For most of enterprise software’s history, companies had a clear model for how technology entered the organization. IT selected applications, provisioned access, and put security controls around a known set of systems. AI is changing that because employees can now adopt powerful software themselves, connect it to company data and tools, and increasingly allow it to act on their behalf.
Before starting Kelasa, I spent several months working with companies on AI transformation. I expected adoption to be the hard part, but found the opposite: employees were already using AI, leadership wanted more of the productivity, and IT and Security teams were looking for ways to enable it. What they lacked was a way to say yes without losing visibility or control.
AI Needs a Control Plane
Today, most employees still use AI to write, research, analyze information, or generate code. The larger shift will come as AI begins participating in workflows by searching company systems, updating records, running code, requesting approvals, and completing tasks across multiple tools.
That changes what access control means. Knowing that an employee can access Salesforce does not tell you what an AI acting for that employee should be allowed to do inside Salesforce. A user may legitimately be able to issue a refund, delete a record, or deploy code, but every agent they use should not automatically inherit the same authority.
As AI becomes more capable, companies will need to know which AI is acting, on whose behalf, what it is trying to do, and whether that action should be allowed. Because employees will use many models, clients, agents, and AI-enabled applications, those controls cannot live inside any one AI product. The company should not need to own the interface; it should own the rules.
We believe that requires an independent control plane for AI.
Trust Creates Adoption
A lot of AI security starts with how to stop employees from doing something unsafe. We think the more important question is how to make it safe for employees to use much more AI.
If someone wants to connect an AI assistant to Salesforce, the ideal outcome is not simply to block it. It is to give them a trusted way to do it, with the right identity, permissions, policies, approvals, and audit trail around the actions that AI takes. Good security should not just stop the wrong thing; it should make the right thing easier.
Kelasa helps IT and Security teams understand how AI is being used, make approved tools and capabilities easy for employees to access, and enforce consistent controls around what those systems can see and do. Employees can keep using the AI products that work best for them, while the company has one place to define the rules.
At Integration Labs, we built infrastructure that let software safely access external systems and data. AI brings many of the same problems back, with one important difference: the software accessing those systems can now decide what to do next.
Trust creates adoption, and our goal is to give companies the confidence to say yes to more AI.
Build a Trusted Path to AI
We believe every employee will eventually work with many AI systems, and that the biggest gains will come when AI begins participating in the actual work of the company. The models and interfaces will keep changing; the trusted path underneath them should not.
If you are an IT, Security, or AI leader figuring out how to safely roll out AI, we would love to work with you and show you what we are building.
Best Regards,
Ishwar