An AI agent that can independently complete tasks, process data across systems, and knows when to involve a human. Not a concept. Not a prototype. A working solution already used in real operations today. Piotr Lobaczewski, Regional Leader at Salesforce, spoke about the future of AI in business during his keynote at our Digital Transformation Summit 2025. Even though the keynote was delivered in the Czech Republic, the vision and examples resonate with businesses across industries and geographies.
More than technology. A real colleague on your team
We are entering the third wave of artificial intelligence. After predictive and generative phases comes the era of autonomous agents. Agents that can make decisions, communicate, perform tasks and collaborate with human teams.
Piotr Lobaczewski
Salesforce
According to Salesforce, Agentforce is the answer to what companies truly need today: a smart, transparent, scalable solution that does not require years of development, custom models or dozens of data specialists. Instead, it is built on a combination of AI, data and human collaboration.
Atlas Reasoning Engine: an agent that understands
Agentforce is not a new Star Wars series. It is a carefully designed solution helping companies of all sizes scale AI and autonomous agents.
At its core lies the so called Atlas Reasoning Engine. As Lobaczewski put it, it “thinks like a human”: it understands context, works with data, evaluates situations, and proposes or directly executes the right actions.
In practice, if an order is delayed, the agent can check its history, verify warehouse status, review compensation terms, and suggest a solution such as a partial refund or an alternative product. All without human intervention.
And if the agent cannot resolve it, it passes the conversation to a live operator.
The agent does not replace people, it complements them. A digital colleague within the team.
But automation must not mean isolation. Customers should always be able to reach an agent if needed. Monitoring drop offs in digital journeys helps refine processes and improve the overall experience.
“Agentforce brings together all the key elements: people, artificial intelligence, data and the actions required. Their integration is what makes agents efficient and meaningful,” added Lobaczewski.
Roles, data, actions and boundaries
One of the biggest advantages of Agentforce is the ability to precisely define how the agent operates within a company. Is it an assistant, a fast operator, or an analyst?
Each company decides:
- what role the agent plays
- what data it has access to
- what actions it can trigger
- where its boundaries lie, for example what it cannot communicate or modify
That makes Agentforce not just a technical tool but a clear framework for human and AI collaboration, whether deployed in marketing, sales, or service.
Real life demo: Wilson at WY bookstore
One of the most engaging parts of the keynote was a live demo of Wilson, an agent running in WY’s online bookstore. A customer asked whether they could download a textbook to their phone and study offline, and Wilson replied.
Then came a tougher request: Can I add a book to an existing order? Wilson could not handle it yet and passed the query to an operator. That is when Lobaczewski showed how easily the agent’s skills can be expanded.
In Agent Builder, a no code graphical interface, he added new instructions, introduced two actions, order verification and modification, and launched a test conversation. The result was that Wilson completed the new task successfully. No coding required.
Another benefit is that the agent is not a fixed cost. You pay for usage, for example only during peak season. This gives companies more flexibility, faster reactions to unexpected demand, and careful cost control.
Thanks to Agentforce, WY handled 40% more customer requests without hiring new staff.
A good agent starts with good data
“Without quality data, there cannot be a quality agent,” Lobaczewski stressed. That is why Salesforce builds on Data Cloud, which unifies structured and unstructured data from multiple systems without physically copying it.
Thanks to a zero copy approach, data can be used where it resides, avoiding costly transfers or replications. An agent can, for example, pull information from SAP in real time, check inventory, or issue a refund. Likewise, it can work with data stored in Snowflake, Google Cloud, Oracle, or AWS.
Another strength of Agentforce is its openness to other technologies and ecosystems. Agents built on this platform can communicate with those from other vendors such as IBM, Workday, or industry specific solutions. This enables companies to build connected environments without barriers and avoid vendor lock in.
As Lobaczewski summed up: “We do not need to teach the model to handle data. We just need to prepare the data so the agent can use it effectively.”
And what about pricing and real world use
Salesforce has moved Agentforce to an outcome-based pricing model. Companies can either pay per conversation or use Flex Credits, where each completed action consumes credits. For larger deployments, there are also add-on licenses and full editions that bundle credits with advanced features. This flexible approach makes costs easier to control and scale, whether for seasonal peaks or company-wide rollouts.
Want to see more details or watch the full keynote with live demos?
If you’d like access to the full recording in English, let us know. And if you’re facing similar challenges and want to explore how this approach could work for your business, book a free consultation with our colleague Martin Havelka, Business Unit Leader for the Netherlands.
