A walkthrough of Profit.co’s AI plan assistant: natural language commands, screenshot import, and iterative adjustment without touching a single cell.
The Old Way vs. The AI Way
Here’s what modifying a KR plan used to look like: open the plan table, scan 13 rows of weekly targets, mentally calculate the redistribution, click the first cell, type a number, tab to the next cell, type another number, realize you’ve overshot the total, go back and adjust three previous cells, verify the incremental targets sum correctly, and hope you haven’t introduced a rounding error. Elapsed time: four to ten minutes per KR, depending on complexity.
Here’s what it looks like with the AI assistant: open the plan table, click Import with AI, type “shift 15% from February into March,” review the preview, click Apply. Elapsed time: about 30 seconds.
This article walks through every step of the AI-assisted workflow — from opening the panel to applying the final values — with real examples you can use immediately.
Opening the AI Panel
-
Navigate to the Key Result and open the Modify Plan modal. Your existing plan (or auto-generated linear plan) loads in the table.
-
In the modal footer, click the Import with AI button (marked with a sparkle icon). The conversation panel slides in from the right side of the modal.
-
The assistant greets you with context-aware instructions. If you have an existing customized plan, it says: “Your current plan is loaded. You can replace it or tell me how to adjust it.” If the plan is a default linear distribution, it says: “I can help you set up a custom plan. Paste data, upload a file, or describe the shape you want.”
The panel is 380 pixels wide and doesn’t compress your plan table — the modal expands to accommodate both. You can see the full table and the conversation side by side, which is critical for reviewing AI-generated values against the current plan.
Five Real-World Scenarios, Step by Step
Scenario 1: Shifting Targets Between Months
Your dependency slipped. You need to move 20% of the January target into February and March.
You type: “Shift 20% of January’s target into February and March equally.”
The assistant processes the command in about two seconds. It identifies all January check-in rows, calculates the total January incremental target, reduces it by 20%, and distributes that 20% equally across the February and March rows. The plan table updates with a diff view: January rows show old values struck through with new lower values in blue; February and March rows show new higher values in blue.
You scan the diff, confirm it looks right, and click Update. Total time: 15 seconds.
Scenario 2: Changing the Plan Shape
Your plan was auto-generated as a linear distribution, but your team knows the work is back-loaded — the first month is setup, the real progress happens in months two and three.
You type: “Make it an S-curve. Slow start in January, steep ramp in February, tapering in March.”
The assistant generates a new distribution curve while preserving your From and To values. It shows a summary: “January now covers 12% of the total (was 33%). February covers 52% (was 33%). March covers 36% (was 34%).” The plan table shows the new values highlighted in blue with sparkle indicators on each AI-generated cell.
You notice February feels too aggressive. You type: “Make February a bit less steep. Move 8% into March.” The assistant adjusts. You’re iterating on the plan shape in natural language, without touching a single cell. Total time: 40 seconds for two iterations.
Scenario 3: Importing from a Spreadsheet Mid-Conversation
You’ve been discussing the plan with the AI, but then you realize your finance team already published revised targets in a Google Sheet.
You switch to the Google Sheet, select the relevant rows and columns, copy them, switch back to Profit.co, click Paste Data, and press Ctrl+V.
The assistant receives the pasted data and says: “I found 13 rows with dates and target values. The targets use a cumulative percentage format. I’ve converted to incremental targets for the plan table. Confidence: 94%.” It shows a compact preview card in the conversation thread with the monthly summaries.
You click Apply to Plan. The table populates. You notice one value looks off and click the cell directly to override it — the blue background fades to white, indicating it’s now user-confirmed. Total time: 25 seconds.
Scenario 4: Pinning Some Values, Filling Others
You know exactly what January should look like — you’ve manually set those targets. But you want the AI to handle February and March based on the remaining target.
You type: “Keep January as-is. Distribute the remaining target evenly across February and March.”
The assistant reads the current January values from the plan table, calculates the remaining target (To minus the January endpoint), and distributes it equally across the February and March check-in periods. January cells remain unchanged (no blue highlight); February and March cells show new AI-generated values.
This “pin and fill” pattern is one of the most common commands. It lets you combine manual precision where you have conviction with AI efficiency where you don’t. Total time: 10 seconds.
Scenario 5: Building a Plan from a Screenshot
You’re migrating from another OKR tool. The plan data is visible on your screen but there’s no export function.
You take a screenshot (Cmd+Shift+4 on Mac, Win+Shift+S on Windows), click the Screenshot chip in the AI panel, and select the captured image.
The vision model analyzes the image and extracts the table. The assistant reports: “I found a table with 12 rows. Dates appear to run from January to March with weekly intervals. The target column shows values from 0 to 100. Confidence: 89%.” It presents the extracted values in a preview card.
Two values look slightly off — the vision model read a “7” as a “1” in one cell. You tell the assistant: “The February 9th target should be 47, not 41.” It corrects and updates the preview. You apply, and the plan is set. Total time: 45 seconds, including the correction.
Understanding the Visual Cues
When the AI generates or modifies plan values, the plan table uses visual treatments to help you distinguish AI output from your own input:
| Visual Cue | Meaning | What You Should Do |
|---|---|---|
| Blue cell background with sparkle icon | This value was generated by the AI and has not been manually reviewed. | Scan it for reasonableness. Click to override if needed. |
| White cell background (normal) | This value was manually entered or has been explicitly confirmed by you. | No action needed. The AI will not overwrite this value in subsequent iterations unless you ask. |
| Struck-through gray text above blue text | Diff mode: the gray text is the old value, the blue text is the AI’s new value. Visible only during review. | Compare old vs. new. If the change doesn’t look right, click the cell to revert or type a new value. |
| Shimmer animation on a column | Recalculation indicator. Appears briefly when From/To values change, showing that incremental targets are being recalculated. | Wait for the shimmer to complete (under one second), then review the new values. |
Iteration Without Fear
One of the most powerful aspects of the AI panel is that you can iterate as many times as you want before saving. Each command the AI executes updates the plan table preview, but nothing is persisted until you click Update.
This means you can experiment freely:
-
Try a shape and undo it. “Make it linear” → review → “Actually, make it back-loaded” → review → “Shift 10% from March into February” → review → Update. Each command replaces the preview; only the final state is saved.
-
Combine AI and manual edits. Let the AI generate the broad shape, then click individual cells to fine-tune specific weeks. The AI respects your manual overrides in subsequent commands unless you explicitly target those cells.
-
Ask the AI to explain its work. Type “Explain the distribution” and the assistant describes the current plan shape: what percentage is allocated to each month, where the steepest ramps are, and whether any check-in has an unusually high incremental target.
The conversation panel is a scratchpad, not a commitment. Experiment with different distributions, compare shapes, and don’t worry about getting it right on the first try. The plan table only locks in when you explicitly save.
Tips for Effective Commands
The AI assistant understands natural language, but some phrasing patterns produce better results than others:
| Tip | Example |
|---|---|
| Be specific about time periods. | “Shift 15% from February into March” is better than “make February lower.” |
| Use percentages or absolute values. | “Reduce January by 10 percentage points” or “Set the Jan 19 target to 25%.” |
| Name the shape you want. | “Linear,” “S-curve,” “back-loaded,” “front-loaded,” “flat then ramp” are all recognized patterns. |
| Reference specific dates when needed. | “Keep everything before February 10 as-is” gives the AI a precise boundary. |
| Chain commands for complex changes. | “First make it linear, then double the first two weeks, then reduce the last week to 2%.” The AI processes sequentially. |
| Ask for what you want, not how to get there. | “I want 70% done by mid-February” is better than “Set weeks 1–6 to average 11.67% incremental.” Let the AI handle the math. |
When to Use AI vs. Manual Editing
The AI assistant and the manual plan table are complementary, not competing. Here’s when each is the better choice:
| Use the AI When… | Edit Manually When… |
|---|---|
| You need to change the overall distribution shape. | You need to tweak one or two individual cells. |
| You’re moving targets between months or weeks. | You know the exact number for a specific check-in date. |
| You’re importing data from an external source. | You’re making a small correction to an AI-generated value. |
| You want to experiment with different plan shapes. | You’re confirming values that the AI generated correctly. |
| You’re describing a plan concept (“slow start, fast finish”). | You’re entering a single value that was communicated to you. |
The ideal workflow combines both: use the AI to get 90% of the way there in seconds, then manually adjust the remaining 10% to match your exact judgment. That combination is faster and more accurate than either approach alone.
Thirty seconds. That’s the new plan modification benchmark.
Profit.co’s AI plan assistant turns your words into structured plans instantly. Describe the shape, paste the data, or upload a screenshot — and watch the plan table populate in real time.Start your free trial.
Related Articles
-
Top-Down Plan Distribution: How to Cascade Targets Without Losing Team Autonomy
Setting targets from the top is efficient, but rigid cascades kill ownership. Here’s how to distribute plans downward while preserving... Read more
-
Bottom-Up Planning: When Ground-Level Insights Should Override Executive Targets
Teams closest to execution often have the most accurate view of what’s achievable. A framework for when bottom-up plans should... Read more
-
The Aggregation Gap: What Happens When Bottom-Up Plans Don’t Meet the Top-Down Target
When team-level plans are aggregated upward and the total falls short of the organizational target, you have an alignment gap.... Read more
-
Plan Propagation Rules: Controlling What Changes Cascade and What Stays Local
Not every parent plan change should ripple downward. Not every child change should silently alter the parent view. Propagation rules... Read more