Notion AI for SMEs: when the wiki starts thinking

For many SMEs, Notion is the place where everything is finally written down: workflows (A workflow specifies what information goes in, what happens to it and who checks the result.), customer notes, project lists, the handbook for new people. Whoever built that created something valuable. The question is whether the AI in it should do more than search and write.

This article belongs to our series of comparisons: the same six questions for every tool. We read Notion’s information on 14 October 2026; the sources are listed at the end.

What you use today

Notion AI is included in the Business and Enterprise plans. Business costs USD 20.– per person per month billed yearly and USD 24.– billed monthly; in Switzerland Notion adds VAT. For five people that is USD 100.– to 120.– a month. The Free and Plus plans get only a limited number of AI responses. Custom agents (An AI agent carries out several steps of a defined task. A person decides before it sends anything outside your firm or spends money.) that work in the background use credits at USD 10.– per 1,000.

Where Notion AI is strong

Notion AI works where your knowledge already lives. Notion Agent creates pages and databases, changes properties, reads PDFs and spreadsheets and drafts emails in Gmail that you confirm before sending. Meeting notes work without a bot, and search reaches beyond Notion into Slack, Google Drive, SharePoint, Outlook and GitHub (A Git repository is storage for files that records every change with its date and author; GitHub is a well-known service for it.).

Custom agents work in the background: on a schedule, when a page is added to a database, when a meeting ends or when something happens in Slack. They may call external MCP servers (A server is a computer that runs all the time and provides services to other devices, such as email, files or a website.), through the open protocol AI uses for tools, and write actions ask for confirmation by default. In the other direction, Notion provides its own MCP server (MCP (Model Context Protocol) is an open standard for connecting an AI harness to other software and data sources.), through which Claude (An AI assistant is a chat program such as ChatGPT, Claude, Gemini or Perplexity: you write in plain language, and a language model answers.), ChatGPT or Cursor read and edit your pages.

By default Notion trains no models on your data, the AI providers (An AI provider runs language models and sells access to them, for example OpenAI (ChatGPT), Anthropic (Claude) or Google (Gemini).) keep data for 30 days at most, and there are SOC 2 and ISO 27001. For a team that likes to document, it is a very good environment.

Where Notion AI stops

• The work ends at the edge of the workspace. Files on the company drive, the accounts, the supplier portal and local programs sit outside. Your own code runs as a worker on Notion’s servers, and external MCP servers must be reachable from the internet.
• Automations stay in Notion. Database automations change properties, add pages, send notifications, emails or webhooks. That is all, and one automation cannot trigger another.
• The data lives in the US. EU data residency is for Enterprise only, on request, and without mail and calendar. The AI runs through providers such as OpenAI, Anthropic and Google, also outside the region (For cloud providers, a region is the area where the data centres that process your data are located, such as the EU or Switzerland.).
• Agents cost extra. Credits for custom agents are only available on Business and Enterprise; when they run out, the agents pause.
• Keeping order is manual work. A wiki is only as good as its upkeep. The AI also finds the outdated page from 2023 if nobody archived it.

Notion connected to our platform

The first way lets Notion stay your wiki and connects it to the Jules Solutions Platform in both directions. A custom agent in Notion calls the platform’s MCP gateway when it needs more than Notion: reading an invoice, producing a document from your template (A template is a prepared structure, for a document, an email or a folder for example, which you only complete for each case.), starting a workflow that talks to Bexio or Abacus within the agreed scope. In the other direction, the platform’s AI agents read and write your pages through Notion’s MCP server, with the rights of the person who approved the connection.

The knowledge stays where your team looks for it, and the work happens where it belongs. The platform sends a message to the outside or a payment only once a person has decided. Each connection sees only the projects you grant it, and every step is in the log (A log records what a program or an AI agent did and when, so that afterwards you can trace who triggered what and what happened.).

For five people: Operator System (Operator System is our package for teams at CHF 3'500.– to 5'000.–: the shared base with workflows for the team and connected systems.) as the shared base, CHF 3'500.– to 5'000.– once, with six months of Business seat (The Business seat is one person’s access to our platform (open alpha); Pro Setup includes three months, and with Operator System it is six per person.) per person; then CHF 199.– per person per month, or CHF 169.– billed yearly. You keep paying for Notion as you do today.

The whole system

In the whole system, the wiki becomes a vault (The vault is a folder of your working files on your computer; we create it from the vault template.): a folder on each person’s computer, organised by PARA (PARA organises storage into four parts: projects, areas, resources and archives.), written in Markdown (Markdown is a simple way of writing text files: characters such as # or * mark headings, lists and emphasis.) and readable in Obsidian (Obsidian is a note-taking program that shows a folder of text files on your computer as linked notes.). Markdown is text that any program opens, today and in twenty years. What the team shares lives in the shared project repository, with versions and someone who accepts changes. You export your existing Notion pages as Markdown; we sort them into the vaults during setup.

An AI harness (An AI harness is the workspace on your computer in which AI reads files and carries out agreed steps.) works in it; we recommend Claude Code (Claude Code is an AI harness made by Anthropic: it runs on your computer, reads files and carries out steps you describe in plain language.). The platform searches your project files and shows the passages an answer rests on; each agent sees only its own project. Workflows start with a short command (A command starts a defined workflow. In Pro Setup, we set up three to five commands for your work.), with an incoming email or on a schedule, and run on your device, on a computer in the office, on our platform or with a cloud provider (The cloud means computers in a provider’s data centres on which programs run and data is kept, instead of on your own device.), as the quote sets out. The platform turns voice notes into text.

The platform runs on our own infrastructure in Switzerland and is in open alpha (Open alpha means the platform is at an early stage: we use it ourselves, and it changes all the time.); we run our own company on it. For five people: Operator System plus one Pro Setup (Pro Setup is our recommended package at CHF 1'490.–: vault, workflows with AI agents, first integrations, an introduction session and three months of Business seat.) per person, CHF 10'950.– to 12'450.– once, with 60 days of follow-up support (Follow-up support means that questions and adjustments by email and phone are included for 30 days (Pro Setup) or 60 days (Operator System) after handover.), then the Business seats and model usage (Model usage means accessing an AI provider’s language model. The cost is not included in the setup price.).

Our verdict

If your team likes living in Notion and the AI is mainly there to search, summarise and keep pages up to date, Notion Business is a very good tool at USD 20.– a person. Stay with it, and spend the time on keeping your wiki in order.

But once the AI should do things outside Notion, read invoices, talk to systems, reach files on the drive, the first way is the smallest step: Notion stays the wiki, and the platform does the work. The whole system pays off when your knowledge should sit in an open format on your own computers and workflows should also run where Notion cannot go.

The questions to settle before you decide are in ‘Five questions before you commission AI automation’.