Skip to content
batoon.ai
DE EN Deutsch: /abgrenzung/

Differentiation

Two kinds of vendor are building governance. Both have a stake.

Above you sit the vendors selling a model. Below you sit the ones selling a channel. Both have built serious controls. And both help decide something they earn money on.

Research as of August 2026. We re-check quarterly.

Where things stand

What the others can do by now.

Model vendors and MarTech suites have built a lot of governance since 2025. Four things are standard today. We list them because a comparison that misrepresents the field collapses at the first follow-up question.

A check before the action
Large suites verify authorisation before an agent calls a tool. On the model side, central enforcement points evaluate every tool call against rules.
Human approvals
Per-tool switches, multi-level approval modes and intervention points in workflows. No serious vendor lets agents write unasked today.
Logs
Audit trails, session traces and compliance APIs. What an agent did can be read back everywhere.
Reach beyond their own boundary
Through the open MCP standard, suite agents now reach third-party systems. Platforms bring third-party agents under their own governance.

Looking up

The chat harness.

A model vendor with enterprise features: roles, permissions down to tool level, logs and spend limits. For a capable individual user that is strong.

  • Permissions down to tool level
  • Logs across all actions
  • Approvals in several gradations
  • Spend limits per user

Where the line runs

Whoever acts approves. There is no instance where somebody else has to agree, and no threshold by amount or risk. That is the design: the harness models a user. A marketing team is an organisation. The plan goes to the campaign lead, the budget overrun to the CMO, the EU segment to legal. And the spend limit caps what the model costs, not what it spends.

Looking down

The suite agent.

Your CRM, your email platform, your content suite: each of these now has agents, often with carefully built governance at the platform layer.

  • Authorisation before the tool call
  • Consent checks at the platform layer
  • Audit trail and observability
  • Third-party systems via open standards

Where the line runs

Look at which third-party systems are actually connected. They are productivity and sales tools. The competitor's delivery channel is not among them and will not be. On top of that, the governance belongs to the platform. One vendor writes in its own documentation that every application and every agent inherits the platform's rules rather than enforcing its own. That is meant as a strength and it is one. It also means that leaving the platform means losing it.

What remains

Four differences no release closes.

01 Position

Everywhere else, the orchestrator has a stake

Whoever enforces the rules also sells the systems being decided between. And benefits from how a budget conflict between two channels turns out. That cannot be fixed. It is the business position.

02 Reach

Everyone opens up, nobody to the competitor

Connecting third-party tools is normal by now. Making a competitor's delivery channel a first-class governed path inside your own product is not. Coordinating across CRM, email, ads, CMS and the call centre needs someone who sells none of those channels.

03 Measurement

A control group only holds if it holds everywhere

A holdout in the email platform excludes those held back from the email. The call centre rings them anyway. Nobody who cannot see the other channels can make that promise.

04 Budget

Everyone caps what the AI costs. Nobody caps what it spends.

Token allowances and credits cap AI consumption. They do not cap the €25k of media budget distributed across Meta, Google, email and the call centre. The CMO answers for that.

Where batoon stands

We claim no lead on the mechanics.

The vendors above and below us are further along on governance mechanics than comparison pages usually admit. Our difference is the position, not the feature count. So here is where batoon itself stands.

Policy gate before every tool call
built
Approvals by role, risk and amount
built
Audit trail with checked rationale
built
Cross-channel control groups
built, so far shown in the simulated world
Model neutrality across six providers
configurable, proof on a second provider still open

Proof on a real stack is still outstanding. That is what we are looking for design partners for.

The full status