Ledger A · The security record
Which agents exist, what they can reach, what they did, how risky it was.
- Owned by security
- Reported to the board as posture
- Silent on cost
What we’re solving
Your agents are already at work. The challenge is knowing what they’re doing
The business shipped agents in months. Making sense of the rollout has become one of the most important jobs in the business.
The enterprise gained a new labor ecosystem
Agents research, analyze, build and communicate. They make decisions, call tools and take actions across the same systems, data and workflows people depend on every day. Work is moving beyond a copilot in a sidebar, into software that decides for itself what to do next.
Which means every organization now needs a new understanding of the same operating environment: which agents are working, what they are doing, what they can access, how they are behaving, what is happening as they work, what it costs, and where attention needs to go.
Those questions span security, AI, technology and the business itself. Most organizations have no governance infrastructure built to answer them, because agents are systems, not surfaces – their configuration, identity, tools, permissions and activity span cloud, code, endpoint and browser at the same time.
Key challenges posed by the agent governance gap
-
01
You are governing the inventory, not the estate
Most teams cannot compile a complete, current list of what agents are deployed, where they operate and who owns them, so posture, compliance and control decisions are all made on incomplete data.
327% more agents than one customer’s own security team estimated
-
02
Audit logs are optional, and lack behavioral context
Not every agent produces logs, and those that do often allow the session to be rewritten. Without behavioral context, a risky session reads like routine work.
-
03
Agents span surfaces. Controls sit on one.
Endpoint-only sensors stop at the device, and gateways see only the traffic that crosses them. Neither can govern a system that lives everywhere at once.
-
04
Risk is financial too, and it compounds
Developer agents wasting tokens on side projects. Agents that keep running long after they should have stopped. Three departments paying separately for the same work. It comes up in every other first meeting, and often before security does.
You’re governing the agents you know about
Every organization we have deployed into found more agents than it expected, in one case 327% more. The gap between the inventory and the estate is where risk, spend and accountability all go missing at once.
In the register reviewed, owned, approved
Running anyway shadow, citizen-built, developer preference, duplicated across departments
Unapproved customer-facing agents processing sensitive complaints. Developer agents running side projects on company tokens. A marketing agent built by someone who never knew a review process existed. None of it appears in the register, and none of it stops running while the register is being written.
- 09:14:02 Repository read · looks like development
- 09:14:38 Credential used · looks like development
- 09:15:11 Commit pushed · the credential belongs to a competitor
Without behavioral context, all three rows are the same row.
Routine and risky look identical in a log line
A coding agent interacting with source code while holding credentials associated with a competitor reads as normal development activity. An agent pushing corporate secrets to a public repository reads as a routine commit. The difference is context, and context is exactly what a boundary log strips out.
Then there is the record itself: not every agent produces logs, and those that do often permit the session to be rewritten. Without an immutable behavioral record, incident response begins after the damage rather than during it.
Agents cover multiple surfaces. Most tools just watch one.
| Tool | Cloud | Code | Endpoint | Browser |
|---|---|---|---|---|
| Total agent coverage | covered | covered | covered | covered |
| Endpoint sensor | not covered | not covered | covered | not covered |
| MCP gateway | covered | not covered | not covered | not covered |
| Cloud posture tool | covered | not covered | not covered | not covered |
| Identity platform | permissions only, not activity | permissions only, not activity | permissions only, not activity | permissions only, not activity |
| Cost dashboard | not covered | not covered | not covered | not covered |
- covered
- permissions only, not activity
- not covered
Every one of these tools is doing its job correctly. The problem is that the agent is a single system and each tool is looking at one wall of it. Assembling five partial answers by hand, weekly, gives you a history of the estate rather than a working view of it.
Risk is financial too, and it compounds
Agent spend is now a board-level line item, and it is the one line the person accountable for it cannot break down. The total is accurate, but it has no work attached to it.
What you get today
API and model usage $418,900
One aggregate, with no agent, owner or task attached, and no answer to “was any of that worth it?”
What the question actually needs
| Agent | Work it was doing | Owner | Spend |
|---|---|---|---|
| claims-triage | Claims intake | Claims ops | $96,400 |
| market-research | Competitor research | Strategy | $61,200 |
| pr-review-agent | Retrying a failed job, 3,400 times | Platform eng | $73,800 |
| unnamed-dev-agent | Personal side project | Unattributed | $44,100 |
| 3 × content-drafter | The same task, three departments | Marketing, Sales, CS | $143,400 |
Three of these five rows are waste, and none of them are visible until spend is attached to behavior.
Agent waste has four shapes, and none of them are on a dashboard
-
Mode 01
Work nobody asked for
Developer agents running side projects on company tokens. Perfectly reasonable behavior by the person, entirely invisible to the budget holder.
-
Mode 02
Agents that never stopped
Uncontrolled cost scaling from agents that kept running when they should have exited – retry loops, orphaned child sessions, work that finished hours ago.
-
Mode 03
The same work, paid for three times
Duplicate agents doing identical work in different departments, because no department can see the others’ estate.
-
Mode 04
Liberal model choice
Unnecessary frontier models for routine tasks. Employees forgetting to reset after a complex and intensive project.
Two ledgers, one set of agents, no shared line
Security tools do not surface financial risk.
Cost tools do not surface security risk.
The same agent, doing the same thing, appears in both records – and the two records never meet.
Ledger B · The financial record
How many tokens, on which model, in which month, at what total.
- Owned by finance or IT
- Reported to the board as spend
- Silent on what the money bought
Six things that have already happened
Every one of these is drawn from a real deployment. Every one of them looked like ordinary activity right up to the moment somebody understood the context.
-
Critical Supply chain
An agent connected to a malicious skill
It looked legitimate until somebody understood what the agent was doing with it. Skills sit in the agent’s own configuration and generate no outbound traffic for a gateway to intercept.
-
High Data exposure
An unapproved agent processing sensitive complaints
Customer-facing, never reviewed, handling exactly the material a review would have caught. It was not in any inventory.
-
Critical IP leakage
Developer agents pushing corporate secrets to public repos
In the commit history it reads as routine work. The distinction only exists if you can see what the agent was holding at the time.
-
Medium Propagation
A worm propagating via agent hooks
Agent-to-agent spread through the same mechanisms that make agents useful. Nothing in the traditional stack is watching that path.
-
High Insider threat
A coding agent running on credentials linked to a competitor
An employee ran an unsanctioned agent with credentials linked to a rival, exposing customer data and company IP. Endpoint and network controls saw normal development.
-
Medium Financial risk
Developer agents burning tokens on side projects
Spend runs without an owner, attribution or a cap. On the invoice it looks like a usage spike, and it only reads as a security problem once you can see whose credentials paid for it.
A patchwork approach offers no audit trail, no behavioral history, no cost attribution and no fast path to remediation for any of the six.
Business pressure to adopt agents is outpacing the security team’s ability to evaluate them.
Security says wait
The bottleneck. Adoption slows, and the team becomes the thing standing in the way of the strategy the board announced.
The business moves anyway
Unmanaged expansion. The estate grows without review, and the first accurate inventory is produced by an incident.
Both outcomes create risk. The choice is only forced because the evidence needed to make a real decision does not exist yet.
A typical timeline, and the cost of waiting
-
Quarter 1
A handful of pilots, one spreadsheet, everything feels manageable
This is the cheapest moment to establish governance, and the moment it feels least necessary.
-
Quarter 2
Citizen development starts, and the spreadsheet stops being true
Business teams build their own. Developers adopt the tool they prefer over the one they were given. Nobody is doing anything wrong.
-
Quarter 3
Spend jumps and nobody can explain the increase
Finance asks for a breakdown. The honest answer is a total and a shrug. This is usually the quarter the cost conversation opens the security conversation.
-
Quarter 4
Something surfaces, and the audit trail is the thing you needed six months ago
Behavioral history cannot be reconstructed retrospectively. Whatever was not recorded at the time is simply gone.
The window to act is now, while the infrastructure is still forming
-
33%
of enterprise software applications will include agentic AI by 2028
Gartner
-
15%
of day-to-day operations completed autonomously by 2028
Gartner
-
40%
of agentic AI projects will face cancellation due to unmanaged risks
Gartner
-
12
months in which every major platform shipped agent capability
Anthropic, Microsoft, OpenAI, GitHub, AWS
Closing the agent governance gap starts with getting closer to the agent
Every requirement on this page points the same direction: understanding has to come from inside the agent’s own execution path, not from a boundary near it. From there, a security team can approve more agents, with evidence behind every decision.