AI AGENTS & WORKFLOWS
AI agents and workflows: the AI does the groundwork, you sign off
Specialised agents use your price lists, rules and documents. In the workflow editor, you connect agents, knowledge search and tools into reliable processes – and at the steps that matter, a person decides.
Agents with your knowledge
- Your own agents for tasks and teams – from templates or created by you.
- Every agent has instructions for its role, tone and rules, a model, tools, knowledge and conversation starters.
- The tools decide what an agent is allowed to do – read cases or create tasks, for example.
- Sources and context appear right in the chat, and every tool call can be traced.
- Examples: quote assistant, sales coach, onboarding buddy, shift assistant and content studio.
Workflows: single steps become a process
- Visual processes with conditions: read knowledge, have an agent check something, evaluate conditions.
- Start them by hand, on a schedule or with a new note.
- Before anything is created or sent out, the workflow asks you.
- Every run appears later under “Activity” – with each step.
Nothing goes out without your approval
New insights are filed as proposals; your team reviews them before they count as knowledge. You add a human approval to critical steps in workflows – a quote, a content draft or an onboarding plan, for example.
Transcript
Workflows handle recurring work step by step. Each workflow shows its blocks and when it last ran – one click on “Run” starts it. It only asks for what it needs. “Insert example” fills in the fields so you can try it out. You can watch the steps run: reading knowledge, having an agent check the company, evaluating conditions. Before anything is created or sent out, the workflow asks you. You check it – and approve or reject. The result is at the bottom. You’ll find every run later under “Activity” – with each step.
Transcript
Agents are specialists for your tasks. Here are all the agents – from templates or created by you, sorted by areas such as Acquisition or Recruiting. “Configure” opens an agent: its instructions, the model, the tools, the knowledge and the conversation starters. The instructions set the role, tone and rules. The tools decide what the agent is allowed to do – for example read cases or create tasks. Two switches are worth a look: “Research first, then answer” makes it look things up first. And an agent that may work through assigned tasks on its own starts as soon as it’s free.
From the first agent to a process
- Choose a template – for your industry, your role or your project.
- Add your rules, prices and text templates to the handbook.
- Try agents out in the chat; every workflow comes with an example.
- Set up processes with a trigger and approval – and follow them under “Activity”.
Frequently asked questions
How is DENKRAUM different from an AI chat?+
DENKRAUM combines chats with your own agents, knowledge spaces and reusable workflows. Answers can draw on your verified company knowledge. New insights are proposed for review and become part of the knowledge base once approved.
Can the AI approve changes or messages by itself?+
No. New insights are proposals that your team reviews. Critical steps in workflows can require human approval.
Which templates are there?+
Thirteen: for industries such as trades, estate agents and property management, for roles such as sales team, customer service, project management and product owner, and for projects such as winning clients, client projects, recruiting, onboarding, content marketing and events.
More topics
See DENKRAUM AI with your own example
In a 45-minute demo, we show you suitable examples. Then we define a first use case together, along with the knowledge sources needed and the operating concept.