7xAIan All7x Solutions Brand

SAP Agentic SOC

Video: The Agentic SOC and SAP

Explainer from Alessandro Banzer's article on the agentic SOC and SAP. Briefing based on Alessandro Banzer, US President of Xiting.

The briefing explains why SAP alerts stall in a security operations center and how an agentic model can triage, investigate, and respond. It uses Alessandro Banzer’s account of that gap, including the cost of detection without a fast response, and ends with three questions for leaders.

Read the article

Transcript

What this briefing covers (00:00:01)

This briefing is based on a recent LinkedIn article by Alessandro Banzer, US President of the SAP security company Xiting, on what agentic security operations mean for SAP.

SOC and agentic AI (00:00:15)

Start with two definitions. A security operations center, or SOC, is the team that monitors an organization around the clock and responds to attacks. Its core tool is the SIEM, short for security information and event management, which collects logs from across the business and raises alerts. Across the industry, SOCs are now being rebuilt around agentic AI: software agents that carry out multi-step work on their own, such as triaging alerts, assembling investigations, and increasingly, executing the response. People set the standards and handle the exceptions.

Why SAP alerts stall (00:00:52)

Banzer's argument is that SAP has been left out of this shift. SAP is the enterprise software that runs the ledger, payroll, the supply chain, and customer master data. It also writes some of the best security logs in the company. Yet in most SOCs, an SAP alert arrives, the analyst on shift cannot interpret it, and it is closed with a note: SAP team to review. Nobody on the SAP team ever sees it.

The Louvre example (00:01:23)

Banzer illustrates the cost with the Louvre robbery of October 2025. Thieves used a furniture lift to reach a gallery balcony and took eight pieces of the French crown jewels, worth 88 million euros, in about seven minutes. The alarm worked. The response was too slow: the control room lacked screens for every feed, police were first sent to the wrong place, and the thieves left about thirty seconds before anyone arrived. Detection without a timely response leaves the organization with a precise record of its loss.

Three reasons the gap is wider (00:01:57)

In SAP, Banzer identifies three reasons that gap is wider. First, the signals require SAP knowledge. A debugger session in production, an emergency firefighter account used outside its approved window, or a remote connection changed at two in the morning only means something to someone who understands SAP. Second, the people with that knowledge, the Basis team that administers SAP systems and the SAP security staff, work office hours and are rarely on call. Third, nobody has built the response actions. Isolating a laptop is one click in endpoint security tools. Locking an SAP user and ending an emergency session usually has no equivalent button.

Attackers are faster (00:02:44)

Meanwhile, attackers are accelerating. In November 2025, Anthropic documented an espionage campaign in which an AI agent performed 80 to 90 percent of the work. An agent like that can read decades of custom SAP code in an afternoon.

One alert, six missing answers (00:03:02)

Consider one alert as a SIEM shows it: a remote logon by a legacy communication user, calling a function that can read any table it is authorized for, in production, at 3:14 in the morning. Every field is accurate. Judging it requires six answers that appear nowhere in the log: whether the user should do this, whether the function is dangerous here, what data was touched, whether this has happened before, whether it is a scheduled job, and what to do in the next five minutes. By hand, that takes thirty to sixty minutes.

Falcora: triage, investigation, response (00:03:37)

Xiting applies the agentic model to SAP through its product, Falcora, in three stages. Triage: every SAP alert is classified by AI, with user names pseudonymized and the customer's own procedures applied, as a false positive, an improper operation, or malicious. Investigation: an agent runs read-only lookups across audit logs, change records, and emergency sessions, then writes a conclusion with recommended severity and next steps, and every step is recorded. Response: autonomy is earned through four steps, observe, recommend, approve, and automate. Actions with real business impact require human approval. Xiting's own benchmarks for Falcora report up to 70 percent less investigation effort per SAP alert, and Banzer advises measuring it in your own landscape.

What changes for people (00:04:33)

The effect on people matters as much. SAP security experts move from reading logs to designing controls. The SOC takes ownership of SAP alerts. Emergency access is reviewed as it happens, and audit evidence becomes a by-product of daily work.

Three questions for leaders (00:04:51)

Banzer closes with three questions for leaders. Which SAP log sources reach your SOC today, including development systems? Who reads an SAP alert at three in the morning, and can they judge it? And what would you allow a machine to do in production without asking first?