Frequently asked questions

Stillhelm is an executive management layer for people and, over time, the organizations around them.

The essentials

What exactly is Stillhelm?

Stillhelm is an executive management layer for people and, over time, the organizations around them. It starts with the individual: building one working picture of priorities, decisions, commitments and outcomes, then using that context to notice drift, direct attention and coordinate the AI agents you connect. The longer-term architecture extends that same management layer to teams, businesses and enterprise.

How is Stillhelm different from ChatGPT or another AI assistant?

ChatGPT is primarily something you ask; Stillhelm is designed to keep managing context between those conversations. It maintains a persistent picture of your goals, commitments, routines and decisions, surfaces issues without waiting for the perfect prompt, and can coordinate other AI capabilities around you. Stillhelm is therefore not trying to replace foundation models — it is the management layer that helps decide when and how to use them.

What can Stillhelm actually do for me?

It takes the coordination off your plate, so your attention goes to the decisions that need you. Plan: Helps you pick a realistic Top 3 for today, next to the time your calendar actually leaves. Adapt: When a meeting runs over or something new lands, it works out what no longer fits and proposes an adjustment for you to approve. Follow through: Brings slipping commitments, blocked work and routines that are starting to break to your attention while there is still room to act. Review: Prepares a weekly review of what moved, what stalled and what deserves your attention next. Coordinate: Puts what the AI agents you connect notice and propose into one queue of decisions, instead of separate chats.

Does Stillhelm make decisions for me?

No. Stillhelm notices, weighs and recommends; you decide. You approve, decline or postpone each recommendation. Changes to your plan, tasks or commitments wait for your go-ahead, unless you have switched on a narrow exception yourself. Stillhelm keeps a record of approved changes, and supported actions can be reversed. What Stillhelm observes can make something relevant; it never makes it permitted.

What does Stillhelm have access to?

Only what you connect or add yourself; it never gets access to your whole phone or computer. That means what you keep in Stillhelm, the Google Calendar and Todoist accounts you connect, and on iPhone only the sources you allow one by one: Apple Health, contacts, HomeKit, screen activity and location. Stillhelm does not currently have general access to your email, messages or files; you can share something with it on purpose. A connected agent needs a separate permission for health data, screen activity, location and your home, and it sees the relationships you confirmed in Stillhelm, never your phone's address book.

Privacy and control

Does Stillhelm train on my personal data?

No. Your private data is not used to train shared AI models. Stillhelm can use your own context, decisions and outcomes to build a more useful experience for you over time. That learning stays part of your personal Stillhelm experience; it is not used to train a model for other users, and Stillhelm does not sell your personal data.

How do permissions work?

Seeing and acting are separate permissions, and both are yours to give and take back. What Stillhelm can see: Each source is its own connection or phone permission; disconnect one and Stillhelm stops reading it. Health data, screen activity, location and your home never enter Stillhelm's chat automatically. What an agent may do: An agent states what it needs, you approve a specific set of permissions, and it cannot widen them itself. You can pause or remove it in one step. Access it has on its own platform stays with that platform. What may happen without asking: Nothing consequential by default. You can switch on a narrow exception, such as letting Stillhelm put a task on today's list when its link to your goal is unambiguous, and switch it off again. Permissions are checked again at the moment anything runs, so a permission you withdraw stops counting straight away.

What happens when Stillhelm gets something wrong?

It will sometimes, so it is built to be checked rather than trusted blindly. Stillhelm is designed to show the evidence behind each recommendation, so a wrong assumption is visible instead of hidden in a confident sentence. Declining a recommendation does not execute it. Stillhelm can reconsider the issue later if the underlying situation materially changes. Stillhelm keeps a record of approved changes, supported actions can be reversed, and approved recommendations are checked later against what actually happened. Where it can't confirm something, it says so instead of guessing.

How it works

Does Stillhelm work on its own, or do I have to keep prompting it?

You don't have to keep prompting it. When you ask: Ask Stillhelm answers from your own goals, tasks and calendar, and proposes changes that are only made after you confirm. Without asking: Stillhelm keeps your picture current from what you connect, prepares your weekly review and reports, reminds you on the schedule you set, and opens with what needs you first. With an agent connected: Stillhelm also re-checks your plans and commitments through the day, wakes the agent responsible when something in its area changes, and brings back what it notices or proposes. Waits for you: Anything that changes your plan, tasks or commitments, unless you have switched on a narrow exception yourself. Not yet: Acting in the outside world on its own, such as sending messages, booking or paying.

How does Stillhelm know what matters to me?

You tell it, and it keeps checking that against what actually happens. Your goals, life areas, commitments and the boundaries you want to protect set the direction; your calendar, tasks, routines and connected sources show what is really happening. What you said you want outranks what Stillhelm observes, what it doesn't know stays an open question instead of a guess, and it never assigns you a personality. You can review and correct that picture on one page: What Stillhelm knows.

Does Stillhelm learn over time?

Yes. Stillhelm becomes more useful as it builds up your own context, decisions and outcomes. Approved recommendations are checked afterwards against what actually happened, so the next one can draw on your own track record, and your patterns are compared with your own earlier weeks, never with other people. Both run in Stillhelm's checks through the day, which start once you connect an agent. That learning stays part of your personal Stillhelm experience; it is not used to train a model for other users, and Stillhelm does not sell your personal data.

AI agents

Is Stillhelm an assistant, a COO or an AI team?

Stillhelm is the executive partner; the AI team is optional and works under it. An assistant handles the request in front of it. A COO-style partner keeps sight of the whole: priorities, trade-offs and follow-through. Stillhelm takes on that role as software, starting with the individual, and the specialist AI agents you connect form the team it coordinates. If you already work with a human assistant, Stillhelm doesn't replace them; it holds the bigger picture their work fits into.

Can Stillhelm run a team of AI agents for me?

Today, Stillhelm can coordinate agents you connect; it does not yet provide or host the full agent workforce itself. You can pair supported external agents with Stillhelm, after which Stillhelm can give them work within the permissions you grant. Stillhelm doesn't need to be the best specialist at every task: its job is knowing what needs to happen, which agent should handle it, what context that agent needs and what the result means. Available: A coordinating agent with specialists, each with its own permissions; Routine Packs from the marketplace that give them a defined job; and results checked against what was asked, with you as the final approver. In development: Paid packs from creators. Direction: Building toward a more complete managed AI workforce is part of the direction, but it should not be confused with what is hosted by Stillhelm today.

Getting started

Who is Stillhelm for?

Stillhelm starts with individuals carrying significant responsibility — founders, executives, operators and ambitious professionals. The product is being built individual-first, but the underlying management model is intended to extend to teams, businesses and enterprise as Stillhelm evolves. It's a weaker fit if you only want a simple to-do list, or full automation without review.

Do I need to be technical?

No. You work with goals, decisions, approvals and outcomes; the technical machinery stays out of sight. Google Calendar connects with a sign-in, Todoist with a key you copy from its settings, and each iPhone source is one permission prompt. You need no agent to start. Bringing your own agents is the one step that still takes setup today: the agent has to exist first, for example on Grok, and you pair it with a code from Stillhelm, which then handles its permissions, approvals and history.

How do I get access?

Stillhelm runs in your browser, with an iPhone app for the sources that live on your phone, in English and Dutch. Start small: one goal, one routine, your calendar.

The bigger picture

Why a management layer instead of more AI tools?

Because the bottleneck is shifting from access to intelligence to coordination of intelligence. You can already have multiple models, agents, calendars, task systems and data sources. But somebody still has to decide what matters, keep them aligned, notice when something changes and determine when you need to step in. Today, that coordinator is usually you. Stillhelm is being built to become that layer.

What is Stillhelm ultimately trying to become?

The management layer between people and a growing supply of intelligence: first for individuals, then for the teams and organizations around them. Intelligence and execution capacity are growing faster than anyone can direct by hand. What is missing is the layer that decides what matters, which capability takes on which job, with what context, whether the result was good enough and when a person needs to step in. That is the layer Stillhelm is building, and its first job is protecting the one resource that doesn't scale: your attention.