Agentic Studio is the workspace inside SitecoreAI where marketing teams create, run and govern AI agents. It ships with prebuilt agents for common jobs (Sitecore's documentation lists agents such as Blog Writer, Content Generator, Email Writer, Researcher and Competitor Analyzer), and it lets you build custom agents for your own workflows. This guide explains the building blocks and how to design a first custom agent that is useful and safe to run.
The building blocks
| Component | What it is (per Sitecore docs) | Why it matters |
|---|---|---|
| Standard agent | A conversational custom agent that decides how to perform tasks from its instructions and the tools, context and skills you select | Flexible: good for research, drafting and analysis |
| Workflow agent | A custom agent for structured, repeatable, multi-step work, with a predefined sequence, checkpoints and approvals | Predictable: good for production processes |
| Skill | A capability package with instructions, reference files, schemas and scripts | Reusable know-how you can share across agents |
| Brand kit | A structured resource built from your uploaded brand documents, attachable as default context | Keeps tone, terms and claims on-brand |
| Agent API | Gives an agent access to Sitecore features: sites, pages, content, components, assets, personalization, jobs, briefs and brand kits | Lets agents act in the CMS, not just write text |
| Space | A collaboration workspace for chats, items, context, agents, runs and artifacts | Where a team works with agents on one initiative |
Standard or workflow agent?
Sitecore's guidance is that the choice depends on how complex the process is and how much human involvement you want. In practice:
- Choose a standard agent when the task varies each time: researching a topic, drafting variants, summarising feedback.
- Choose a workflow agent when the steps are the same every time and someone must sign off: localising a page into five markets, preparing a campaign landing page, migrating legacy content into new components.
Design your first custom agent in six steps
- Pick one repeatable job with a clear owner. Example: "create a localised variant of a product page for each market, ready for regional review".
- Write the instructions as a brief. Goal, inputs, output format, what the agent must never do (invent claims, change prices, publish).
- Attach the brand kit so tone, terminology and approved claims come from your documents, not from the model's guesses.
- Give it only the tools it needs. If it drafts content, it does not need publish rights. Least privilege applies to agents too.
- Add checkpoints. In a workflow agent, put a human approval before anything reaches a live page.
- Measure it. Track time per item before and after, and the share of drafts approved without major edits.
Governance checklist
- Who can create, edit and run custom agents (Sitecore controls this by role and Builder licence)?
- Which agents may write to which sites and languages?
- Where are approvals required, and who approves?
- How are agent changes versioned and tested before teams use them?
- What data may agents use? Sitecore states SitecoreAI does not share customer data with external providers or use it to train external AI models; check your contract for your own requirements.
Good first agents for enterprise teams
- Regional variant drafter (workflow): builds market variants of a page, flags claims that need local legal review.
- Content migration assistant (workflow): maps legacy rich text into new components during an XP to SitecoreAI migration.
- SEO and AI-visibility reviewer (standard): checks titles, descriptions, headings and FAQ coverage before publishing.
- Campaign brief to page (workflow): turns an approved brief into a draft landing page and personalization variants.
How Kogifi helps
We run SitecoreAI agent adoption programmes: we pick the workflows worth automating, set up brand kits, skills and approvals, and measure time saved. Kogifi is a two-time Sitecore Digital Impact Awards winner (2025). See our Sitecore and SitecoreAI services or read what SitecoreAI is.
FAQs
What is the difference between a standard agent and a workflow agent in SitecoreAI?
A standard agent is conversational and decides how to complete a task from its instructions, tools, context and skills. A workflow agent follows a predefined multi-step sequence with checkpoints and approvals.
What is a brand kit in Agentic Studio?
A structured resource built from your uploaded brand documents. You attach it as default context so agents follow your brand voice and terminology.
Can SitecoreAI agents change content in Sitecore?
Yes, through the Agent API, which gives agents access to sites, pages, content, components, assets, personalization and more. Use permissions and approval checkpoints to control what reaches live pages.
Who can build custom agents?
Creating and editing custom agents depends on the user's role and Builder licence in SitecoreAI.
Sources: Sitecore documentation, "Understanding agents" and "The Agentic Studio toolkit" (doc.sitecore.com), checked October 2026.
