially just to signal modernity.

Third question: has a new management layer emerged?

The more autonomy a system receives, the greater the need for governance.

Companies need to know who can create agents, which data each agent can access, which actions are allowed, how errors are identified, when a person needs to take control, and how decisions can be audited.

This operational layer is one of the most important signals. When organizations begin managing agents as permanent resources, with their own policies, metrics, and responsibilities, the change goes beyond the interface.

Fourth question: does human work change function?

The narrative of an “agentic” company proposes that professionals coordinate a combination of people, automations, and agents.

This can change managerial roles. Part of the work stops being direct execution and begins to involve defining objectives, reviewing exceptions, quality control, and system design.

The change can also generate resistance. The analysis of loss aversion helps explain why perceived threats to status, autonomy, and existing skills may weigh more heavily than future benefits that are still abstract.

Credibility depends on operational evidence

A category does not become established because a company declares that it exists. The market needs to observe repeatable results.

Some signals are especially important.

• Relevant tasks completed with consistent quality.

• Measurable reduction in time or cost.

• Governance sufficient for regulated and complex environments.

• Integration with real systems, not just isolated demonstrations.

• Clear human escalation when agent confidence is insufficient.

Without these elements, the risk is that “agent” becomes a new label for familiar automation and assistance.

The parallel with Salesforce's early years

In the company's early years, Salesforce also presented a transition between eras. Installed software represented the past; internet-delivered service represented the future.

Today the narrative structure repeats itself. Work concentrated in people and traditional automation would be limited; digital agents would represent a new operational layer.

The difference lies in the company's position. Today's Salesforce has to prove the change within a large ecosystem of accumulated customers, data, partners, and processes.

The answer still depends on what companies do

Agentforce already has a clear narrative and a product infrastructure around agents. The open question is the depth of the transformation in everyday operations.

If agents become a permanent unit of work, with their own budget, governance, metrics, and organizational design, the idea of a new category gains strength.

If they remain mainly an interface and automation layer within existing software, the change will still be important, but closer to an evolution of the current category.

Continue exploring EuAlan

Also read

How to use AI without making your text generic, Loss Aversion and Authority Bias.

#sf