Gains:
- Ability to divide a project into stages from concept to delivery and determine the role of artificial intelligence, the verification to be activated and the responsible expert at each stage.
- Asking the risk bucket question at the end of each stage and being able to confirm the local plant, legislation and cost at its own stage.
- Ability to integrate AI produces, human decides, every critical output is verified, signature is with the expert principles in an end-to-end workflow
In the previous eleven units, we learned the parts one by one: concept, visual, site analysis, plant, water, sustainability, technique, presentation, cost, legislation. In this unit we combine them all into one real project. The aim is to establish a workflow that can use artificial intelligence (AI) fluently and safely from the beginning of the project to its delivery. Workflow is the set of sequential steps that a job follows from beginning to end. This unit is not a summary, but an integration rehearsal: at each stage, where AI accelerates, where it stops, what validation comes into play.
Sample project: 4 ha city park by the stream
Let's consider the project we mentioned throughout the module from start to finish. The municipality wants to transform an idle stream bank at risk of flood into a city park that both buffers water and brings usability to the neighborhood. The budget is limited, the duration is three months, the climate is mild-rainy.
Phase 1 — Framework and concept (Units 1-2). Project framework and 6 themes are produced with AI; The team chooses the theme of "the water line that returns the stream that the city has forgotten" and translates it into principles. AI diversifies and selects teams.
Stage 2 — Field analysis (Unit 4). Slope, orientation, water flow are obtained from GIS; A field trip is made. AI synthesizes measured data + observation and produces an opportunity-constraint list. Measurements from the tool, synthesis verified.
Stage 3 — Water and climate strategy (Unit 6). YZ proposes a combination of rain garden + bioswal + permeable soil; settlement logic is established. Dimensioning is delegated to the hydraulic engineer.
Stage 4 — Plant palette (Unit 5). AI produces feature profile and candidate list; The team verifies each species for local climate zone, invasive list, toxicity and sapling availability. Unverified types are not included.
Stage 5 — Sustainability check (Unit 7). The design is passed through the water-permeability-local rate-shade-biodiversity scorecard; weak titles are improved.
Stage 6 — Visual and technical (Units 3, 8). AI produces atmospheric images and detail logic/material comparison; Scale sheets are drawn in CAD, static details are drawn by the engineer.
Stage 7 — Cost and schedule (Unit 10). AI organizes item list and table; Actual quantities and current prices are entered, and the work schedule is set up according to the planting season.
Stage 8 — Legislation and rollout (Units 9, 11). Permission checklist and institution questions are prepared, items are confirmed with the official; The presentation is designed in different languages for three audiences (municipality, neighborhood residents, jury).
Tip: At the end of each phase, ask one question: “Is this deliverable low, medium, or high risk; who will verify it?” This single question keeps the entire workflow secure.
Golden rule between stages
In the end-to-end flow, three principles are repeated at every stage: (1) AI varieties and accelerates, humans select and own; (2) each numeric/local/security-critical output is verified; (3) the decision and signature are with the competent expert. These three principles do not change; Only the subject to which they are applied changes.
three mini cases
Case 1 — End-to-end speed. One office ran a similar stream park with an AI-powered workflow: concept and analysis went from 3 weeks to 10 days; The time saved was devoted to detail and verification. Delivery quality improved because the team gained time to “decide” rather than “produce.”
Case 2 — Chain verification. In the same project, four separate AI bugs were caught at different stages: an invasive plant (Unit 5 verification), a fabricated regulatory clause (Unit 11 verification), an exaggerated image (Unit 3 check), a missing cost item (Unit 10 list). None reached submission; discipline worked.
Case 3 — Team transformation. A team made a lot of mistakes at first, thinking that AI was "magic that does everything." Once they adopted the risk bucket discipline (Unit 1), they started using AI at the right layer: free at low risk, validating at medium risk, only helpful at high risk. Efficiency and confidence increased together.
Four copyable templates
1) Project roadmap:
Your role: landscape project coordinator. Project: [description, duration, budget, climate]. Give me an end-to-end workflow roadmap: concept, analysis, water strategy, plant, sustainability, visual, technical, cost, legislation, presentation. Specify the AI's role at each stage and which verification/expert will be involved.
2) End of phase risk audit:
Check the output of the following phase [stage and output] for risk: separate low/medium/high risk parts, who and how to verify for each medium/high, list gaps that need to be closed before delivery.
3) Holistic control before delivery:
Check this project holistically before delivery: concept-design consistency, whether local plant verification has been made, whether the engineering strategy has been approved, whether the cost is realistic, whether the legislation has been confirmed, whether the visuals are representative. The missing verifications appear in a list.
4) Lesson learned:
When this project is finished, draw out the lessons learned from using AI: at which stage did AI work very well, at what stage was the risk of error high, how should I improve the workflow in the next project? Write in concrete and actionable items.
Weak prompt / Strong prompt
Weak prompt:
Do this project for me from start to finish.
This desire that wants everything at once; It produces an unverified, superficial and risky pile.
Powerful prompt:
Your role: project coordinator. Project: 4 ha park by the stream, 3 months, limited budget, mild-rainy climate, priority flood buffering and neighborhood use. A step by step workflow emerges for me; At each stage, (1) the concrete contribution of AI, (2) the verification that will come into play, and (3) the responsible expert/authority should be specified. Mark high-risk stages separately. Don't say "AI completes any stage on its own".
Progressive prompting with verification and responsibility assignment turns AI into a true coordination tool.
End-to-end role summary table
Stage
AI contribution
verification
Responsible
concept
variation
Budget/climate filter
designer
Analysis
synthesis
Field + GIS
field analyst
water/climate
Strategy
hydraulic calculator
engineer
plant
candidate list
Local verification
landscape architect
technical
Logic/text
CAD + static
Architect/engineer
Cost
Table
Actual quantity/price
quantity surveyor
legislation
Summary/list
Authorized confirmation
competent authority
presentation
Narrative
Integrity check
designer
Common mistakes
- Wanting everything in one request. The end-to-end flow is gradual; Each stage is verified separately.
- Skipping the end-of-stage risk audit. Errors are caught during phase transitions.
- Accumulating affirmations and leaving them until last. Local plant, legislation and cost confirmation should be done at its own stage.
- Embedding human decision in AI flow. Choice, ownership and signature rest with people at every stage.
- Not recording the lesson learned. Every project is an opportunity to improve the next workflow.
In summary
End-to-end workflow is where all previous units come together in a single project. AI diversifies and accelerates at every stage; but at every stage people choose, verify and own. Three immutable principles keep the flow safe: AI produces, human decides, every critical output is verified, the signature is with the competent expert. Divide the project into phases, ask the risk bucket question at the end of each phase, perform verifications in its own phase, and learn from each project. Thus, AI becomes a partner that strengthens your profession, not weakens it.
Application task
Create an end-to-end workflow with the "Project roadmap" template for a real project of your choice; Verify and assign responsibility for each stage. Then select two phases and audit them with the "End of phase risk audit" template. Finally, list your missing verifications with the "Holistic pre-delivery check" template.
checklist
- [ ] I divided the project into stages; I verified and assigned responsibility for each stage.
- [ ] I asked the risk bucket question at the end of each stage.
- [ ] I made the local plant, legislation and cost confirmation at its own stage.
- [ ] I linked high-risk stages to expert approval.
- [ ] I noted the lessons learned from the project for the next workflow.
Module Exam
1. Which of the following is the most accurate positioning for artificial intelligence in landscape architecture?
- A) AI is an accelerator and blueprint generator; Responsibility for site, ecology, security and regulatory decisions rests with humans ✔
- B) Artificial intelligence can complete and sign the design on its own
- C) Artificial intelligence only works in producing visuals, it has nothing to do with other stages of design
- D) Since artificial intelligence is more objective than humans, legislative decisions should be left to it.
Description: Artificial intelligence is an idea multiplier and draft generator that accelerates concept, visual, analysis and communication; Field reality, local ecology, safety and legal responsibility remain with the competent expert. In engineering and safety-critical decisions, AI is not a substitute for expert approval.
2. When you break down decisions on a landscaping project by risk level, which of the following would fall into the 'high risk (expert approval required)' bucket?
- A) Writing a concept text
- B) Creating a moodboard idea
- C) Retaining wall statics and children's playground safety ✔
- D) Getting a presentation title suggestion
Description: Engineering and safety-critical issues such as retaining wall statics, tree root-substructure conflict, toxic/invasive plants, and children's playground safety are high risk; Here artificial intelligence is only an assistant, the final decision belongs to the expert. Concept text and moodboard are low risk.
3. What is the most efficient way to use artificial intelligence during the concept development phase?
- A) Creating many different themes and selecting and deepening them by passing them through a context filter ✔
- B) Asking for a single 'best concept' and implementing it directly
- C) Asking for a general concept without giving context
- D) Delivering the concept without any criticism
Explanation: The human mind tends to fixate on the first 2-3 ideas. The real power of artificial intelligence is that it produces and compares many different themes; The choice, deepening and ownership belong to the human being. Asking for a single 'best' recommendation wastes that power.
4. Which principle is essential when presenting a render (image) produced with artificial intelligence to the customer?
- A) The visual is presented as a definitive and exact view to be applied.
- B) Considered a visual scale technical document
- C) Visual representation is the atmosphere; inspected for authenticity and supplied with technical sheets ✔
- D) The image is transferred directly to the application without verification
Description: Artificial intelligence visuals are a means of communication and atmosphere; It is not to scale and may contain elements that cannot be installed in reality (exaggerated density, plants unfamiliar to the climate). Therefore, it should be stated that it is a 'representative atmosphere' and should be presented with technical sheets.
5. Why is it wrong to ask artificial intelligence about the average slope of a plot in area analysis?
- A) Artificial intelligence cannot make positional measurements and can make up the number; measurement taken from GIS/field measurement ✔
- B) Artificial intelligence calculates the slope correctly but is slow
- C) Slope is an unimportant data, there is no need to ask
- D) Artificial intelligence can only calculate square meters, not slope, but calculate distance
Description: Artificial intelligence does not see the plot and cannot make spatial measurements; such a question may produce a confident but unfounded number. Measurements such as slope, area and distance are taken from GIS or field survey; Artificial intelligence is only used to interpret the measured data given.
6. What is the most critical rule of using artificial intelligence in plant selection?
- A) Use AI to candidate list and validate each species with local climate, invasive list, toxicity and market ✔
- B) Putting the list suggested by AI directly into the project, because it is based on global data
- C) Selecting only species that look visually appealing
- D) Preferring species that are practical because they grow quickly, without verifying them.
Explanation: Because AI learns from the global average, it can be wrong about climate zone, invasive status, toxicity, and market. So it is used only as a candidate list generator; each species is verified for local climate zone, invasive species list, toxicity, and sapling availability. Unverified type does not enter the project.
7. How does the role of artificial intelligence and the role of the engineer differ in stormwater management?
- A) Artificial intelligence makes both strategy and capacity calculations safely
- B) Artificial intelligence generates strategy and placement ideas; Dimensioning and capacity are engineer calculation and approval ✔
- C) The engineer only determines the strategy, artificial intelligence does the calculations
- D) Sizing is unimportant; The direction in which the water flows is sufficient
Description: The idea and placement of solutions such as rain gardens and bioswales are design decisions, and artificial intelligence helps here. However, sizing, capacity and flood safety require hydraulic calculations and engineer approval; Artificial intelligence cannot calculate flow rate/capacity because incorrect sizing means flood or collapse.
8. What is the most accurate approach when evaluating the sustainability of a design?
- A) The greener the design looks, the more sustainable it is
- B) Evaluating sustainability in measurable topics such as water, permeability, local rate, shade, maintenance and biodiversity ✔
- C) Just increasing the number of plants is enough
- D) Intensive planting without taking into account maintenance costs
Explanation: 'Looking green' is not sustainability; A single-species green space that consumes a lot of water can be ecologically poor and costly. The right approach is to score sustainability on measurable topics such as water, permeability, native species rate, shade, maintenance and biodiversity and improve weak points.
9. What should be done for the measurement and reinforcement values in the 'sample section' of a retaining wall given by artificial intelligence at the technical drawing stage?
- A) Values can be copied directly into the project because AI knows the standards
- B) Values are representative; Static detail is determined by ground survey and engineer calculation, it is not included in the drawing without verification ✔
- C) Values are for aesthetic purposes only, they have nothing to do with statics
- D) The equipment value provided by artificial intelligence always remains on the safe side
Explanation: Artificial intelligence does not produce scaled/static drawings and the measurements, elevations and reinforcement values it provides are representative. Static detail; It is determined by ground survey and engineer calculations. These values cannot be transferred to the drawing without verification, this is especially vital for security-critical details.
10. What is the most effective approach when presenting the same landscaping project to different audiences?
- A) Presenting the same technical text to everyone
- B) Exaggerating savings and benefit figures to increase impact
- C) Adapting the design essence to tell each audience what they care about in their own language ✔
- D) It is sufficient to prepare a presentation only for the technical jury.
Description: The investor cares about cost and value, the neighborhood cares about daily benefits and safety, the jury cares about concept and technique. Explaining a single design essence to every audience what they care about, in a language they can understand; Taking advantage of the frame-translating power of artificial intelligence is the most effective way. It is a common mistake to explain it to everyone in the same language.
11. Why is it wrong to have artificial intelligence directly calculate the total cost of a park?
- A) Artificial intelligence always knows the current market prices
- B) Artificial intelligence calculates the cost but is wrong only on large projects
- C) Artificial intelligence does not know the actual length of the project and current prices; The amount given is representative, not an offer ✔
- D) Cost is unimportant, what matters is the beauty of the design
Explanation: Artificial intelligence neither knows the actual length of your project nor the current local unit prices; The amount given is purely representative and not an offer. The correct way is to take the actual quantity from the project data and the unit prices from the market and have the artificial intelligence only edit the table.
12. What is the most dangerous risk and correct precaution when using artificial intelligence regarding legislation?
- A) Since artificial intelligence knows the legislation by heart, its articles can be trusted
- B) Legislation is universal, there is no need for local confirmation
- C) The item number given by artificial intelligence is always up to date.
- D) Artificial intelligence can make up non-existent matter; Every provision must be confirmed from the official up-to-date source and authority ✔
Description: Legislation is local, current and open to interpretation; AI may make up a non-existent item number or rule (hallucination). The correct precaution is not to use any provision without verifying it; It is to use artificial intelligence only to summarize the official text you provide, to generate a checklist and questions to the official, and to confirm the items from the official current source.
13. Which principle applies to the use of artificial intelligence with data such as customer contract and parcel information?
- A) All documents can be uploaded as is, because speed is important
- B) Contract data is unimportant for AI, it is ignored
- C) Parcel information can be shared freely as it is not considered personal data.
- D) Confidential and personal data are not shared without control; Used anonymized if necessary ✔
Explanation: Customer identity, parcel information, contract and personal data should not be uploaded to artificial intelligence tools in an uncontrolled manner. If necessary, data is shared by anonymization. Confidentiality is one of the key ethical principles emphasized throughout the module.
14. What are the three constant principles when running a project end-to-end with artificial intelligence?
- A) Request everything in one request, verify the result, deliver
- B) Artificial intelligence decides, humans only implement
- C) Accumulate verifications and do them all at once at the end of the project
- D) Artificial intelligence produces, humans select and own; every critical output is verified; Decision and signature belong to the expert ✔
Description: In an end-to-end workflow, three principles are repeated at every stage: AI types and accelerates, humans choose and own; every numeric, local, and security-critical output is verified; The decision and signature belong to the competent expert. These principles remain constant even if only the subject to which they are applied changes.