New AI tools launch at a pace that's impossible to follow in detail. But the landscape is simpler than the LinkedIn feed makes it look. Here's my map — sorted by what the tools actually do, with honest notes on which ones I use myself and why.
First, let me take the drama out of it: you don't need to know every tool in this article. Most people get a long way with a single good AI tool they've come to know deeply. But it helps to understand the map — which categories exist, what sets them apart and where things are heading. That's the map I'm drawing here, based on what I use every day, both in building Familjegaraget and in Advanto's AI team.
The conversation partner, the writing help, the sounding board — the category most people meet first.
My own workhorse — the one that built Familjegaraget alongside me and that runs all of Advanto's AI team. Claude is known for long, coherent reasoning, strong coding and a tone that feels more like a colleague than a search engine. Anthropic has also built a whole family of specialised products around Claude — more on those below.
For you if: you want a thinking partner for text, analysis and building — and you value one that pushes back when you're wrong.
The tool that kicked off the whole wave, and still the world's most-used AI assistant. A broad ecosystem with image generation, voice mode and custom GPTs. For a lot of organisations, ChatGPT is synonymous with AI, and it's often the natural starting point.
For you if: you want an all-rounder with the biggest ecosystem and the most built-in features in one place.
Google's answer, deeply woven into Gmail, Docs and the rest of the Google world. The strength is exactly that integration — if your organisation lives in Google Workspace, the AI is already where you work.
For you if: you already live in Google's ecosystem.
Microsoft's AI, built into Word, Excel, Teams and Windows. For the Swedish public sector this is often the first AI that actually gets rolled out widely — because it arrives through agreements that already exist. That makes it important, whatever you think of it next to the competition.
For you if: you work in an M365 organisation and want AI in the tools that are already procured.
AI-driven search: ask a question, get a compiled answer with source references. It's become many knowledge workers' replacement for the classic Google search.
For you if: more than anything you want quick, sourced answers to factual questions.
The category that let a project manager with no coding skills submit an app to the App Store.
A code editor with AI built into the core. You describe what you want in a chat box — in Swedish if you like — and the AI writes and changes the code in your project. Cursor is where all of Familjegaraget was built, with Claude as the engine under the hood. For non-technical people, this is the most accessible way into building for real.
For you if: you want to build something of your own — an app, a website, a tool — without first learning to program.
Anthropic's own developer tool: an AI agent that works straight in the terminal or in an app, and can take on whole tasks by itself — "finish this feature, test it and fix what breaks". Where Cursor is an editor you sit in, Claude Code is more like a developer you delegate to. It's aimed mainly at developers, but that line is moving fast.
For you if: you want to delegate whole builds rather than steer them line by line.
The pioneer among AI coding assistants, now standard kit for many professional developers. Suggests and writes code right inside the developer's usual environment.
For you if: you're already a developer in a Microsoft/GitHub environment.
A Swedish upstart (worth a mention!) that removes one more step: you describe the app in plain text in the browser and get a running product straight away, without ever seeing any code. Perfect for prototypes and simpler services.
For you if: you want to go from idea to clickable prototype in an afternoon.
Through 2026, Anthropic has grown Claude into a whole product family — a pattern I think we'll see more follow.
Launched in April 2026: a visual workspace where you create prototypes, presentations and one-pagers by talking to Claude. You describe, Claude draws, and you refine with comments straight in the design. It can learn your organisation's visual identity and apply it automatically, and it exports to PowerPoint and Canva among others. For all of us with ideas but no design background, this is a big deal.
For you if: you need to produce visual material — sketches, presentations, prototypes — with no designer on the team.
Brand new — launched in beta at the end of June 2026. A "workbench" for researchers that brings databases, analysis tools and compute together in one environment, focused on the life sciences: genomics, protein structure, drug development. Every result comes with traceable code and history, so it can be reviewed and reproduced. Most of us will never use it — but it shows clearly where the AI agents are heading: from the chat window to independent, reviewable specialist work.
For you if: you work in research and data analysis — or want to understand where agent AI is heading.
Claude Code's sibling for those of us who aren't developers: an agent app for knowledge work that can handle documents, spreadsheets and a browser to solve whole tasks — not just answer questions. Think "go through these twenty documents and put together a report" rather than "what do you make of this paragraph?".
For you if: you want to delegate whole tasks — reviews, summaries, document handling.
The category that gave Advanto's AI colleagues their voices.
The market leader in AI voices: text to natural speech, voice cloning and complete voice agents that can hold a conversation in real time. It's ElevenLabs that gives Ulla and Viktor in Advanto's AI team their voices — and that cloned my own voice into AI-Christian. The quality in Swedish has become genuinely good, which opens the door to everything from accessibility to phone agents. That comes with a responsibility too: anyone who clones voices has to be transparent about it — something the EU AI Act also requires.
For you if: you want to give text a voice — for podcasts, training material, accessibility or voice agents.
From concept images to finished film — in minutes rather than weeks.
Image generation has become an everyday tool: Midjourney for the artistic, ChatGPT's built-in image creation for the quick and practical (that's how the portraits of Advanto's AI colleagues were made), and video generators like OpenAI's Sora and Google's Veo for moving material. My rule of thumb: brilliant for concepts, illustration and internal material — but always be transparent about what's AI-generated.
For you if: you need images and illustrations but have no agency budget.
Tools that aren't AI themselves, but are the precondition for anything you build with AI to become real.
The hosting service that takes the code the AI wrote and puts it on the internet — automatically, on every change. Familjegaraget lives here. Vercel also has its own AI efforts, among them the tool v0, which generates interfaces from text descriptions.
For you if: you've built something with AI and want the world to be able to use it.
Database, login and real-time features as a ready-made service. This is where AI-built apps connect their "memory". Free to start with and surprisingly simple even for non-technical people, because the AI handles the wiring.
For you if: you're building something where users need to log in and data needs to be saved.
Email sending for apps: confirmation emails, password resets, notifications. One of those services nobody talks about but every app needs. All of Familjegaraget's Swedish emails go through Resend.
For you if: you're building a service that needs to email its users.
If I have to boil the landscape down to a single thought, it's this: the AI tools are sorting themselves into three layers. At the top, the assistants you talk to (Claude, ChatGPT, Gemini, Copilot). In the middle, the agents that carry out whole tasks (Claude Code, Claude Cowork, Claude Design, Claude Science). At the bottom, the infrastructure that makes the result real (Vercel, Supabase, Resend — and specialists like ElevenLabs).
The trend between the layers is clear: the weight is shifting from the top to the middle. In 2024 we asked the AI for advice. In 2025 we let it write drafts. In 2026 we delegate whole tasks and review the result. That's exactly the shift I'm exploring with Advanto's AI team — seven colleagues, one of them human.
Want to see the tools in practice? Read the guide How I built a real app without knowing how to code — the whole Familjegaraget journey, step by step.
This article reflects the state of things in July 2026 — a landscape that moves fast. Prices, features and product names all change before you know it.
I share what I build, test and learn as I go — both here and on LinkedIn.
