Being stuck in the ticket queue was always a pain. Yet it's begrudgingly tolerated for lack of a better alternative.
It stopped being tolerable for software engineering once software started eating the world. Cloud was that better alternative, because it let us automate all the things. And after years of searching for the right abstraction to get both speed and quality, we landed on platform engineering.
That's almost two decades of my career summarized in three sentences.
What many don't seem to realize: in the agentic organization, tickets stop being tolerable. Period.
Where I'm coming from
As software was eating the world, from PaaS to platform engineering, I've been all about helping dev teams skip that ticket queue.
And assuming AI doesn't destroy the world, organizations will put agents to work wherever they make things faster. With that, it's no longer just software development that outpaces the ticket queue, it's everything agentic.
Now some may argue agents resolving tickets will speed up the queue.
But a speedier queue is still a queue.
Self-service is what got software teams out of the queue, and I believe what I learned doing that will matter even more for the agentic organization. That's the bet I made when I quit my job to work on Accession1 full-time.
What I'm building
Accession1 makes resource provisioning and access management self-service, continuously reconciled, and context-aware.
Self-service means users can create resources and share them with human and agentic team members within boundaries set by administrators without waiting for a ticket.
Continuously reconciled means identities synced from your identity provider are mapped to resources by matching labels, so any change to identities, resources, or the organizational structure is reflected in grants and resource configuration across all tools automatically.
Context-aware means policies are designed with AI assistance, based on the organization's legal, regulatory, and business realities. As the organization changes, the AI takes over the tedious part of keeping that context and the taxonomy current.
AI assists, humans approve, and the AI is never in the path that grants access. That path is deterministic policy reconciliation, and every grant it produces is auditable.
How it compares
Agentic AI creates new challenges for IT and security teams. And there is no shortage of vendors and open source projects providing micro VMs, harnesses, agent discovery, tool call auditing, and access-enforcing MCP gateways. Many of these capabilities will be required. But those teams are already managing too many tools. The answer can't be to just add more.
Accession1 closes the loop, so you can control agentic AI security as one instead of playing whack-a-mole.
Top comments (0)