⏱️ 5 min read
Mark AI is built on six core concepts. Together, they form the Content Operating System that powers everything you do on the platform. This article gives you a working understanding of each one: what it is, how it behaves, and when you'll actually use it.
📌 What you'll learn
How agents, workflows, and templates stack into a complete content engine
The role of workspaces in organizing brands, teams, and markets
How the knowledge base and glossary ground every piece of content in your truth
Best practices to get the most value out of each concept
1. Agents
The building block of Mark AI.
An agent is a purpose-built AI worker dedicated to a single content mission. An SEO writer agent produces optimized articles. A LinkedIn editorial agent writes posts in your voice. A brand translator agent localizes content across markets. Each agent is configured once, then runs consistently every time you call it.
The AI Studio of your workspace lists every specialized AI worker available to your team.
What makes a Mark AI agent different from a prompt in ChatGPT or Claude is the three layers of context it operates within:
Your brand voice. Tone, style, editorial line, formatting rules.
Your knowledge base. Guidelines, product sheets, case studies, customer interviews.
Your editorial guardrails. Preferred terminology, banned terms, compliance requirements, governance rules.
You design the mission once. The agent executes it consistently, at scale, across every team and every market.
The agent edit page for a Blog Post Editorial agent, showing the agent brief, desired content examples, voice/tone fields.
Example: a LinkedIn Editorial agent
A B2B SaaS company configures a LinkedIn agent with:
Voice: insightful, never promotional, first-person
Length: 1,200 to 1,500 characters
Structure: hook, story, takeaway, soft CTA
Banned terms: "revolutionary", "game-changer", "synergy"
Knowledge base: case studies, founder interviews, product roadmap
Every time a marketer triggers this agent, the output respects all of these rules automatically. The marketer doesn't have to re-explain the brand each time.
💡 Best practice
Start with 3 to 5 agents that cover your highest-volume content needs (typically: SEO articles, LinkedIn posts, emails, newsletters, translations). Add more as you identify recurring use cases.
2. Workflows
Agents chained into a multi-step production line.
A single agent produces a single output. A workflow chains several agents together to execute a multi-step content process, where the output of one agent becomes the input of the next.
A typical content workflow might look like this:
Research agent gathers insights from the web and your knowledge base → Copywriter agent drafts the article → SEO agent optimizes for target keywords → Translator agent localizes for 3 markets → Repurposing/LinkedIn agent turns the article into 5 LinkedIn posts.
Workflows turn isolated agents into a content factory. Instead of triggering 10 agents manually, you trigger one workflow and get a complete deliverable.
In practice, most teams never build a workflow from scratch. They launch their workflows from a template (see next section), which is a pre-built blueprint kept in the Templates library inside the Workflows module.
The workflow builder lets you chain multiple agents into a single, repeatable content process.
Example: an event content workflow
For an industry event, you might build a workflow that:
Generates a teaser email 2 weeks before
Drafts a launch LinkedIn post 1 week before
Writes a recap blog post the day after
Repurposes the blog into a newsletter section
Translates everything into your 3 priority markets
One trigger, one full event campaign.
💡 Best practice
Map your existing content processes on paper before building a workflow. The clearer your process, the more value the workflow will deliver. Don't try to automate chaos.
3. Templates
Pre-built workflows, saved in your library and ready to relaunch.
A template is a saved blueprint of a workflow, stored in your Templates library inside the Workflows module. While a workflow is the actual execution of a multi-step content process, a template is the reusable plan that defines that process — already built, already tested, ready to be instantiated whenever your team needs to run it again.
The relationship is straightforward:
A template is the recipe in your cookbook
A workflow is the meal you cook from that recipe
Every time you launch a template, you create a new workflow run
The Templates library gives your team one-click access to every pre-built workflow
Templates exist for complete business outcomes, not just single pieces of content:
Campaign launch template: from teaser to recap, across email, social, and blog
Event activation template: invitations, reminders, live posts, follow-ups
Product release template: announcement, technical docs, sales enablement, customer email
Webinar repurpose template: turn one webinar into 10+ content pieces
When a new campaign comes up, your team picks the right template, plugs in the campaign-specific inputs (topic, date, target audience), and launches a workflow run that delivers a complete content package.
Example: the Campaign Launch template at work
Your marketing team is launching a new feature next month. They open the Campaign Launch template, fill in:
Feature name and value proposition
Target launch date
Priority markets
In one run, the resulting workflow generates: an announcement blog post, 3 LinkedIn posts (founder, company, customer voice), a launch email, a follow-up sequence (3 emails), a sales enablement one-pager, and translations for each market.
💡 Best practice
Templates become exponentially more valuable as your team grows. The first hour of building a template saves dozens of hours over the next year. Prioritize templates for recurring campaigns (monthly newsletter, quarterly product update, weekly social cadence).
4. Workspaces
Dedicated environments to keep your work organized.
A workspace is a container that holds its own agents, templates, knowledge base, glossary, and users. Most organizations use multiple workspaces to cleanly separate work that should not mix.
Common reasons to create a new workspace:
One workspace per brand. If you manage multiple brands (e.g. a parent company with sub-brands), each gets its own workspace with its own voice and rules.
One workspace per market or language. If your French and Spanish teams operate independently with localized agents and glossaries.
One workspace per team. If your editorial team and your sales enablement team need different agents and content rules.
One workspace per client. Common setup for agencies serving multiple clients on Mark AI.
Each workspace is fully isolated: agents in one workspace cannot access knowledge or glossary from another, ensuring there's no cross-contamination.
Example: a multi-brand setup
A retail group with 4 distinct brands creates 4 workspaces. Each brand gets its own:
Voice and editorial guidelines
Product knowledge base
Glossary of brand-specific terms
Set of agents and templates
Content produced in the "Brand A" workspace will never accidentally use vocabulary or examples from "Brand B".
💡 Best practice
Don't over-segment. Start with one workspace per brand, and only add more if you have clear governance reasons (different legal entities, fully independent teams, client confidentiality). More workspaces means more maintenance.
5. Knowledge base
The reference library that grounds your agents in your truth.
The knowledge base is the space where you upload everything your agents should know about your business. When an agent generates content, it retrieves the relevant pieces from the knowledge base to stay accurate, on-brand, and free of hallucinations.
Your knowledge base centralizes every document your agents can rely on, from product sheets to case studies.
Documents to upload into your knowledge base:
Editorial guidelines (tone of voice, style guide)
Product sheets (features, specs, pricing, positioning)
Case studies (real customer stories, ROI numbers, quotes)
White papers and research reports
Customer interviews and testimonials
Sales decks and pitch materials
Past content that performed well (to reinforce style and structure)
The richer your knowledge base, the more accurate your generated content becomes. An agent without a knowledge base will produce generic content. An agent with a deep knowledge base will produce content that sounds like it was written by your most senior team member.
How the knowledge base works at generation time
When you trigger an agent, Mark AI:
Analyzes the request and identifies relevant documents in the knowledge base
Retrieves the most relevant passages
Passes them to the agent as grounding context
Generates content that's anchored in your real information, not in the model's training data
This is what guarantees no hallucinations on brand-critical facts (product specs, pricing, customer names, case study numbers).
💡 Best practice
Refresh your knowledge base every quarter. Outdated product sheets or old case studies will produce outdated content. Set a recurring review in your team's calendar.
⚠️ Important
The knowledge base is workspace-scoped. If you operate multiple brands, each workspace needs its own knowledge base. Don't try to share one across brands.
6. Glossary
Your brand vocabulary, enforced on every output.
While the knowledge base provides what your agents know, the glossary controls how they speak. It's the dictionary of terms, expressions, and phrasing rules that apply across every piece of content your team produces.
The glossary defines the vocabulary every agent must respect, across all your markets.
A glossary typically contains four types of entries:
Preferred terms. "Customer" not "user". "Platform" not "tool". "Investment" not "spend".
Banned terms. Words you never want to appear in your content (overused buzzwords, competitor names, deprecated product names).
Acronyms. What does ARR mean in your context? POC? MRR? The glossary disambiguates them.
Translation equivalents. For each preferred term, the exact equivalent across each of your target languages.
Agents check the glossary on every generation. If banned terms appear in a draft, they're flagged or replaced. If a preferred term has a translation, it's used consistently across markets.
Example: enforcing voice consistency at scale
A financial services company has a glossary that:
Bans "cheap" and replaces with "affordable" or "accessible"
Bans "guaranteed returns" (regulatory issue) so the term never appears in any output
Translates "wealth management" into "gestion de patrimoine" (FR), "gestione patrimoniale" (IT)
Every email, blog post, and LinkedIn update produced on Mark AI respects these rules automatically. No more team-by-team inconsistencies. No more accidental compliance breaches.
💡 Best practice
Start your glossary with your top 20 entries. Identify the terms your team argues about most often, or the ones you've had to correct repeatedly in past content. Expand from there.
How the 6 concepts work together
Think of the stack from the bottom up:
The knowledge base and glossary form the foundation — your single source of truth and your vocabulary.
Agents sit on top of this foundation, each specialized for a content mission.
Workflows chain agents into a production line.
Templates save those workflows for one-click reuse.
The workspace wraps everything in a clean container, isolated from other brands or teams.
Each layer reinforces the next. A great agent without a knowledge base is just a generic LLM. A great workflow without a glossary will drift in tone. A great template without a workspace will leak across brands.
This is why Mark AI is called a Content Operating System: it's the full stack, not a single tool.
➡️ Next step
Now that you understand the building blocks, it's time to see the platform in action.
Continue with Platform tour to discover the interface and where each of these concepts lives in the UI.
If you're ready to produce content, jump to Your first generation in 3 minutes.