Key Takeaways

• Dust is a workspace where your team builds custom AI agents on company knowledge. You write instructions, wire in data sources, pick from 20+ models, and run your agent fleet inside Dust's own app.
• Viktor is a single AI employee working in Slack or Microsoft Teams. You @mention him where your team already talks, and he carries the task through to a finished deliverable.
• The deepest difference is where the work happens and what comes back. Dust gives you agents and answers in its workspace; Viktor hands finished work products back in your channels: PDFs, dashboards, drafted emails, code PRs.
• Dust rewards an owner; Viktor rewards a delegator. Dust gets better the more your team invests in designing agents and curating knowledge. Viktor gets better simply by being corrected as he works.
• Both are serious products. Teams who want model choice and a searchable knowledge layer will like Dust. Teams who want output on day one, with approvals guarding real actions, should hire the employee.

Two answers to the same question

A 30-person company decides this is the quarter AI stops being a personal chatbot habit and starts doing team work. The ops lead gets the mandate, opens two tabs, and finds two products that describe themselves in almost opposite ways. Dust calls itself a multiplayer workspace where people and agents collaborate. Viktor calls himself an AI employee working in Slack. Both descriptions are accurate, and the difference between them decides which one fits your team.

This is the honest comparison. Dust is a well-built product with a consistent philosophy, and for some teams it is the right choice. The goal here is to make the trade-off visible, not to pretend one product does everything.

What is Dust?

Dust is a workspace for building and running custom AI agents on your company's knowledge. Your team connects data sources such as Notion, Google Drive, GitHub, and Slack, then creates agents in a builder: instructions describe the job, knowledge sources give the agent context, and capabilities let it act, from web search to drafting emails via Gmail or Outlook. A built-in helper called Sidekick drafts agent instructions for you and suggests improvements as reviewable diffs.

Out of the box, Dust ships global agents like @dust, which is connected to all your workspace data, and @deep-dive, which decomposes complex research into sub-tasks handled by sub-agents. You can also talk directly to raw models: @gpt, @claude, @gemini, @mistral and more, with 20+ frontier models available and the freedom to switch as new ones ship. Agents can run on schedules and event triggers, call other agents, and render interactive outputs called Frames. The primary surface is Dust's own app plus a Chrome extension, with connectors reaching into the rest of your stack.

It is a genuinely thoughtful platform, and the philosophy is consistent: your team designs the agents, so your team controls exactly what each one knows and does.

What is Viktor?

Viktor is an AI employee, singular on purpose. You install him from the Slack App Directory or into Microsoft Teams, connect the tools behind your recurring work, and delegate the way you would to a new hire: in a channel, in a thread, in plain language. He pulls the pipeline numbers from HubSpot and Stripe, drafts the follow-up email, assembles the board pack as a PDF, opens the pull request, and ships the internal dashboard as a live app your team opens from a link.

There is no agent builder step between installing and delegating. Viktor arrives with his capabilities and learns your company the way an employee does: through the work. Corrections and preferences accumulate as persistent skills shared across the team, so the tenth week looks smarter than the first. Sensitive actions, like sending a customer email or pushing code, wait for a human approval in the same thread, which is how review-first delegation stays safe without slowing everything down.

The quick comparison

Where Dust is strong

Naming a competitor's real strengths is the only way a comparison earns trust, so here they are.

Model choice. If your team cares which model runs which job, Dust gives you 20+ options and lets you switch when a better one ships. Viktor manages the model for you, which is simpler but removes that dial.

Deliberate agent design. A custom agent with hand-written instructions and a curated knowledge scope behaves predictably. For narrow, high-volume jobs, such as a support agent answering from a specific documentation set, that precision is valuable.

A semantic layer over company knowledge. Dust invests heavily in making connected knowledge searchable and understood across Notion, Drive, SharePoint and more. Teams whose main pain is "our answers are buried in our docs" will feel that immediately.

Multi-agent orchestration. Agents calling agents, on schedules and triggers, is a native pattern in Dust, and it appeals to teams who enjoy composing systems out of specialized parts.

Where the AI employee model wins

Delegation is the whole interface. The quiet cost of an agent platform is the person who has to own it: writing instructions, curating knowledge scopes, maintaining the fleet as tools and processes change. That is real, ongoing work. With Viktor, the setup phase is a Slack message, and the first useful output usually happens the same afternoon, in the channel where the question came up.

He works where your team already is. Your team will not adopt a tool it has to remember to open. Slack and Teams are already open. A team rollout succeeds or fails on this: when the AI is a colleague in the channel rather than a destination, usage compounds instead of decaying.

Finished work, not just answers. The difference shows up after the answer. An agent grounded in your knowledge tells you which deals stalled. An employee with write access drafts the three follow-up emails, holds them for your approval, updates the CRM after you approve, and files the summary as a PDF for Friday's review.

@Viktor our Q3 planning offsite is in two weeks. Pull last quarter's OKR
scores from Notion, the actuals from our metrics sheet, and build a
one-page PDF per team: goal, result, one-line diagnosis. Post drafts
here for review before you file them in the offsite Drive folder.

One message, five tools, and a deliverable another human can hold. Requests like this rarely fit a pre-designed agent, because nobody designed an agent for them ahead of time. General delegation is the point of hiring.

One employee, whole-company memory. Because Viktor is one identity rather than many, what he learns in #finance about how you define MRR is the same knowledge he uses in the board pack. Skills persist and are shared across the team, so knowledge does not fragment across a fleet.

When to choose which

Choose Dust if your team has a motivated builder, wants to pick models per job, and your central pain is getting answers out of a large, well-maintained knowledge base. It is a strong platform for teams who enjoy operating one.

Choose Viktor if you want the work done rather than the system built: recurring reports that show up on time, research that ends in a document, follow-ups that get drafted and sent after your approval. If the sentence "who on our team will maintain the agents?" produces silence, that silence is your answer.

Frequently Asked Questions

What is the main difference between Viktor and Dust?

Dust is a workspace where your team builds and runs custom AI agents on your company knowledge, inside Dust's own app. Viktor is a single AI employee in Slack or Microsoft Teams who turns delegated requests into finished deliverables. In short: Dust gives you agents and answers in its workspace; Viktor hands completed work back in your channels.

Does Dust work inside Slack like Viktor does?

Dust connects Slack as one of its data-source connectors, and its primary surface is its own app plus a Chrome extension. Viktor is Slack-native and Teams-native: the chat workspace is his only office, and every request, approval, and deliverable happens in your channels and threads.

Can Viktor match a custom-built Dust agent on a narrow task?

For a narrow, repetitive job with a fixed knowledge scope, a hand-designed agent is a fine tool. Viktor covers the same ground through skills: you correct him once, the correction persists, and the recurring task runs on schedule. The difference is that nobody has to maintain the setup as your process changes.

Which is faster to get value from?

Viktor, in most teams. No configuration stage sits between installing and delegating, so the first real output typically lands the same day. Dust's value arrives after your team connects data sources and designs its first agents, which rewards teams willing to invest that setup time.

Can a team run Viktor and Dust side by side?

They solve different shapes of problem, so coexistence is plausible: Dust as an internal answers-and-agents layer over company knowledge, Viktor as the employee who owns recurring deliverables and cross-tool actions in chat. If budget forces a choice, decide based on whether you want to operate a platform or manage an employee.

Does Viktor let me choose which AI model to use?

No. Viktor manages the model for you, running on frontier models from named providers with no-training agreements, and upgrades automatically as better models ship. Dust makes model choice a first-class feature. Managed simplicity versus configurable control is a genuine trade-off, and teams that want the dial should weigh Dust for that reason.