AI Suggestions
Choosing what goes on a site is the slowest part of designing one. The suggestion tools shortlist plants, animals, structures, technologies and soils for your conditions, and you decide what survives the shortlist.
1. What a suggestion is
A suggestion is a proposal with a reason attached, drawn from the catalogue in the app and filtered against your project. It is never applied on its own. Nothing enters your design until you accept it, and rejecting something costs nothing.
These tools live in the Design stage
They are available once Design is unlocked for that project — a one-time, per-project activation, not a subscription or a plan you upgrade to.
2. Where suggestions appear
Element selection
Flora, fauna, structures, technologies and soils, each with its own selector. This is where most of your component list comes from.
Zone components
Once you have drawn permaculture zones, this proposes components for one specific zone, given how that zone is used and what conditions it has.
Sector analysis
Reads the sectors you drew — sun, wind, water, access — and reports what they imply for placement.
Guild composition
The Plant Guild Designer's own pipeline, which proposes whole plant communities rather than single species.
3. How a shortlist is built
Element selection runs in three passes, and each pass has a different job.
- Pre-filter. A fast, cheap pass narrows the catalogue to what could plausibly survive on your site at all. It works on your climate, soil and water data.
- Selection. A larger model proposes a set from what is left, with a reason for each entry — why this species, in this position, for this project.
- Suitability review. A third agent checks the proposal back against your conditions and flags anything marginal, so a plausible-sounding choice does not slip through unexamined.
You see the result of all three. A flagged entry is not removed for you — it is marked, because whether the risk is acceptable is your call.
4. What the suggestions are based on
Every selector reads the same two things: your project's site data — climate, soil, water and existing vegetation — and the findings the research stage extracted. That is the whole basis. It does not know your budget, your local nursery, your regulations or your taste.
Fill the site data before you ask
Empty tables do not stop a selector; they make it fall back on generalities about your region. Ten minutes spent on your soil and climate records changes the quality of every suggestion that follows.
5. Judging a suggestion
Read the reason, not just the name. A suggestion is worth accepting when its reason refers to something true about your site — a drainage problem, a frost date, a slope. A reason that would apply to any site anywhere is a sign that the model had little to work with.
Check anything that will be there for decades
Trees, earthworks and buildings are expensive to be wrong about. Verify species against a local source, and check what is invasive or restricted where you are — the app does not know your jurisdiction.
6. What suggestions cost
Each suggestion run is a metered AI response, charged in credits on the amount of text involved. The pre-filter pass is deliberately small and cheap; the selection pass is the expensive one. Reviewing and rejecting what came back is free — you are charged for generating a shortlist, not for reading it.
7. Where to go next
Design tools
The whole design stage — canvas, guilds, floor plans and visualisation.
The agents behind them
Which agent does which pass, and how they hand work along.
AI research
Where the findings the selectors read come from.
Credits
What each run costs and how credits are topped up.