Working With Foundry's AI
9 AI features, finite monthly budgets, and a clear rule for which to use when.
By the end of this module
The Operating Stance
Foundry's AI is a structured second opinion, not an answer machine. Every feature is designed around a single principle: the AI reads your workspace data, suggests something, and you decide.
This matters because the worst way to use AI in a tool like this is to outsource judgment. The second-worst way is to dismiss it as a gimmick. The right way is to treat the output as input, useful when you read it critically, harmful when you accept it uncritically.
Two practical implications: never blindly accept AI scores (review every one before saving), and never skip running AI on a high-stakes decision because "you've already thought about it" (the point is catching what you missed).
The Nine Features and Their Right Moment
| Feature | Monthly Limit | Use When |
|---|---|---|
| --------- | -------- | ---------- |
| AI Idea Categorization | 200 | Always-on (auto-suggests type and tags at capture) |
| AI Score | 100 | Triaging 5+ ideas in one session, or when scoring fatigue would otherwise compress your range |
| AI Idea Expansion | 100 | A captured idea feels promising but is too thin to triage |
| AI Decision Prep | 50 | Before any decision where wrongness costs >$1K or >a week |
| AI Decision Review | 50 | After drafting a decision but before committing, blind-spot scan |
| AI Founder Briefing | 30 | Start of week or before a strategic planning session |
| AI Weekly Summary | 10 | During weekly review Step 5, before setting intentions |
| AI Pattern Detection | 10 | Monthly, or during quarterly portfolio review |
| AI Pricing Decision Helper | (varies) | When making a pricing or packaging decision |
All require Pro. Your data is processed by Claude (Anthropic) and is never used to train models.
Budgeting Your Monthly AI
The limits exist for a reason: AI is expensive, and unlimited usage trains low-quality habits. Treat the monthly allowance like a budget.
A sustainable monthly profile for a typical solo founder:
- ~40 AI Scores during weekly triage (4 sessions × 10 ideas)
- ~8 Decision Reviews for the meaningful decisions of the month
- ~4 Founder Briefings (one per week if Mondays warrant it)
- 2 Pattern Detections (one monthly, one mid-month)
- 4 Weekly Summaries (one per review)
That leaves headroom for unplanned heavy weeks. Burn through 100% of every limit every month and you're using AI as a crutch, not a tool, back off the volume and use the human judgment that the AI was meant to augment.
AI Best Practices, Concretely
On AI Decision Review: Use the "missing alternatives" section hardest. The AI is best at catching what you didn't consider; it's only OK at evaluating what you did. Read alternatives twice; read the rest once.
On AI Founder Briefing: Read the "blind spots" section first. Save the "top priority" for last and only accept it if you'd have written something similar yourself, otherwise you're letting the AI set your day.
On AI Pattern Detection: Run it the day after a quarterly portfolio review, not the day of. Sleep on the patterns. The features that pattern-match across 60 days produce conclusions you can't process in real time, give them time to settle before acting.
When Not to Use AI
For trivial captures. A Quick Note doesn't need AI categorization. The auto-suggest fires anyway, but don't deliberately invoke AI Expansion on a captured grocery-list-shaped thought.
When you already have a strong intuition. The AI is most valuable when you're genuinely uncertain. Burning a Decision Review on a call you've already made is wasted budget, use it on the next genuinely hard decision.
Out of habit. Re-read this section if you're hitting AI limits before the 20th of the month. The system isn't broken, your usage pattern is.
Practice Checkpoints
Sign in to track your progress and mark checkpoints complete.
Next in Foundations Path
Quarterly Planning From Data
Use 90 days of Foundry data to plan the next quarter, not vibes, not aspiration.