Embedded AI
and Joule Assistant
in SAP SuccessFactors

When Mango unveiled an advertising campaign entirely generated by artificial intelligence in 2024, reactions were immediate and extreme. Admiration for the technology was mixed with outrage. The campaign did not involve a classic creative background. Instead, there were AI specialists, and the model… did not exist. There was a question that returns today in the discussion of artificial intelligence. Will AI take our jobs? The reaction of many customers was also critical. Will an outfit that looks great on a generated model lie equally well on a real person?

The case of the Mango campaign is an apt metaphor for what is happening around AI in business.
First – fear of losing jobs. Second – lack of confidence in the effects of artificial intelligence.

Lack of confidence in the results of artificial intelligence is not a contrived problem – it has been well described in scientific research. A high-profile article on the so-called transparency dilemma (The transparency dilemma: How AI disclosure erodes trust, Organizational Behavior and Human Decision Processes, May 2025) analyzed more than a dozen experiments involving the use of AI, including in four HR scenarios: recruitment, employee evaluations, hiring or layoff decisions. The conclusions were disturbing. When artificial intelligence was used in these areas, confidence declined not only in the decisions themselves, but also in the people who used AI. For example, managers supported by AI in appraisals were seen as less trustworthy and their decisions as less fair. This leads to the conclusion that the incompetent implementation of AI in HR processes can generate real psychological and social costs.

The authors of the aforementioned study did not stop at diagnosis. They also pointed out how to minimize the risk of mistrust. The key concept here is the so-called collective validity of AI, which is achieved when the use of artificial intelligence becomes an organizational norm rather than an exception that arouses suspicion. The conditions are two. First, AI must be available not only to managers, but also to other employees. Second, its credibility should be confirmed by internal authorities – people who have real influence over opinion in the organization.

Fear of losing one’s job is the other universal dilemma associated with AI. Lists of occupations at risk of automation, published by Microsoft Research, among others, circulate regularly in the media. HR specialties are not on there. Why? Because HR has ceased to be a purely administrative function. It is today an area based on communication, empathy, conversation and relationship building.

Of course, where HR is limited to handling administrative processes, the risk of job cuts is real. But for the modern HR function, artificial intelligence is becoming a support rather than a substitute. This model is well illustrated by the “centaur" metaphor proposed by Garry Kasparov: a human supported by artificial intelligence is more effective than each of them separately. HR in this view does not compete with AI, but uses it as if it were an exoskeleton.

HR analyst Josh Bersin promotes a similar concept, talking about “super employees" and “super managers." – people who can use artificial intelligence in their daily work. This makes them more effective than employees without AI enhancement.

In SAP SuccessFactors, the vision of the super-employee is realized in two ways. On the one hand, through embedded AI – that’s currently 40 AI scenarios supporting recruitment, development, job descriptions or competency matching. On the other hand, through the intelligent assistant Joule, which enables communication with the system in natural language.

SuccessFactors’ AI applications are surprisingly down-to-earth – and that’s where their strength lies. Generating SMART goals, analyzing the competency gap against the role an employee aspires to, or automatically prompting development activities. On top of that, there are personalized recommendations for training, projects and mentoring in the Opportunity Marketplace. And, of course, Joule, which can answer questions, guide a user through a transaction or initiate a workflow – from a vacation request to changing employee data. These tools speed up the workflow, increase the quality of results and reduce the cognitive cost of using the system.

We are on the eve of a most interesting development at SuccessFactors. SAP is intensively developing AI agents – autonomous entities that execute complex processes with minimal user input. The first of these supports the preparation of employee evaluations. In practice, this means that the manager does not start with a blank sheet of paper. The agent collects data from the system, analyzes the employee’s performance history, previous goals, 360° feedback, progress (recorded in Continuous Performance Management), and generates recommended issues for conversation. It also initiates specific actions: scheduling a 1:1 meeting, or sending a request for additional feedback. This dramatically reduces the preparation time of the conversation, but most importantly, it increases its quality.

IBM’s experience in implementing the AskHR solution shows that technology alone does not guarantee success. IBM was one of the pioneers in the use of AI in the HR field, and quickly learned that the key challenge is not about the algorithm, but about adaptation. The first implementations revealed a classic problem: users expect a conversation similar to human interaction from an intelligent assistant, not communication more akin to filling out a form or spreadsheet.

There was a need for quality improvements, a change in the approach to communication, and better preparation of the organization for the change. IBM gradually moved away from the “big bang" deployment model to pilots, iterative fixes and intensive user support. A key role was played by managers and internal opinion leaders, who supported the dissemination of the solution with their authority. The result? Significant improvements in user satisfaction ratings in subsequent years (from NPS -35 to +75).

Artificial intelligence in HR is becoming part of an organization’s decision-making infrastructure. And that is why the most important question is not about functionality, but about accountability. AI today can collect data faster, analyze it more broadly and propose solutions more efficiently than a human. However, it will not bear the consequences of decisions or take responsibility for their impact on humans. That still remains with us.

So perhaps the real challenge is not to implement artificial intelligence in HR, but to redefine the role of the manager in a world where some of the operational thinking is taken over by an algorithm. And this is no longer a technology question – it’s a question about the maturity of the organization.