Categorized answers about ajan.io — agents, workflows, models, channels, and security.
25 questions
ajan.io is an enterprise AI platform where companies can create and manage their own AI agents. These agents can use your knowledge sources, answer employee or customer questions, work with the systems you use, and run business processes within the permissions you grant.
The core idea is to move AI beyond a chat-only tool by combining three capabilities: access knowledge, take action, and manage processes. You can build AI workers that do more than answer — they help finish real work.
An AI agent is more than a model that answers a question. It can reach knowledge, use tools, and take the next steps needed to complete the task it was given.
In ajan.io you can define what an agent works on, which company knowledge it can access, which systems and tools it may use, which model it runs, and which actions require human approval.
That lets you create agents with different jobs and permissions for sales, HR, operations, finance, and more.
General-purpose apps like ChatGPT are built to answer a wide range of questions. ajan.io is an enterprise platform designed to run AI with your company knowledge, users, systems, and processes.
An agent on ajan.io can find information in company documents, show which sources an answer is based on, pull data from company systems, act with defined tools, and start workflows when needed.
In short: general-purpose AI mostly answers; ajan.io agents can find knowledge, answer, act, and run processes.
No. Chat is only one way to interact with agents.
ajan.io goes beyond a chat box: it runs AI with your knowledge and operations stack. Agents can use knowledge sources, talk to enterprise systems, act through tools, and run multi-step processes.
The same agent capabilities can be used via web chat, embedded chat, workplace channels, and API.
Yes. Each agent can be configured for your company’s needs.
You can define a role and job; choose knowledge bases, AI models, integrations, tools, budgets, and user access. Language, tone, and behavior instructions can also be set per agent.
For example, sales can have an agent that knows customers and proposals, while HR has a different agent focused on policies and people processes.
Yes. You can configure specialized agents to work together.
A user may talk to one primary agent that routes a request to a sales, finance, or HR specialist agent.
That keeps the setup manageable instead of loading one agent with every company responsibility.
Yes. Your documents and other supported sources can become knowledge bases agents can use.
PDFs, office files, text, web sources, and supported connectors can be processed for agents.
People get what they need from the agent instead of searching folders and apps.
Yes. Answers grounded in company knowledge can show the sources and relevant document sections.
That matters for policies, procedures, product docs, and other cases that need verifiable information.
Users see not only an AI answer, but which company knowledge it rests on.
ajan.io lets you control which knowledge sources agents use and what happens when company knowledge is missing.
Finding suitable content, scoring results, re-ranking, and linking answers to sources work together.
You can configure agents to answer only from approved company sources. When no reliable source exists, the system can say so clearly or fall back to general model knowledge — depending on the scenario.
Yes. A core goal is to run AI with the software you already use.
APIs, HTTP services, supported connectors, and tool integrations can open CRM, ERP, operations systems, or custom apps to agents.
Instead of replacing systems, AI becomes a layer on top that can access information and act for the user.
Yes. With defined tools and permissions, agents can perform real operations in company systems.
An agent can query data, create a record, update an existing record, or start an operation.
You decide which agent may use which tool and which actions need human approval — so AI only gets the authority it needs.
Yes. Controlled operation is part of ajan.io’s enterprise approach.
You define which knowledge, tools, and systems an agent can access, and which agents users may use.
Tool policies and human approval mean you do not have to give every agent or action the same autonomy.
Yes. Agents can prepare work up to a point and wait for an authorized person before a critical step.
For example, AI can draft a proposal but wait for a sales manager before sending it to the customer. Purchasing or operations can also pause for a human decision.
You keep company controls while using AI’s speed and automation.
Workflows manage multi-step work with AI and automation — not just a single action.
A process can combine gathering information, condition checks, AI evaluation, actions in systems, waiting for approval, pausing for a period, or starting another process.
Agents become part of an end-to-end defined process instead of only firing one action.
Your company sets the autonomy level for each scenario.
An agent can be an assistant that only shares information, one that offers options and leaves the decision to the user, or one that acts within permissions and follows a workflow end to end.
ajan.io aims for controlled autonomy within policy — not uncontrolled maximum autonomy.
ajan.io is not locked to one model or provider.
Supported provider models can be used, and suitable options can be set per agent. Capabilities such as tool use, vision, reasoning, streaming, or context size may differ.
You can pick for quality, cost, and use case instead of one model for everything.
Yes. On supported providers you can use your own API key (BYOK — Bring Your Own Key).
You may use platform-managed access or your own provider account. Because model choice is per agent, different keys and access methods can coexist in one organization.
That helps companies with existing provider agreements or that want to manage model spend on their own account.
Yes. Different jobs can use different models.
Simple, high-volume tasks may use a more economical model; complex analysis or reasoning may use a stronger one.
You can allow multiple models per agent and define fallback order for a balance of quality, cost, and continuity.
Agents are not limited to the ajan.io chat screen.
Besides embeddable web chat, they can run on supported channels such as WhatsApp, Telegram, Slack, and Microsoft Teams.
You can bring agents to the channels employees or customers already use.
Yes. With API and partner integration paths, you can embed agents in your products and digital experiences.
A SaaS app, customer portal, mobile app, or internal system can talk to agents on behalf of its users.
Users get platform AI capabilities inside your product without switching to the ajan.io UI.
Yes. AI usage can be tracked by user, agent, model, and similar dimensions.
Model calls, token use, and AI Credit consumption can be monitored. Budgets or limits can be set for agents and users.
You can see what consumes resources and manage AI spend with more control.
ajan.io treats enterprise AI as an operations area to manage — not only model access.
Role and permission management, agent and knowledge access, action approvals, tool policies, guardrails, and audit records can work together.
Critical admin actions and AI operations can be made inspectable so you can manage who works with which agent under which authority.
ajan.io is designed for scales from small teams to larger organizations.
A small company can start with one agent and a few knowledge sources. Larger orgs can manage department agents, user groups, knowledge bases, models, integrations, and workflows together.
Start with a focused scenario and expand as needs grow.
You can start by creating a free account.
Add sample documents or sources to build a knowledge base, then define your first agent that uses it. Add users, models, integrations, and workflows as needed.
The best start is a clear, measurable use case. After the first agent works well, expand the same pattern to other departments and processes.
Write your question and we'll route it to the right team. Setup and API details are in the docs.