Agentic Studio is the workspace inside SitecoreAI where marketers run AI agents and review what they produce. The agents take the first draft and the first check off your plate. They do not make decisions for you, and what you get out depends on what you set up before your first run. This post covers what the agents do, what you still have to do yourself, and what to have in place before you start.
What Agentic Studio Is
An agent is a piece of AI that carries out a task, such as research or a first draft, and hands the result back to you to review. Agentic Studio is where those agents live inside SitecoreAI. Here is how it works. You pick an agent, give it inputs, and it works inside a space. A space holds the agent's output, the context it used and the chat around it. Each run produces an artifact. You can review it, refine it, compare it with earlier versions, export it, or convert it into a structured SitecoreAI item.
Agents come in two forms. Pre-built agents cover common marketing tasks. An admin chooses which tools, skills, schemas and templates agents can use on the Settings page. Custom agents are ones you build yourself, either as a standard agent for chat-style work or as a workflow agent for a fixed, repeatable process. Flows tie several agents together.
A flow is a sequence of stages, each handled by the right agent, with each agent's output passed to the next. A flow can also include steps where you review or approve before it moves on. For example, a flow might run research, then drafting, then a tone check, with you signing off before anything goes live.
What the Agents Take Off Your Plate
Sitecore's agent documentation lists the pre-built agents in one flat table. I grouped them according to the jobs they do for a marketing team.
| Job | Agent | What it does |
|---|---|---|
| Research | Researcher | Runs structured research on a topic using web and market sources |
| Research | Competitor analyzer | Reviews competitor content and messaging for ways to differentiate |
| Research | Market signals | Turns trends, updates and industry signals into market insights |
| Research | Audience summary | Profiles the target audience: who they are, what they care about, how to reach them |
| Research | AEO/SEO researcher | Assesses visibility, ranking signals and competitiveness in Google and AI-powered search |
| Research | Account data enricher | Fills in missing account details such as size, industry, location and key contacts from an uploaded list |
| Create | Brief generator | Drafts structured briefs for campaigns or content projects |
| Create | Blog writer | Drafts structured blog articles for a defined audience and goal |
| Create | Email writer | Writes personalized emails for newsletters, promotions or onboarding |
| Create | Content generator | Creates ready-to-use content from a single prompt |
| Create | Bulk content generator | Creates content for several audiences and channels at once |
| Create | Snackable content generator | Turns long-form content into short pieces |
| Review and reuse | Persona content auditor | Checks whether content matches the tone and needs of target personas |
| Review and reuse | Translator | Translates and localizes content while keeping brand tone |
| Review and reuse | Summarizer | Condenses long documents or articles into short summaries |
| Review and reuse | Quote extractor | Pulls key quotes or statements out of long content |
| Review and reuse | Structured content extractor | Pulls key information from unstructured text and organizes it for reuse |
Six of the 17 work on content you already have: the persona auditor, translator, summarizer, quote extractor, structured content extractor and snackable content generator. The rest start from a prompt, an audience, a website or an uploaded list.
What the Agents Do Not Do
They do not know your brand until you write it down. Agents write from the brand context you give them: your audiences, messaging, product details, content guidelines and customer evidence. A brand kit adds your tone and writing rules. If that setup is thin, the agents give you generic drafts. Expect to edit heavily for the first few weeks while you tune it. Set it up properly before your first run.
They do not fix a messy content model. Agents work from your content, so if you have unstructured pages and an inconsistent taxonomy, you'll end up with weaker results.
They do not set your strategy. Agents work toward the goals you give them, so have clear audiences and real campaign objectives before you start.
Governance: Where the Review Has to Come From
Agentic Studio gives you controls over who can use agents and over how you trace what they produce. It does not decide where you have to approve the output. That part is up to you.
What the studio gives you
- Roles and licenses. What you can do in Agentic Studio depends on your role and on whether you hold a Builder license.
- Agent visibility. Admins can set an agent as private, public, or shared with specific users.
- Traceability on outputs. Each field in an artifact shows a confidence score, labeled Known, Assumed or Generated, and artifacts show their source references. Sitecore saves versions automatically, and you can compare them.
- Run reports. A post-run report shows runtime, resource usage and artifact counts.
How the review checkpoints work
You set the review checkpoints when you build a flow or a workflow agent. In a flow, you can add review or approval steps. In a workflow agent, the Approval action pauses the workflow until a person reviews and approves. If you build without a checkpoint, nothing stops the run for review. Before you rely on a checkpoint, build a test flow and confirm it holds the artifact until someone approves it.
What we think. Decide who owns the flows before anyone builds one. Creating custom agents requires a Builder license, so whoever holds that license in your tenant can build them, whether that is your team or your implementation partner. A sensible split: your partner builds and tests the first flows with the person who will own them, you write down the sign-off points first, and then your partner hands them over. Drafts for internal use can move freely. Have a person approve anything a customer will see, anything with legal or compliance weight, and every translation before it goes live. Do not count on the tool to enforce the review for you.
How to Start
- Write down your brand and your sign-off rules. Set up your brand context and brand kit, and decide who approves what before anyone runs an agent. An organization admin or owner sets up brand context, so involve one early.
- Decide who builds the flows. Creating custom agents needs a Builder license. Five are included by default, and only admins can assign them. Choose whether that sits with your team or your implementation partner, and who owns the flows afterwards.
- Start with low-risk agents. Begin with the Researcher and the Summarizer. Their output stays internal, so mistakes cost little while you learn how the studio behaves.
- Run the Brief generator early. Every agent and person downstream works better from a structured brief.
- Pilot one flow. Pick one real workload with one named owner. Measure cycle time before and after, then expand.
- Keep checking. Run the Persona content auditor regularly instead of once, so tone drift shows up early.
If you build your own agents, give each step one job and build and test one step at a time. Start from the output you want and work backwards. Keep a copy of a working prompt before you change it, because a change can make results worse.