AI agents enhance productivity, but also expand the attack surface.
aging
When the pandemic forced the displacement to remote work, many are supposed to be temporary. But “everywhere workplace” quickly became permanent, because workers and organizations realized that flexibility and productivity profits were valuable to abandon.
We are at a similar turning point now with the rise of workplaces-driven by AI, where autonomous digital assistants not only receive information, but actually act on behalf of employees. These agents are planning meetings, document drafting, analyzing conversations and negotiations on all platforms. The job is not just faster. In some cases, he feels like he is running himself.
This is exciting, but it also creates a well -known concern. When something feels unhindered, it is usually because the complexity – and the risk – is hidden.
A new blind seat spot
Agentic workplaces are strong, but they are also opaque.
If an AI agent sums up a confidential meeting, where are this data stored? If he planning a legal or financial analysis, who else can access the framework behind it? Without viewing these flows, organizations trust the processes that cannot be fully seen.
It reminds me a bit of the double sword of QR codes. They explained during the pandemic as a way without friction to access menu or pay bills. But this simplicity covers a serious danger: people have prepared to trust QR codes without knowing where they really drive. The workplaces of the practices have the same dynamic. The convenience is undeniable, but trust is often blind.
Scott CrawfordThe head of the Information Security Research at 451 Research / S&P Global collapses because this blind spot is different from the previous challenges: “Agentic AI is distinguished by three basic possibilities: Modeling of reasoning based on inflows and the available memory framework, Practical actions can make it more difficult for organizations to determine restrictions that mitigate these risks. “
This unpredictable makes visibility and protective messages necessary – and explains why security sellers are struggling to deal with it.
The first approach of the ProofPoint industry
A blog by the Managing Director Proofpoint Sumit Dhawan It defines a cohesive strategy for the challenges of data cooperation and security in the emerging workplace and describes a series of solutions and corporate relationships from Proofpoint to allow it.
Proofpoint is the first major security seller Announce a solution expressly designed for this new environment. Instead of laying more checks on end points or networks, they incorporate security directly into the flow of conversations, documents and actions leading to the c.
This is important because in a workplace, the “user” is not always a person. He is an autonomous agent who makes decisions on behalf of a person. Traditional security checks were never made for this scenario. By expanding visibility and protection to these work flows, Proofpoint aims to preserve both cooperation and sensitive data exchanged.
Lessons from everywhere workplace
The parallels with the displacement of the workplace driven by the pandemic are impressive. Investigations have shown that 87% of employees never wanted to return to the full -time office and many were still willing to exchange compensation for flexibility. Leaders who assume that the change was temporary had been caught unprepared.
Agentic workplaces are on a similar orbit. As soon as employees experience productivity profits to unload repetitive work on AI, there will be no return. Organizations facing it as a short-lived experiment are at risk of being re-found-at this time from security gaps.
Where the real risks will occur
Todd thiemannThe main analyst of Omdia, points out that many current AI agent developments occur in SAAS applications such as SalesForce Agentforce, working day agents and Microsoft Copilot. These “Walled Gardens” offer some security because sellers control authentication and authorization. But the real challenge will come when agents begin to interact directly with applications, where they can access more sensitive systems and data.
Thiemann warns: “The real value for the Agentic AI will come when agents begin to touch the basic applications of businesses, where businesses will unlock a significant value of AI, but also face a significant risk of security from AI agents. Agic AI expands the surface of the business.”
This risk is exacerbated by the speed of adoption. Businesses are largely investing in Agentic AI, often rushing pilots into production to demonstrate the ROI. Security is usually a second thought. Excessive factors or weak protective messages could quickly lead to incidents such as data violations or fraud. Thiemann believes he may need a high profile safety failure to push the organizations into strict controls.
Because it echoes
I have written for years on technologies that promise to facilitate life, but to introduce new dangers. New technologies usually follow a well -known pattern. We embrace the benefits quickly, but security is usually delayed. Proofpoint’s announcement feels like an attempt to break this cycle of building Guardrails in workplaces from scratch.
Crawford made the point clearly: “There is no time like the present for sellers to play a leading role in providing assistance to organizations to find functional ways to contain these risks and help the agency functionality fulfill its promise.” Proofpoint enters this role and others are likely to follow.
The road ahead
Agentic AI is not a fury – it is the logical next step in automation. We have moved from macros to chatbots to agents who adapt, decide and act. This displacement makes the data layer to secure, not only the perimeter, necessary.
If the story is any driver, the standards and the best practices will eventually appear to determine how to secure the workplaces of the techniques. But for now, organizations do not have the luxury to wait. Sellers and businesses that move early will form these protective messages – and give themselves a start while maintaining this new era of safe work.
