How AgentM uses AI agents.
What AI agents actually do at AgentM
The boring truth: most of what you've heard about AI agencies is marketing. What we actually do, in plain English:
- Planning agents draft strategy docs, research briefs, and architecture.
- Building agents write code, copy, and produce design variants.
- Growth agents draft content, run keyword analysis, and assemble paid creative.
Senior humans set the brief, choose between options, and approve every shipped artifact. The agents are tools, not employees.
What we will not do
- Ship AI-generated content without human review.
- Train models on your data.
- Use consumer-tier AI products for client work — only enterprise endpoints with audit logs.
- Hide which models we used. The disclosure on this page is the floor, not the ceiling.
Frequently asked
Which AI models does AgentM use?
Across studio work and product development we use Anthropic Claude (planning, code review, writing), OpenAI GPT (research, multi-step tasks), Google Gemini (multimodal — image and video reasoning), and Perplexity (live web research). Specific deployment depends on the task.
Is my data sent to AI vendors?
For client engagements: yes, with explicit scope agreed in the SOW. We use enterprise endpoints with no-training data policies (Anthropic Workbench, OpenAI Enterprise, Google Vertex AI). For our own products, the privacy policy on each product page details exactly what is sent and to whom.
Do you train models on client data?
No. All vendor relationships are configured to opt out of training. We do not have an internal training pipeline.
Who reviews AI output before it ships to a client?
Always a senior human. AI agents draft; humans review and ship. The exact ratio depends on the task — code is reviewed line by line; bulk content production uses sample review with full audit trail.
How can I see your subprocessor list?
Our complete subprocessor list is public at /trust/subprocessors/.
Can I have a DPA?
Yes. Our standard DPA is at /trust/dpa/. We can also negotiate custom terms for enterprise engagements.