Chat Workflows — Help Center

How to ask better questions, verify sources, correct facts, and keep long threads useful.

How do guided workflows work?

Guided workflows are step-by-step due-diligence walkthroughs inside project chat. Use them when you want a structured path instead of deciding every prompt yourself.

Entry points:

• Start a new chat and choose a workflow card when the empty-chat state appears.
• Use the checklist button in the message bar to open the workflow picker later.
• After each answer, use the next-step chips when you want the agent to continue the selected workflow.

Current workflow paths:

Best practice: run one workflow at a time, review citations after each major answer, and upload project-specific studies before starting workflow steps that ask the agent to review consultant reports.

How does the Screening & Risk Analysis workflow work?

The Screening & Risk Analysis workflow screens a site for what could kill the deal before you spend on it, then closes by generating a full feasibility assessment that weighs the risks it surfaced. Launch it from the workflow picker in Project Chat or from your project briefing.

It runs in six steps, worst-risk-first:

1. Use & Fatal Flaws — is what you want to build permitted in your zoning district, and is anything on this site an outright deal-killer? If the use is not by right, the agent names the realistic path (conditional or special use, the most conservative rezone target, or a planned unit development).
2. Flood & Water — which FEMA flood zones touch the parcel, whether a special flood hazard area or floodway affects the build, the base flood elevation if one exists, and what flood status changes about design, insurance, and approvals.
3. Wetlands, Soils & Slope — mapped wetlands, hydric soils, high water table, and steep slopes, and what each one limits about where you can build and what it adds to cost.
4. Access & Roads — street frontage, who controls access, whether a driveway or DOT permit is needed, and — for a single point of access — whether that is an emergency-access concern at the scale you described.
5. Utilities & Providers — which utilities actually reach the site, who serves each one, what is confirmed from real data versus assumed, and how to reach each provider to verify.
6. Approvals & Feasibility — the permits and approvals the project needs and how long each typically takes to review, then a full feasibility assessment for the site (factoring in any site studies you have uploaded) as the deliverable.

How to read the results:

• Each step is a screening pass, not a confirmed finding. Values drawn from public data (flood, wetlands, soils, utilities) are screening-grade — verify surveyed elevations, wetland delineation, and provider service commitments with the agency or provider before relying on them.
• The workflow does not replace the deal-killer judgment — it surfaces the risks so you can decide. When an answer sits on a decision fork, the agent names the alternative before answering.
• The closing step produces a Feasibility Assessment deliverable — a three-tier risk view (deterministic site risks, use-case feasibility, and an alternative-use comparison) — that is versioned and downloadable as DOCX.

How does the Site Cost Estimate workflow work?

The Site Cost Estimate workflow builds a screening-level opinion of probable cost for horizontal site development, then hands you an editable Excel cost workbook. Launch it from the workflow picker in Project Chat or from your project briefing.

It runs in six steps, smallest scope to largest:

1. Scope & Program — agree on the program being priced (use, building or lot scale, and whether existing structures come down) before any pricing.
2. Site Cost Drivers — surface the terrain, soils, stormwater, utility proximity, and frontage conditions that push costs up or down.
3. Earthwork & Grading — rough cut and fill, topsoil, erosion control, and detention-basin excavation.
4. Wet Utilities — sanitary, water, and storm mains or service connections, run lengths, tie-ins, and fire protection where triggered.
5. Paving & Access — internal streets or the parking field, curb and gutter, driveway aprons, markings, and accessible ramps.
6. Cost Summary — everything rolled up into itemized earthwork, utility, and paving tables with mobilization and a visible contingency, then the downloadable workbook.

How to read the numbers:

• This is a Class 5 conceptual estimate (AACE analog) with an expected accuracy range of -30% / +75%. Treat it as an order-of-magnitude screen, not a bid.
• Quantities are parametric screening values. Each line item states whether it was measured from your site, derived from code requirements, or a screening default — and lists what was excluded and why. Verify every quantity against an engineered grading, utility, and paving design before relying on it.
• The contingency is visible and editable in the workbook, and the underlying cells use live formulas so your team can adjust and extend the estimate.

Use your own pricing: the estimate ships with national baseline unit prices adjusted for your state. To reprice against your local costs, paste or upload a unit-price list in chat and ask the agent to apply it — your prices override the baseline while the quantities stay the same.

Can I ask the agent how to use PrecisionDocs?

Yes. Ask the project chat for PrecisionDocs workflow help when you need a product SOP, a control location, or the right sequence for a task.

Good use cases:

• How do I export GIS layers for Civil 3D?
• Where do I check whether bootstrap is still waiting on the county?
• What is the safest workflow for correcting zoning?
• How should I troubleshoot a missing wetland or soil layer?
• How does Email my agent routing work?

The agent can search the same Help Center content that appears on the public and in-app help pages, then summarize the article that best matches your question.

Use project chat for workflow guidance, but use the LifeBuoy/support path when you need the PrecisionDocs team to correct a source, investigate a failed bootstrap, or review account-specific billing state.

How should I ask questions to get the best answers?

The agent performs best when you ask task-shaped questions instead of broad prompts.

Good pattern:

1. Name the subject: parcel, report, ordinance topic, or document.
2. State the outcome you want: summary, comparison, risks, checklist, or draft language.
3. Add the scope: one parcel, one uploaded report, one code section, or one engineering topic.
4. Ask for sources when the answer matters operationally.

Examples:

• Summarize the floodplain constraints for this site and cite the source data.
• Search the uploaded geotech report for groundwater depth and bearing recommendations.
• What ordinance sections control parking, setbacks, and stormwater for this parcel?
• Draft a client-ready risk summary for the Phase I ESA findings.

Weak prompt example: Tell me everything about this project.

Strong prompt example: Give me the top 5 development constraints for this site with citations and next steps.

How do I check sources and citations?

Use citations as your first quality check.

What to look for:

• Inline citation markers should appear with factual claims.
• Government API and ordinance sources are the strongest references for regulatory answers.
• Uploaded documents are strong when they come from your consultants or jurisdictional correspondence.
• Web results are supplementary and should be verified against the official record if the finding affects design, permitting, or fee exposure.
• User-corrected facts (e.g., a zoning override you set) are flagged so you know the agent is using your value, not the assessor's.

Recommended review SOP:

1. Click the citation badge or citation panel entry.
2. Confirm the source type: ordinance, API, parcel data, document, web, or user-corrected.
3. Compare the claim to the cited text or map output.
4. If two sources conflict, keep both in view and escalate the discrepancy instead of forcing a conclusion.

If you are preparing a client deliverable, verify every major constraint directly from the linked source before sending it out.

What should I do when the agent cannot find information?

Treat a "not found" answer as a source-routing problem first, not as the final answer.

Recovery sequence:

1. Name the source set. Ask again using one source class: ordinance, uploaded documents, parcel data, government APIs, GIS layers, or web search.
2. Ask for a narrower target. Replace "site constraints" with "parking count", "front setback", "FIRM panel", "NWI wetland class", or "stormwater detention trigger".
3. Check readiness. Open the project home and confirm bootstrap, supplemental ingests, and uploaded files are finished before relying on the answer.
4. Add the missing source. Upload the consultant report, use Add ordinance source, paste a URL for the agent to ingest, or ask it to find and index a separate zoning/LDC/subdivision code product.
5. Ask for the evidence path. Show which sources you searched and what was missing. This helps distinguish "not in corpus" from "source exists but the prompt was too broad."

Good recovery prompts:

• Search only the uploaded ALTA survey for access easements and tell me if the answer is not in the survey.
• If the zoning standards are in a separate zoning code, find and index that source before answering.
• Use official FEMA/NWI/USDA sources only and tell me which dataset returned no data.

How do I correct a parcel's zoning when the assessor data is wrong?

Regrid pulls zoning from the county assessor record, which is sometimes stale or incorrectly classified for the actual ordinance designation. There are two correction paths and both produce the same provenance trail.

Path 1 — UI pencil edit (owner only):

1. Open the project home and find the parcel on the map card.
2. Click the small pencil icon next to the zoning chip.
3. Type the correct designation and press Enter (Escape to cancel).

Path 2 — chat (any project member, gated by ordinance verification):

• Tell the agent the corrected designation directly: This parcel is actually zoned R-2, not R-1 — verify against the ordinance.
• The verify_zoning tool will check the ordinance text. If it confirms your assertion, the override is auto-written by the agent on your behalf (still owner-scoped at the database layer — non-owners silently fail rather than mutate).

Either path:

• Replaces the assessor value with yours in all subsequent parcel lookups, reports, and feasibility assessments.
• Renders the parcel chip as a cyan "User-corrected from {original}" pill (hover for the original assessor value).
• Writes provenance to the project record (who, when, original value) for audit.

Are there conversation limits or long-thread behaviors I should know about?

The chat is designed to preserve continuity, but long threads are managed automatically.

What to know:

• The agent can perform multiple lookups per question when gathering evidence.
• Recent conversation context is prioritized so answers stay relevant.
• Older conversation history is summarized automatically so the thread can continue without losing important context.
• If you are revisiting a decision from earlier in the project, restate the key fact or ask the agent to recap it first.

Best practice for long projects:

1. Keep each thread focused on one decision stream when possible.
2. When you pivot topics, start with a short summary of the context you want carried forward.
3. If the thread has become noisy, clear chat history and start a new clean conversation.

This keeps answers tighter and reduces the chance of stale assumptions carrying forward.

Can I stop a response, clear a thread, or restart a line of work?

Yes.

• Use the Cancel control while the assistant is still streaming if you already have what you need or the answer is headed in the wrong direction.
• Use Clear chat history from the project chat menu when you want a fresh thread without deleting the project, documents, or outputs.
• If you want to restart a line of work, begin the new prompt with a short statement of the decision or source set you want the agent to use.

Good reset prompt example:

Start over. Only use the uploaded geotech report and the FEMA citations, and give me a construction-risk summary.

Other help topics

• Getting Started
• Email my agent
• API & connectors
• Collaboration
• Tools & Data
• Bootstrap & Ordinances
• Documents & Reports
• Credits & Usage
• Billing & Subscriptions
• Account & Settings
• Troubleshooting

• All help topics