Using AI Tools
Arketic's assistants don't just answer questions — they can look things up for you and take real actions in the platform. You describe what you need in plain language, and the assistant does the work while showing you exactly what it is doing.
This page explains what the assistant can do, how you can tell what is happening, and when it will stop and ask you first.
The Two Kinds of Tool Activity
Everything the assistant does falls into one of two groups.
| Kind | What it is | What you see | Do you approve it? |
|---|---|---|---|
| Looking things up | Searching documents, checking tasks, reading the org chart, searching the web | A live status chip in the chat | No — nothing is changed |
| Making a change | Creating a task, updating a status, starting a process, sending something to a connected app | An Action Requires Approval card | Yes — it waits for you |
No action that creates or changes data is ever taken without your explicit approval. If you don't press Approve, nothing happens.
Status Chips: Watching the Assistant Work
While the assistant looks something up, a small bar appears in the chat with an icon, a short label and three animated dots. When the step finishes, the dots are replaced by a green check and the label changes to a completed message.
Some examples of what you'll see:
| While it runs | When it finishes |
|---|---|
| Searching documents... | Documents found |
| Reading document... | Document loaded |
| Summarizing document... | Summary ready |
| Searching the web... | Web results ready |
| Loading your tasks... | Tasks loaded |
| Analyzing deadline risks... | Risk analysis complete |
| Loading org chart... | Org chart loaded |
| Creating task... | Task created |
If a step doesn't have its own label, you'll simply see Processing... followed by Done.
More than one chip can appear for a single answer. A question like "Find the travel policy and check whether my expense task is overdue" may run a document search and a task lookup, one after the other.
Turning Search On and Off
The Tools button next to the message box lets you decide which sources the assistant may use for your next message.
| Toggle | What it does |
|---|---|
| Web Search | Lets the assistant search the internet and cite the pages it used |
| Document Search | Lets the assistant look inside your organization's documents and knowledge |
Each row has a sliding switch that turns sky-blue when it's on.
A toggle only appears if the assistant you're using is allowed to use it. If your assistant was set up without web access, you won't see the Web Search row. In a project chat, both are available.
Knowledge and Document Search
Ask a question about your company's material and the assistant searches your documents and knowledge, then answers using what it found.
"What does the travel policy say about international flights?" "Summarize the employee handbook's section on remote work." "Find the document that describes our onboarding checklist."
Sources under the answer
When an answer is based on documents or web pages, a collapsible Sources pill appears underneath it with a number badge. Click it to expand the list.
| Source type | What the row shows |
|---|---|
| Document | The document name and a relevance score as a percentage |
| Web page | A globe icon and the page title — click to open it in a new tab |
The relevance badge is colour-coded so you can judge at a glance how closely a document matched your question:
| Badge colour | Relevance |
|---|---|
| Green | 80% and above — a strong match |
| Yellow | 60% to 79% — a partial match |
| Grey | Below 60% — a weak match |
Open the sources before you act on an answer. If the relevance scores are low or the documents listed aren't the ones you expected, rephrase your question with the exact document or policy name.
When an answer is well supported by its sources, a small green Source-backed response badge appears next to the message timestamp when you hover over it.
Web Search
With Web Search on, the assistant can look something up on the internet and show you where the information came from.
While the search runs, a sky-blue block appears with Searching the web... and animated dots. When it's done, the block header shows Web search: "your query" and the body contains:
- A short summary of what was found
- A Sources list — each entry is the page title (with a snippet where available) and opens in a new tab when clicked
The block is collapsible, so you can fold it away once you've read it.
Links from the internet are shown only when they are safe, standard web addresses. Anything else is listed but not clickable. Web results come from outside your organization — check the linked page before relying on it for anything important.
Actions That Need Your Approval
When the assistant is about to create, change or trigger something, it stops and shows an Action Requires Approval card in the chat instead of just doing it.
What's on the card
| Part | What it tells you |
|---|---|
| Action Requires Approval (with a shield icon) | The assistant is waiting for you |
| A one-line description | What will happen — e.g. "This will create a new task in your project." |
| The action name | Which action is about to run |
| A list of details | Every value that will be used, as label-and-value rows |
| Approve / Cancel | Your decision |
The detail rows change depending on the action. For a new task you'll typically see Title, Description, Priority, Assigned to, Deadline and Department. For a status change you'll see the Task and its New Status. For a time entry you'll see the Task, the Duration and a Description.
Descriptions you may see
| The assistant wants to | The card says |
|---|---|
| Create a task | This will create a new task in your project. |
| Change a task's status | This will change the status of a task. |
| Add a comment to a task | This will add a comment to a task. |
| Log time on a task | This will log a time entry for a task. |
| Update a position in the org chart | This will modify a position in the org chart. |
| Update an employee's profile | This will update an employee's profile data. |
| Save something to a project's memory | This will save information to the project's memory. |
| Create a service request | This will create a new service request. |
| Start a blueprint | This will start a new blueprint workflow instance. |
| Run a workflow | This will run one of your workflows on your behalf. |
| Act in a connected app | This will run an action on one of your connected third-party services on your behalf. |
If an action has no description of its own, the card falls back to "This action will modify data. Please review before approving."
Approving or cancelling
- Approve (green, with a check icon) — the action runs. The card switches to a spinner and Executing... until it finishes.
- Cancel (white, with a cancel icon) — nothing runs. The assistant is told you declined and can suggest something else.
Read the detail rows before approving — this is the moment to catch a wrong assignee, a wrong date or a typo in a title. If something is off, press Cancel and rephrase your request rather than fixing it afterwards.
Working in Your Connected Apps
The assistant can also work in the third-party services your organization has connected — for example Slack, Notion, Confluence and Google. It finds which actions are available, checks what information each one needs, sees which services you've connected, and then runs the action.
Anything that sends or changes data in a connected service always shows an approval card first, listing the exact Service, the Action, a What it does line, and every value that will be sent.
This is switched off unless your admin turns it on for the assistant you're using. See Connected Apps for the full picture.
What You Can Ask For
| Category | Examples |
|---|---|
| Tasks | "Create a high-priority task for Sarah due Friday", "Move the onboarding task to In Review", "Add a comment to the budget task", "Log 90 minutes on the audit task", "What's on my plate this week?" |
| Documents & knowledge | "Search for the vacation policy", "Summarize the handbook", "Which documents mention the new supplier?" |
| Organization & HR | "Who manages IT?", "Show me the reporting chain for the design team", "Find someone who knows data analysis", "Which positions are open?" |
| Service desk | "Submit an IT request for a new laptop", "What requests are still open for my department?" |
| Processes | "Start the onboarding process for John" (Blueprints), "Run the weekly report workflow" (Workflows) |
| Analytics | "Are any tasks at risk of missing their deadline?", "How are we doing on our service level targets?", "Prepare a status report for this project" |
| Web | "Search the web for the latest regulations on..." |
| Connected apps | "What can I do with Slack?" |
Which of these are actually available depends on how your admin configured the assistant you're talking to. Different assistants can have different abilities.
Who Can Do What
Some actions are limited by your role in the organization.
| Level | Examples of what it unlocks |
|---|---|
| Everyone | Looking things up — documents, tasks, the org chart, the web |
| Manager and above | Starting a blueprint, updating a position, updating an employee's profile, project status and risk reports, trend analysis |
| Admin or executive | Deleting a task, creating or changing a department, the company-wide overview |
If the assistant says it can't do something, it's usually one of two reasons: the action is restricted to a higher role, or it was turned off for that particular assistant. Ask your admin if you think you should have access.
Asking About an Employee
If you manage people, you can put a team member in context before you ask a question.
- Open the + menu next to the message box.
- Choose Select Employee.
- In the Select Employee window, use Search by name or email... to find the person, then click them.
- A chip with their name appears above the message box.
Now ask your question:
"What's on their plate this week?" · "Summarize their open tasks"
Your message carries an About {name} chip while the context is active. Remove it with the × on the chip.
Select Employee appears in the + menu only if your role allows it — team members don't see it.
Generated Documents
Ask the assistant to produce a real file — a report, a spreadsheet, a presentation — and it delivers a download card in the chat with the file name and format (Word, Excel, PDF, PowerPoint). Use Preview to open it in the side panel, or Download to save it.
See Artifacts & Documents for the full details.
Tips
- Be specific with names — "Create a task for John Smith" works better than "for John".
- Put the details in one message — priority, due date, assignee and department together, so the approval card is right the first time.
- Say where the answer should come from — "using our documents" or "search the web for" removes the guesswork.
- Read the approval card before confirming — it's a full preview of what's about to happen.
- Combine requests — "Show my overdue tasks and prepare a summary report" is a single ask.
- Follow the sources — for anything you'll act on, expand Sources and open the original.
Troubleshooting
| Problem | What to do |
|---|---|
| The assistant says it can't do that | The action is restricted to a higher role, or it's turned off for this assistant |
| No approval card appeared for something that changed data | Nothing that creates or changes data runs without a card — if you didn't see one, the assistant only looked something up |
| Wrong details on the approval card | Press Cancel and rephrase with clearer instructions |
| The action failed after you approved it | Read the message the assistant returns — it usually explains what went wrong |
| No Sources panel under the answer | The answer wasn't based on documents or web pages. Turn on Document Search or Web Search and ask again |
| No Web Search or Document Search toggle | That assistant wasn't set up with it. Try a different assistant, or ask your admin |
| A message says document features aren't available | Some document features can be unavailable depending on your organization's data settings — check with your admin |
Related Pages
- Starting a Chat — the message box, the toolbar and how answers appear
- Connected Apps — working in Slack, Notion, Confluence and other connected services
- Thinking Modes — how much reasoning the assistant puts into an answer
- Attachments — adding your own files to a conversation
- Artifacts & Documents — generated files and the side panel
- Tips & Best Practices — getting better answers