Goals let askROI keep working until the job is done

Try it

Pick something that usually takes you a few rounds of back and forth, and let askROI take the finish line instead of just the first step.

Think about how you use an AI assistant on a real task. You ask, it answers, you read it, you spot what is missing, you ask again. For anything longer than one step, you are the one holding the thread: nudging it along, remembering where you were, deciding when it is actually finished.

That is fine for a quick question. It gets tiring for real work, the kind that takes ten small steps and a couple of dead ends before it is done.

Goals change who holds the thread. You describe the finish line once, and askROI keeps working toward it on its own.

What a goal is

A goal is a finish line, not a task. A normal message says "do this next thing." A goal says "keep working until this is true." askROI takes that outcome, works a step at a time, checks its own progress after each step, and keeps going until the finish line is reached.

The difference sounds small, but it changes how the work feels. You are no longer prompting each move. You set the destination and let the assistant find its way there.

This is a real shift happening across AI right now, and it started in tools built for engineers. OpenAI's Codex added Goals, and Anthropic's Claude Code added a goal command that keeps working across turns until a condition is met. Both let an AI pursue a defined outcome instead of stopping after a single answer. askROI's goals bring that same idea to everyday work, in an app you can use from your phone.

What it means for you

• Less babysitting. You say what "done" looks like once, instead of steering the assistant through every step.
• Multi-step work actually gets finished. Research, cleanup, drafting, analysis: the kind of task that needs several passes can run to the end without you at the wheel.
• You stay in control. askROI shows what it is working on, checks in as it goes, and you can pause or stop it whenever you want.

A fair caveat, because this is honest work and not magic: goals need a clear finish line. "Organize my expenses into a spreadsheet I can download" is something askROI can check for. "Make my life better" is not. If a goal is too open-ended to verify, askROI says so and helps you sharpen it before it starts, which we will get to in a moment.

How askROI works toward a goal on its own

The core idea is a loop with a referee.

1. askROI does one turn of work toward your goal, using whatever the job needs: search, your connected apps, skills, and files.
2. A separate, quick check reads what just happened and asks one thing: is the finish line reached yet?
3. If yes, the goal is done. If not, askROI starts another turn, guided by what the check noticed was still missing.

An abstract goal loop: askROI works, checks progress, reaches the finish line, or loops back

The referee is the part worth pausing on. The model doing the work does not get to declare itself finished. A second, lightweight check looks only at what actually landed in the conversation, the real results rather than what the assistant hoped it did, and decides whether the goal is genuinely met. It is the same principle the developer tools that pioneered this settled on: completion is judged by evidence, not by the assistant's own optimism. It keeps a goal honest, so "done" means done.

Set the finish line before any work starts

Good outcomes start with a good goal, so askROI checks yours before it spends a single step on it. You will see one of three things.

• If your goal is clear and checkable, it starts right away. The card above your chat reads "Pursuing goal," with a counter for how many rounds it has run.
• If your goal is a good idea but underspecified, askROI asks one to three quick questions to pin down the finish line, the scope, and anything that must not change. You answer in the chat, it proposes a sharper version, and you confirm or edit it before it runs.
• If your goal has no finish line it could ever check, like "be more creative forever," it tells you plainly ("Goal needs a clearer finish line") and suggests a rewrite you can tweak and retry.

The goal setup card showing askROI asking for a clearer finish line before it runs

None of this burns a turn on a goal that was never going to work. It is a short setup that makes the autonomous part worth trusting.

Built to stop, not just to run

An assistant that works on its own is only useful if it also knows when to stop. askROI keeps a few limits in place, and you can see all of them.

• A check-in every 25 rounds. Even on a goal that is going well, askROI pauses after 25 autonomous turns and waits for you: "Paused for a check-in after 25 autonomous turns. Resume the goal to keep going." Nothing quietly runs off in the background.
• Pause, resume, or clear anytime. The goal card gives you a stop button whenever you want it. Pause if it is drifting, resume when you are ready, clear it to drop the whole thing.
• It waits when it needs you. If a step needs your approval or a real decision, the goal pauses and asks instead of guessing.
• It stops when the credits run out, and says so. No silent surprises on the bill.

Because a goal can run over many turns, askROI stays quiet while it works. There is no ping for every step, just a single message when it actually stops: finished, paused, or waiting on you. And when it finishes, it does not shout. A quiet line appears under the final answer that reads "Goal complete," with how long it took, what it cost, and how many rounds it ran.

A running goal card and the quiet Goal complete line after a finished answer

What people use goals for

• Turn a folder of messy notes into a clean, sectioned outline with a short summary on top.
• Comb through a spreadsheet and sort it into categories you can download.
• Research a topic across several sources and produce one tidy brief.
• Draft a piece of writing, review it against a checklist, and revise until every item is met.

The pattern is always the same: a job with a real finish line, and enough steps that you would rather not walk it yourself.

Where this is going

The whole industry is moving from assistants that answer to assistants that act. The first wave replied when asked. The next one takes an outcome and works toward it. That showed up first in tools for programmers, where "done" is easy to define: make the tests pass, make the file smaller. askROI's goals bring the same shape to the rest of your work, where the finish line might be a clean spreadsheet or a finished brief instead.

It is a modest promise, made carefully. Give askROI a clear finish line, and it will do the walking, check its own work, and stop when it gets there.

How to set a goal in askROI

The chat composer with Goal mode armed

1. In a chat, tap the Goal button in the composer to arm goal mode. A "Goal" chip appears where the model picker usually sits.
2. Describe your finish line the same way you would ask a question, and send it.
3. If askROI asks, answer its quick setup questions, then confirm the goal it proposes.
4. Watch it work. The card above the chat shows "Pursuing goal" and the round counter. Expand it to read the latest progress.
5. Pause, resume, or clear from that same card at any time. When the goal is met, you will see "Goal complete" under the final answer.