How AI Picks
Your Quadrant
(And When to Override It)

Most “AI productivity tools” generate task lists for you. That's the wrong problem to solve. The harder problem — the one that actually costs you time — is deciding where a task you already know about belongs. That's what Focus Quadrant's AI suggestion feature does: it helps you place tasks faster, not think of them for you.

The ✦ AI suggestion button appears whenever you type a task title in the queue. Click it, and within seconds you get a recommended quadrant, a confidence level, and a brief plain-English reason. You can accept the suggestion, add the task to the queue instead, or skip. The decision is always yours — the AI just gives you a fast first opinion.

How the ✦ Button Works

The workflow is intentionally frictionless. Type a task title — any title — and the ✦ button appears next to the input. Click it. The task title is sent to the AI model, which returns a JSON response containing three fields: the recommended quadrant, a confidence score, and a one-sentence reason. The suggestion renders as a card with three action buttons.

1. 1
   Type your task title

   The ✦ button appears as soon as you start typing. The task title is the only input — no additional context is required.
2. 2
   Click ✦ to get a suggestion

   The title is sent to Claude Haiku via OpenRouter. The model analyzes the language and returns a quadrant recommendation with a confidence level and brief reason — usually within 1–2 seconds.
3. 3
   Review the suggestion card

   The card shows the recommended quadrant (Do, Plan, Delegate, or Drop), the AI's confidence, and its reasoning. Read the reason — it tells you what the AI saw in your task title.
4. 4
   Accept, queue, or skip

   “Place in [Quadrant]” adds the task directly to the suggested quadrant. “Add to Queue” adds it to your staging area for manual placement later. “Skip” dismisses the suggestion without adding the task.

What the AI Is Actually Evaluating

The AI doesn't have access to your calendar, your project deadlines, or your company priorities. It has your task title — and that's it. So what does it look at?

The model evaluates two dimensions simultaneously. For urgency, it looks for explicit deadline language (“today,” “by EOD,” “Thursday,” “ASAP”), time-sensitive action verbs (“fix,” “respond,” “resolve”), and crisis indicators (“outage,” “broken,” “blocked”). For importance, it looks for outcome language (“finalize,” “launch,” “present”), strategic scope (“Q3 roadmap,” “investor,” “architecture”), and the difference between operational tasks (support a process) and value-creating tasks (produce an outcome).

Three Real Examples

Here's how the AI reasons through three different task titles, with the actual suggestion output shown.

When to Accept vs. Override

The AI gets it right most of the time when the task title is specific. The cases where you should override are predictable:

• ▸You have context the title doesn't capture. “Call Marcus” might look like Delegate to the AI, but if Marcus is your biggest client and this is a renewal call, it's Do. Add that context to the title next time — “Call Marcus — Q3 renewal conversation.”
• ▸The confidence is low. A low-confidence suggestion means the AI couldn't read clear urgency or importance signals from your title. Trust your judgment, and consider rewriting the title with more specificity.
• ▸The task has changed since you typed it. The AI evaluated the title you wrote. If the situation has evolved — the deadline moved, the stakes changed — the original title may no longer reflect reality. Update the title before re-suggesting.
• ▸The task is obviously placed. For tasks with crystal-clear priority — “server is down, fix now” — skip the AI button. Save your 10 monthly suggestions for tasks where you genuinely pause on classification.

The 10/Month Limit — And Why It Exists

Focus Quadrant defaults to 10 AI suggestions per month. This isn't an arbitrary constraint — it's a design choice rooted in how the feature works best.

The Eisenhower Matrix is a judgment tool. Its value comes from forcing a decision: is this task genuinely urgent? Is it genuinely important? Those questions are worth asking yourself. The AI suggestion feature is designed to help with the genuinely ambiguous cases — tasks where you stare at the title and can't immediately classify it. If you use the AI button on every task, you're outsourcing the judgment that the matrix is designed to build. The 10/month limit encourages deliberate use. Think of it as 10 “phone a friend” calls — you'll use them more carefully if they're scarce.

Privacy: What Gets Sent (And What Doesn't)

When you click ✦, exactly one piece of data is sent to the AI model: your task title. No account information, no task history, no project names, no email address. The AI request contains only the task title string.

After the suggestion is shown, Focus Quadrant logs an analytics event recording the suggested quadrant and your outcome (accepted, queued, or skipped). This event is used for two purposes: rate limiting (counting your monthly usage) and aggregate product analytics (understanding which quadrant classifications the AI makes most often). The event does not include the full task title — only the classification outcome.