# The Best AI Workflow Automation Tools in 2026, Compared > Published 2026-10-28 > Source: https://viktor.com/blog/ai-workflow-automation-tools This guide compares the AI workflow automation tools that operations leads and founders shortlist most often in 2026, judged on integrations, AI capability, pricing model, skill level and self-hosting, using vendor pricing and trust pages checked in October 2026. The best AI workflow automation tools in 2026 are Viktor, an [AI employee](https://viktor.com/ai-employee) in Slack and Microsoft Teams, for teams that want to hand off the work itself; Zapier for the easiest no-code workflows; Make for visual multi-step scenarios; n8n for self-hosting and developer control; Lindy and Gumloop for agent-first builds; and UiPath for legacy desktop and document-heavy processes. Microsoft Power Automate and Workato round out the list for Microsoft 365 shops and enterprise IT. ## How we evaluated Most lists rank workflow automation tools by the size of their app library. That matters, but it does not decide whether a tool still saves your team time six months from now. We scored every tool on five criteria: 1. **Integrations.** How many apps it connects to, how deep those connections go (read and write, or just triggers), and what happens when an app has no API. 2. **AI capability.** Whether AI is a single step inside a fixed workflow, an agent that picks its own steps, or something that does the whole job end to end. Human-in-the-loop approvals count here too. 3. **Pricing model.** Tasks, credits, executions, seats or contracts. The model matters more than the entry price, because it decides what the bill looks like at ten times your current volume. 4. **Skill level required.** Who on your team can build, change and fix a workflow without help: anyone, an ops person, or only a developer. 5. **Self-hosting and data control.** Whether you can run it on your own servers, where data is stored, and whether customer data is used to train models. Prices come from each vendor’s official pricing page, checked in October 2026. Vendors change plans often, so confirm before you buy. ## Best AI automation tools at a glance ```json:object { "_key": "k40", "_type": "table", "rows": [ { "_key": "k29", "_type": "tableRow", "cells": [ "Tool", "Best for", "AI capability", "Pricing model (checked Oct 2026)", "Self-hosting" ] }, { "_key": "k30", "_type": "tableRow", "cells": [ "1. Viktor", "Delegating recurring cross-tool work from Slack or Teams", "AI employee: takes a goal, works out the steps, asks for approval on sensitive actions", "Usage-based credits shared across the workspace, no per-seat fee. Free with up to $100 in credits, Team $100/mo for 40,000 credits", "No (cloud; EU data residency on Enterprise)" ] }, { "_key": "k31", "_type": "tableRow", "cells": [ "2. Zapier", "Non-technical users building simple app-to-app automations", "AI steps, Copilot, Zapier Agents, Zapier MCP", "Tasks. Free (100 tasks/mo), Professional from $19.99/mo, Team from $69/mo; Agents billed separately", "No" ] }, { "_key": "k32", "_type": "tableRow", "cells": [ "3. Make", "Visual multi-step scenarios with branching", "AI modules and AI agents inside scenarios", "Credits (most actions use one credit). Free plan, then Core, Pro and Teams", "No" ] }, { "_key": "k33", "_type": "tableRow", "cells": [ "4. n8n", "Technical teams that want self-hosting and code", "AI agent nodes, LLM nodes, MCP", "Executions. Cloud Starter €20/mo billed annually for 2,500 executions; self-hosted Community Edition free", "Yes" ] }, { "_key": "k34", "_type": "tableRow", "cells": [ "5. Lindy", "Agent-first workflows for sales and support", "AI agents with triggers, actions and approvals", "Credits per active user, from $29.99/user/mo for 3,000 credits", "No" ] }, { "_key": "k35", "_type": "tableRow", "cells": [ "6. Gumloop", "Agent-first data and marketing pipelines", "AI nodes and agents on a canvas", "Credits. Pro from $37/mo", "No" ] }, { "_key": "k36", "_type": "tableRow", "cells": [ "7. UiPath", "Legacy, desktop and document-heavy enterprise processes", "RPA robots plus agentic automation and document AI", "Free tier, Basic from $25/mo, Standard and Enterprise via sales", "Yes (on-premises or cloud)" ] }, { "_key": "k37", "_type": "tableRow", "cells": [ "8. Microsoft Power Automate", "Companies that run on Microsoft 365", "AI Builder and Copilot inside cloud and desktop flows", "Premium $15/user/mo billed annually", "Desktop flows run on your machines; cloud flows hosted by Microsoft" ] }, { "_key": "k38", "_type": "tableRow", "cells": [ "9. Workato", "Enterprise integration owned by IT", "AI agents and orchestration across governed recipes", "Contract pricing, no public list", "No" ] }, { "_key": "k39", "_type": "tableRow", "cells": [ "Relay.app", "Shut down in September 2026", "n/a", "n/a", "n/a" ] } ] } ``` Relay.app still appears on many older lists, but it shut down in September 2026 and deleted all accounts and workflows. Lindy and Gumloop are the closest agent-first replacements, Make the closest classic builder. ## Our top picks Each pick lists who it is best for, its AI capability, pricing and hosting, plus the one thing to watch out for. ### Best overall: Viktor **Best for:** Startups, operations teams and agencies that want recurring work done across many tools without building or maintaining workflows. Viktor is an AI employee who lives in Slack and Microsoft Teams. Instead of drawing a workflow on a canvas, you message him the way you would message a person on your team: “Every Monday, pull last week’s ad spend from Meta and Google Ads, compare it with Stripe revenue and post a summary with anything unusual to #growth.” He connects to 9,800+ tools, most through one-click OAuth, and when a tool has no API he can work through it in a browser. That is the difference between delegation and rule-building. With a builder, someone on your team designs the trigger, maps every field, adds the branches and fixes the flow when an app changes. With Viktor, you describe the result and he works out the steps each time, so a renamed column or an unexpected edge case does not silently break the job. When something is unclear, he asks in the thread. - **AI capability:** He plans the steps, reads messy inputs such as emails, PDFs and spreadsheets, decides what to do, and reports back where the team already talks. - **Finished deliverables:** Reports, spreadsheets, CRM updates, drafted emails, dashboards, internal apps and code changes, not just a chat reply. - **Approvals:** Money moves, code pushes and customer emails wait for your explicit yes in Slack or Teams. - **Recurring work:** Ask once and he runs the task daily, weekly or on a trigger. - **Pricing:** Usage-based credits shared across your workspace, with no per-seat fee. Start free with up to $100 in credits, then Team is $100 a month for 40,000 credits ([viktor.com/pricing](https://viktor.com/pricing)). - **Hosting:** Cloud only, US-hosted by default, with EU data residency on Enterprise contracts. **Where he is not the right fit:** millions of identical, sub-second events at the lowest possible cost per run. A dedicated builder or an engineer-owned pipeline is cheaper for pure high-volume plumbing. If you want the longer explanation of the model, read [what an AI employee is](https://viktor.com/blog/what-is-an-ai-employee) and how he differs from a workflow builder. The quickest test is a real task. [Add Viktor to Slack and give him one recurring job you would otherwise build in a canvas](https://app.viktor.com/signup). ### Best for non-technical users: Zapier **Best for:** Simple, reliable app-to-app automations set up by people who do not code. Zapier is still the default tool for no-code workflows, and it now pitches itself as an AI orchestration platform. It lists 9,000+ apps and most people can build a working Zap in an afternoon. On top of Zaps it now offers AI steps, Zapier Copilot (build a Zap from a description), Zapier Agents and Zapier MCP, which lets AI chat tools call Zapier actions. - **AI capability:** AI steps inside fixed workflows, plus Agents and MCP as separate products. - **Pricing (checked October 2026):** Free plan with 100 tasks a month and two-step Zaps. Professional starts at $19.99 a month and Team at $69 a month, both scaled by task volume. Agents are billed separately, with Agents Pro at $33.33 a month billed annually. - **Skill level:** Low. - **Self-hosting:** No. **Watch out for:** task-based billing on long, high-volume workflows, because every action counts as a task. For more options, see our [Zapier alternatives](https://viktor.com/blog/zapier-alternative-2026) and the [Viktor vs Zapier Agents](https://viktor.com/blog/viktor-vs-zapier-agents) comparison. ### Best visual builder: Make **Best for:** Ops people and agencies who want to see branching, multi-step logic on a canvas. Make (formerly Integromat) shows the whole scenario as connected modules, which makes routers, iterators and data transformation easier to follow than a linear list of steps. It has added AI modules and AI agents that can sit inside a scenario and choose between tools. For complex flows it usually costs less than Zapier at the same volume. - **AI capability:** AI modules and agents inside scenarios, with your own model keys or Make’s AI provider. - **Pricing (checked October 2026):** Free plan, then Core, Pro and Teams plans billed in credits. Most actions use one credit, while AI features can use more. - **Skill level:** Medium. You need to think in data structures. - **Self-hosting:** No, cloud only. **Watch out for:** dense scenarios that only one person understands. Debugging a failed run means reading module-by-module logs. We compare the two approaches in [Viktor vs Make](https://viktor.com/blog/viktor-vs-make). ### Best for technical teams: n8n **Best for:** Developer-led teams that want self-hosting, code steps and full control over data at low cost. n8n combines a visual builder with JavaScript and Python nodes, strong AI agent and LLM nodes, and MCP support. You can run it in n8n’s cloud or on your own servers. n8n is distributed under its Sustainable Use License, a fair-code license: the source is available and self-hosting for internal use is free, but it is not open source in the OSI sense, and some enterprise features need a paid license. - **AI capability:** AI agent nodes with tools and memory, LLM chains, and code anywhere in the flow. - **Pricing (checked October 2026):** Cloud Starter is €20 a month billed annually for 2,500 workflow executions with unlimited steps, and Pro covers 10,000 executions. Billing counts one execution per full run, however many steps it has. The self-hosted Community Edition is free, but you pay for the server and the time to maintain it. - **Skill level:** High. - **Self-hosting:** Yes. **Watch out for:** hidden ownership cost. Without someone who can host, update and debug it, n8n becomes a liability. See [n8n vs Zapier](https://viktor.com/blog/n8n-vs-zapier), [n8n alternatives](https://viktor.com/blog/n8n-alternatives) and [Viktor vs n8n](https://viktor.com/blog/viktor-vs-n8n). ### Best agent-first platforms: Lindy and Gumloop **Best for:** Teams that want to design their own AI agents, with approvals, for sales, support, research and marketing tasks. Agent-first tools start from the agent rather than the trigger. Instead of mapping every step, you give an agent instructions, tools and a goal, then add human-in-the-loop checkpoints where a person approves before anything is sent. That is the main difference from classic integration builders such as Zapier and Make, where AI is one step in a fixed path. You still design, test and maintain each agent yourself. - **Lindy:** Agents for email triage, meeting notes, lead outreach and support, with triggers, actions, approvals and computer use. Pricing (checked October 2026): every active user is billed from $29.99 a month for 3,000 credits, with top-ups at $10 per 1,000 credits and a 7-day free credit allowance. Lindy lists SOC 2 and GDPR, and HIPAA with a signed BAA on Enterprise. More in [Lindy alternatives](https://viktor.com/blog/lindy-alternatives). - **Gumloop:** A canvas for AI-heavy pipelines such as scraping, enrichment, content and reporting, with AI nodes and agents. Pricing (checked October 2026): Pro starts at $37 a month on credits, with a 14-day free trial and custom Enterprise plans. - **Relay.app**, once the third name in this group, shut down in September 2026. **Watch out for:** credit burn. Agent runs that call large models use far more credits than simple steps, so set spend limits before you scale. ### Best enterprise RPA: UiPath **Best for:** Legacy, desktop and document-heavy workflows that no-code integration tools cannot reach. UiPath automates applications that have no usable API by driving the screen the way a person would, clicking, typing and reading, and it adds document understanding and agentic automation on top. It is strongest in finance, insurance, healthcare and back-office operations where mainframes, Citrix apps and scanned documents are not going away. Automation Anywhere is its closest peer, with a free Community Edition and enterprise plans through sales. - **AI capability:** RPA robots, document AI and agents orchestrated across systems. - **Pricing (checked October 2026):** Free tier, Basic from $25 a month, and Standard and Enterprise plans quoted by sales. - **Skill level:** High. Most rollouts need a partner or an internal automation team. - **Self-hosting:** Yes, on-premises or cloud. **Watch out for:** brittle bots when screens change, and long implementations. Our explainer on [RPA vs AI agents](https://viktor.com/blog/rpa-vs-ai-agents) covers where RPA still wins. ## Other options we considered These tools are worth knowing but placed lower for this audience of founders and operations leads: - **Microsoft Power Automate:** the natural pick if approvals, documents and data already sit in Microsoft 365, SharePoint and Dynamics. It covers cloud flows and desktop RPA, adds AI Builder and Copilot, and IT can govern it centrally. Premium costs $15 per user per month billed annually (checked October 2026). Licensing is confusing, premium connectors are paywalled and it feels heavy outside the Microsoft stack. - **Workato:** enterprise iPaaS that combines integration, governance, environments, role-based access control and AI agents for ERPs, HR systems and data warehouses. Pricing is by contract with no public list, which puts it out of reach for most startups. - **Salesforce Agentforce:** AI agents built into Salesforce. Worth a look if your CRM and service desk already run on Salesforce, less so as a general automation tool. - **Pipedream:** excellent for developers writing code steps against APIs, too technical for business users. - **Activepieces:** open-source, self-hostable builder with AI steps and an MCP server. Younger, with a smaller integration library. - **IFTTT:** fine for personal and smart-home automations, too limited for business workflows. - **Relay.app:** shut down in September 2026. If you mostly want drag-and-drop builders, our list of [no-code automation tools](https://viktor.com/blog/no-code-automation-tools) goes deeper, and [the best automation tools](https://viktor.com/blog/best-automation-tools) covers the wider category beyond AI. ## AI automation vs traditional rule-based automation Traditional automation moves data with fixed if-then rules: when a form is submitted, create a CRM contact, then post to Slack. It is predictable, auditable and cheap per run. It also breaks the moment reality does not match the rule: an email that does not fit the template, a field renamed in an app, a customer who replies with a question instead of a yes. We wrote about this pattern in [why automations die quietly](https://viktor.com/blog/why-automations-die-quietly). AI automation adds reasoning. An AI step can read unstructured text, classify it, extract data and decide which branch to take. That is why the category of AI workflow automation tools exists: a large share of real operations work is not identical every time. There are three levels: 1. **AI as a step.** A fixed workflow with one AI action, such as “summarize this email”. Zapier, Make and n8n all do this well. 2. **AI agents.** The agent chooses which tools to call and in what order, within limits you set. Lindy, Gumloop, Zapier Agents and n8n agent nodes sit here. 3. **AI employee.** You hand over the job, not the workflow. Viktor takes the goal, does the work across your tools and comes back with the finished result. ### Handling AI failure modes AI steps fail differently from rules. A rule either fires or it does not; a model can return a confident wrong answer. Plan for four failure modes before you scale: - **Hallucination.** The model invents a field value, an order number or a policy. Ground it in real data from the source system and validate outputs against a schema before anything is written to your CRM or sent out. - **Prompt drift.** Outputs slowly change as models are updated or inputs shift. Keep prompts versioned, test them on a small set of real examples, and review a sample of runs every week. - **No deterministic fallback.** When the AI step is unsure, the workflow should route the item to a person or a fixed rule, not guess. Builders such as Make and n8n let you add an error branch; agent tools let you set confidence thresholds or ask-first rules. - **Silent failure.** Log every run with its inputs and outputs, and alert a human when a run fails or retries. Viktor handles the last two in the conversation itself: when something is ambiguous he asks in the thread instead of guessing, and he reports what he did after every run. The more autonomy a tool has, the more approvals matter. Anything that sends a customer email, changes a price, issues a refund or pushes code should wait for a human yes, with a log of what was done. Our guide on [keeping a human in the loop](https://viktor.com/blog/how-to-keep-a-human-in-the-loop-with-your-ai-employee) shows how to set those checkpoints. For the broader picture, see the [AI workflow automation guide](https://viktor.com/blog/ai-workflow-automation-guide) and [what an AI workflow is](https://viktor.com/blog/what-is-an-ai-workflow). ## Real pricing: task-based billing, self-hosting and token costs Entry prices hide the real cost, because every vendor counts a run differently. Take one ordinary AI workflow: a new support email triggers four actions (classify it with an LLM, look up the customer in the CRM, draft a reply, post it to Slack for approval). Here is what that workflow consumes at 1,000 runs a month (prices checked October 2026): ```json:object { "_key": "k310", "_type": "table", "rows": [ { "_key": "k303", "_type": "tableRow", "cells": [ "Tool", "Billing unit", "Usage at 1,000 runs/mo", "What it means for the bill" ] }, { "_key": "k304", "_type": "tableRow", "cells": [ "Viktor", "Credits for the work done, shared across the workspace", "One recurring job; credits scale with the work, not the step count", "Free with up to $100 in credits, then Team $100/mo for 40,000 credits; no per-seat fee and no separate model bill" ] }, { "_key": "k305", "_type": "tableRow", "cells": [ "Zapier", "Tasks (each successful action)", "About 4,000 tasks", "Above the entry Professional tier, so you move up the task tiers" ] }, { "_key": "k306", "_type": "tableRow", "cells": [ "Make", "Credits (about one per module run, AI uses more)", "About 5,000 credits plus AI usage", "Usually cheaper than Zapier for the same flow" ] }, { "_key": "k307", "_type": "tableRow", "cells": [ "n8n", "Executions (one per full run)", "1,000 executions", "Fits in Cloud Starter (€20/mo, 2,500 executions); self-hosted costs server time only" ] }, { "_key": "k308", "_type": "tableRow", "cells": [ "Lindy", "Credits per active user", "Depends on model and steps", "From $29.99 per active user per month, plus top-ups at $10 per 1,000 credits" ] }, { "_key": "k309", "_type": "tableRow", "cells": [ "Gumloop", "Credits", "Depends on AI nodes", "Pro from $37/mo, more for heavy AI usage" ] } ] } ``` **Token costs on top.** Builders that let you bring your own model key (n8n, Make, Zapier with your own OpenAI or Anthropic connection) also send you a separate model bill. Assume the LLM step uses about 2,000 input tokens and 500 output tokens per run. At 1,000 runs a month that is 2 million input and 0.5 million output tokens. At published API rates for a small model (about $0.15 per million input tokens and $0.60 per million output), that is roughly $0.60 a month. With a frontier model (about $3 input and $15 output per million), it is roughly $13.50 a month. Agent workflows that loop through several tool calls can use five to ten times more tokens per run, so model your real volume. **Self-hosting is not free.** n8n Community Edition costs nothing to license, but the server, backups, upgrades and developer hours usually cost more than a cloud plan for small teams. Add the hidden costs too: premium connectors, hours spent fixing broken runs, and the rebuild every time a process changes. Viktor’s credits are usage-based and shared across the workspace with no per-seat fee; current details are on [viktor.com/pricing](https://viktor.com/pricing). ## Security, compliance and self-hosting Automation tools hold OAuth tokens to your CRM, inbox and finance systems, so security review is not optional. Here is what each vendor states publicly (trust and pricing pages, checked October 2026). Confirm details with each vendor’s security team before you sign. ```json:object { "_key": "k332", "_type": "table", "rows": [ { "_key": "k324", "_type": "tableRow", "cells": [ "Tool", "SOC 2", "GDPR", "HIPAA", "Audit logs", "Trains on customer data", "Hosting" ] }, { "_key": "k325", "_type": "tableRow", "cells": [ "Viktor", "Type 1 certified, Type 2 in progress", "Aligned, DPA available", "Not stated", "Access logged; admins can revoke any integration in one click", "No, never used to train models", "Cloud, US by default, EU residency on Enterprise" ] }, { "_key": "k326", "_type": "tableRow", "cells": [ "Zapier", "Type II", "Yes", "No", "Enterprise plan", "Check the DPA and AI terms", "Cloud" ] }, { "_key": "k327", "_type": "tableRow", "cells": [ "Make", "Type II", "Yes", "Not stated", "Enterprise plan", "Check the DPA and AI terms", "Cloud (EU and US regions)" ] }, { "_key": "k328", "_type": "tableRow", "cells": [ "n8n", "Type II (cloud)", "Yes", "Self-hosted setups can be configured for it", "Enterprise plan", "Self-hosted data stays on your servers", "Cloud or self-hosted" ] }, { "_key": "k329", "_type": "tableRow", "cells": [ "Lindy", "Yes", "Yes", "With signed BAA on Enterprise", "Enterprise plan", "Check the DPA", "Cloud" ] }, { "_key": "k330", "_type": "tableRow", "cells": [ "Gumloop", "Yes", "Yes", "Not stated", "Enterprise plan", "Check the DPA", "Cloud" ] }, { "_key": "k331", "_type": "tableRow", "cells": [ "UiPath", "Type II", "Yes", "Supported", "Yes", "Check the contract", "Cloud or on-premises" ] } ] } ``` Beyond certifications, ask every vendor about role-based access control (who can build, run and approve), credential handling, data retention and how fast an admin can revoke access. Viktor’s security model is worth spelling out because it answers the questions most reviews start with. Credentials are injected by a backend tool gateway at execution time, so the AI model never sees your API keys or tokens. Data is encrypted with TLS 1.2+ in transit and AES-256 at rest, SAML SSO is supported, and workspaces are fully isolated. Your conversations and files are never used to train a model, neither Viktor’s nor the model providers’. Details are on [viktor.com/security](https://viktor.com/security). **When self-hosting matters:** if regulation or a customer contract says data cannot leave your infrastructure, n8n or UiPath on-premises are the realistic options. For most startups, a cloud vendor with SOC 2, a signed DPA, data-retention controls and approvals on sensitive actions is enough, and far less work. ## How to choose Start with the job, not the vendor. Write down the three tasks you most want off your plate this month, then map your team to a tool: ```json:object { "_key": "k352", "_type": "table", "rows": [ { "_key": "k345", "_type": "tableRow", "cells": [ "Your situation", "Best fit", "Why" ] }, { "_key": "k346", "_type": "tableRow", "cells": [ "You want the work done, not another tool to build in", "Viktor", "You hire an AI employee and delegate in Slack or Teams instead of designing workflows by hand" ] }, { "_key": "k347", "_type": "tableRow", "cells": [ "Small team, no developers, simple handoffs", "Zapier", "Easiest builder, biggest template library" ] }, { "_key": "k348", "_type": "tableRow", "cells": [ "Ops person who likes logic, complex branching", "Make", "Visual canvas, cheaper at volume" ] }, { "_key": "k349", "_type": "tableRow", "cells": [ "Developers on staff, data must stay in-house", "n8n", "Self-hosting, code steps, low cost per run" ] }, { "_key": "k350", "_type": "tableRow", "cells": [ "You want to design and tune your own AI agents", "Lindy or Gumloop", "Agent-first building blocks with approvals" ] }, { "_key": "k351", "_type": "tableRow", "cells": [ "Legacy desktop apps, scanned documents, large enterprise", "UiPath or Automation Anywhere", "RPA reaches systems without APIs" ] } ] } ``` Then ask three questions. Is the task identical every time? If so, a builder is efficient; if it changes with context, an AI employee saves the rebuild. Who will maintain it? If the honest answer is “the founder, on Sunday nights”, choose the lowest maintenance load, not the lowest sticker price. And what does it cost at ten times today’s volume, tokens included? Many teams end up with both: a builder for high-volume plumbing they already have, and Viktor for the work around it that needs reading, deciding and following up. [Start free with up to $100 in credits, no credit card required](https://app.viktor.com/signup), and compare the result with the builder you are considering. ## FAQ ### Zapier vs Make vs n8n: which should I pick? Zapier for non-technical people building simple automations fast, Make for visual control over complex scenarios at a lower cost per action, n8n for developers who want self-hosting and per-execution billing. If nobody wants to maintain workflows at all, delegate the work to an AI employee instead. ### Can I self-host an AI workflow automation tool for privacy? Yes. n8n and Activepieces can be self-hosted, and UiPath supports on-premises deployment. You take on hosting, upgrades and patches, and prompts sent to a hosted LLM still leave your servers unless you run your own model. ### Which tools work for non-technical staff? Zapier is the easiest builder to learn. Viktor needs no building at all: staff message him in Slack or Microsoft Teams in plain language, he works across 9,800+ tools, and he asks when something is unclear. Make and the agent-first tools are manageable for ops people, while n8n and UiPath usually need technical owners. ### When do I need enterprise RPA? When the work happens in desktop or legacy software with no usable API (mainframe screens, Citrix apps, old ERPs) or in large volumes of scanned documents. If all your tools are modern SaaS, RPA is usually more than you need. ### Where do AI agents need a human in the loop? Anywhere a mistake is expensive or public: emails to customers, payments and refunds, price changes, contract terms, code deployments and deleting records. Good tools let you require approval for those actions and keep a log of what was done. Viktor waits for an explicit yes on money moves, code pushes and customer emails. To see how that works on your own tools, [put Viktor on one recurring task this week](https://app.viktor.com/signup).