Every task app is adding AI right now. Motion auto-schedules your calendar. ClickUp Brain generates project plans. Monday builds workflows from a sentence. Todoist writes your task descriptions for you.
The pitch is always the same: AI will manage your tasks so you don’t have to.
And for some people, that works. If you want an algorithm deciding when you do deep work and when you answer emails, Motion is built for that. If you need AI to summarize a hundred tasks across a large team, ClickUp has you covered.
But there’s a problem with the AI-in-everything approach, and it’s the same problem that made task apps overwhelming in the first place: more automation means more complexity, and more complexity means more time managing the system instead of doing the work.
AI feature bloat is still feature bloat
The first generation of task apps added too many features. Views, filters, tags, priorities, project hierarchies, custom fields. The second generation is doing the same thing, except now the features are “AI-powered.”
Auto-scheduling sounds good until the algorithm reschedules your afternoon because you finished a task early and now your entire day looks different. AI-generated subtasks sound helpful until you’re reviewing fifteen machine-generated items you didn’t ask for, trying to figure out which ones actually matter.
The pattern is familiar. The app does more. You manage more. The tool that was supposed to save you time becomes another thing that demands your attention.
Ember was built around the opposite idea: the app should get out of the way. Two views — Now and Next. A daily reset that clears your slate every morning. No priority levels, no project hierarchies, no configurations to maintain.
So when it came time to think about AI, the question wasn’t “what can AI automate?” It was “how do you add AI without breaking what makes the app simple?”
Tag it, don’t prompt it
Ember’s approach to AI is delegation, not automation.
During your normal task review — the same flow where you’re checking things off, adding new tasks, moving things between Now and Next — you can mark a task as delegated. One gesture. Swipe a single task, or select several at once and delegate them in bulk. No prompt. No context-switching. You’re telling the agent “I want help with this” and moving on.
This matters more than it sounds. If you’ve ever stared at a task you know you need help with and thought “I’ll figure it out later” — that’s not laziness. That’s the gap between recognizing what you need and having the activation energy to start. For people whose brains resist initiation (ADHD, executive function challenges, or honestly just a packed day), that gap is where tasks go to die. Delegation closes it. You don’t have to formulate the perfect request. You mark it and move on. The hard part — getting the conversation started — happens for you.
When you sit down with an AI agent later (Claude, or any agent connected through the Model Context Protocol), the agent already knows what you need. It checks your delegated tasks and starts there.
But it doesn’t start working immediately. It starts by asking questions.
An agent that asks before it acts
Say you’ve delegated two tasks during your morning review: “Research birthday venues” and “Plan birthday menu.” The agent picks them up and the conversation goes something like this:
Agent: I see you’ve delegated 4 tasks. Two look related — “Research birthday venues” and “Plan birthday menu” are both in your Personal list. Want to tackle those together? I also see “Compare home insurance quotes” in Now, which seems higher priority. Where would you like to start?
You: Let’s do the birthday ones together. It’s for my mom, she’s turning 60. Somewhere in Austin, budget around $2K for the venue.
Agent: Got it. Let me research venues in Austin that work for a 60th birthday party within $2K…
You never wrote that prompt from scratch. You tagged the tasks during your morning review, and the agent met you where the intent already existed.
This is the core idea: users think in terms of their task list, not prompts. Delegation bridges the gap between recognizing “an agent could help with this” and actually getting that help — without changing your workflow.
What delegation is not
Delegation in Ember is deliberately narrow. It’s worth being clear about what it doesn’t do, because the boundaries are as intentional as the feature itself.
It’s not auto-scheduling. The agent doesn’t rearrange your day or decide when you should work on things. You decide what goes in Now. You decide what stays in Next. The agent helps with the tasks you explicitly hand it.
It’s not auto-prioritizing. The agent uses your existing Now/Next placement as a signal — a delegated task in Now is more urgent than one in Next — but it doesn’t override your judgment. It suggests where to start. You decide.
It’s not a black box. The agent doesn’t disappear with your task and come back with a finished result you didn’t ask for. It clarifies scope first. It asks questions. It presents options and findings. You make the decisions. The agent does the legwork.
It’s not a new interface to learn. There’s no AI sidebar. No chat panel embedded in the task app. No new workflow to configure. You tag tasks in the same views you already use. The conversation happens in whatever agent you already talk to — Claude Code, Claude Desktop, any MCP-connected tool.
Bring your own agent
There’s no built-in chatbot. No AI assistant living inside the app. That’s deliberate.
Ember connects to any agent that speaks the Model Context Protocol — Claude, or whatever tool you already use. And that means the agent you bring already knows your setup. It has your workspace, your tools, your files, your preferences. You’re not starting from zero with a generic assistant embedded in a task app. You’re handing tasks to an agent that’s already configured for your life.
This is an advantage, not a limitation. A built-in chatbot would be a blank slate every time. Your agent has context that Ember doesn’t need to replicate — and Ember has structure that your agent doesn’t need to build. The app captures intent through delegation. The agent acts on it with whatever capabilities it already has. Each does what it’s good at.
The protocol is open. The task data is yours. The agent is whatever you choose. There’s more to say about why this matters — but the short version is that delegation works better when the agent already knows you.
The model: agent as teammate
The pattern Ember follows is closer to how you’d work with a capable colleague than how most AI products work.
You wouldn’t walk up to a coworker and hand them your entire schedule. You’d say “can you look into these two things for me?” and they’d ask follow-up questions before starting. If they found something unexpected, they’d check in. If the work spawned new tasks, they’d flag those too.
That’s the delegation model. The agent identifies what you’ve asked for help with, clarifies what you actually need, does the work using its own tools, and creates follow-up tasks if the work reveals things you need to handle yourself.
And the system keeps the agent honest. Ember’s delegation protocol instructs the agent to ask before acting, respect your Now/Next priorities, and present options instead of making decisions. The agent is constrained by the structure you already use — it can’t ignore your priorities or go rogue on tasks you didn’t delegate. That constraint is the feature. It means you can trust the handoff.
It doesn’t over-automate. It doesn’t take control. It does the part you don’t want to do — the research, the comparison, the information gathering — and hands you the results so you can decide what to do next.
Two views, a daily reset, and an agent that listens
Ember’s product decisions connect to each other.
The Now/Next system keeps your focus view intentionally small. The daily reset prevents accumulation. And AI delegation follows the same philosophy: do one specific thing well, stay out of the way, let the user stay in control.
The app captures your intent. The agent acts on it. Neither one tries to be the other.
If you’re already using Ember, connect your agent and start delegating tasks.
If you haven’t tried Ember yet, it’s on the App Store. Two views. A daily reset. And an approach to AI that respects the simplicity you came for.