Our note on AI
Helm's AI reads your board to write your daily brief, and does the jobs you hand its agent. Here is a plain-English account of what that means.
What AI we use
Helm's AI runs on OpenAI's models. To write your brief, read in a list you import, answer the Helm Organizer or do a job you hand the agent, Helm sends OpenAI the to-dos and words that job needs.
What Helm's AI does well
- Identifying which tasks are blocked, stalled, or overdue.
- Surfacing a short, ordered list of what to work on first.
- Adapting daily priorities as your board changes.
What Helm's AI cannot do
- Guarantee outcomes. A well-prioritised brief increases your focus — it does not guarantee you will finish everything.
- Understand context it cannot see. If a task's urgency depends on information not in your board, Helm's AI cannot factor it in. Add a note or update the card.
- Replace your judgment. Every recommendation is a starting point. You know your work, deadlines, and stakeholders better than any model does. Review and override as needed.
What you tell an agent stays between you
Even our agents work with agent-client privilege: whatever you tell them or share with them stays between you and them. Whatever you type to one of Helm's agents, and anything you share with it, is there to help you and goes no further.
No one but you can read what you keep in Helm — your account holds the only key. Your data explains both in everyday words.
What happens to what you give the agent
When you correct Helm's agent, it learns for you alone. What you teach it is never used on anyone else's to-dos. What you hand the agent is kept with your account so you can see what it did. Memriq's team is never shown what you asked: to decide what the agent should learn next, we see only the kinds of help several people wanted that it could not give, named by each person's own agent. Your data explains what Helm keeps, in everyday words.
Human oversight
Every suggestion is presented to you for review. Nothing is applied to your board automatically unless you instruct it. You are always in control of what changes.
Accuracy and feedback
If a recommendation feels wrong, ignore it. If you see something consistently unhelpful, we want to know: [email protected].