a Sagan program ยท currently in private beta

Landscape Design QA Agent

Your design submission gets instant QA feedback against the checklist and past approvals in under two minutes, so reps fix missing dimensions or material callouts before the production manager opens the queue.
before

Rep submits design through JotForm. Production manager reviews it in his admin time between field work. 3-6 days later, feedback goes back. Project can't be scheduled. Customer calls. Rep is stalled.

after

Rep submits design through JotForm. Agent screens it in under 2 minutes against your QA rules and past approvals. Rep sees what's missing and fixes it or force-submits. Production manager opens the queue and sees only pre-screened work. Scheduling starts without waiting.

landscaping / home services / residential construction / quality assurance / design review / operations / JotForm / Slack / Housecall Pro / OpenRouter / SQLite /  landscaping / home services / residential construction / quality assurance / design review / operations / JotForm / Slack / Housecall Pro / OpenRouter / SQLite / 
the problem

Design QA takes 3-6 days. Your reps stop selling while they wait.

Landscape sales reps finish the sale, hand-draw the design on graph paper, then submit through JotForm.

01
Today's lag

3-6 days from rep submission to QA feedback

02
The blocker

One production manager reviews every design in his spare time

03
Downstream impact

Scheduling can't start, customers call, reps can't move to the next sale

the math, if you want to look

Instant pre-screen feedback before the production manager ever opens the queue.

proof 01
Feedback speed

Under 2 minutes from submission to agent verdict, not 3-6 days

proof 02
What gets caught

Missing dimensions, material callouts, measurement gaps, and other checklist items your team defines

proof 03
Rep behavior

Reps fix obvious issues upstream so the production manager only reviews pre-screened work

proof 04
Production manager focus

Shifts from routine checks to edge cases and judgment calls

The agent reads the sketch photo and form fields the moment the rep submits. It checks against your QA rules and past-approval examples, flags exactly what's missing (dimensions, material callouts, measurements), and returns feedback in under two minutes. Reps fix issues before the production manager sees the submission. The production manager sees only pre-cleaned work and can focus on edge cases instead of routine checks.

how it works

How the agent screens every submission.

The agent sits between the rep's JotForm submission and the production manager's queue. It reads the design sketch and form data, checks against your rules and past approvals, and flags what's missing before the production manager ever sees it.

step 01
Rep submits design sketch and form fields

Sales rep uploads the hand-drawn design photo, job ID, project type, square footage, materials, and notes through JotForm.

step 02
Agent reviews against rules and past approvals

The agent reads the sketch using vision AI, checks it against your QA checklist and past-approval examples, and identifies what's missing or wrong.

step 03
Agent returns instant feedback

Within 1-2 minutes, the rep sees exactly what's flagged: missing dimensions, material callouts, measurement gaps, or other checklist items.

step 04
Rep fixes or force-submits

Rep uploads a corrected version and re-runs the agent, or force-submits if they disagree. Either way, the production manager sees the agent's pre-screen attached.

step 05
Production manager reviews pre-screened work

The production manager opens the QA queue and sees only submissions the agent has already vetted, with flags and overrides logged.

ai agent · estimator console inputs transform outputs public preview
inputs
Design sketch photo +

Hand-drawn 2D layout on graph paper with color-coding, dimensions, and material callouts

Form fields +

Job ID, project type, square footage, materials, and notes from the JotForm submission

QA rules +

Plain-language checklist maintained by your team (e.g., square footage labeled, material callouts present, dimensions readable)

Past-approval corpus +

20-30 example chains of rep submission, production manager feedback, and final approved design

transformation
Vision AI review +

Agent reads the sketch photo using a state-of-the-art vision model to extract visible elements, dimensions, and annotations

Rules matching +

Agent checks each visible element against your QA checklist and flags what's missing or unclear

Corpus comparison +

Agent compares the submission to similar past-approved designs to ground its judgment in your team's actual standards

Confidence scoring +

Agent assigns a confidence score to each verdict so you can track when auto-approval becomes safe

outputs
Agent verdict +

Pass all checks, or list of specific flags (missing dimensions, material callouts, measurement gaps, etc.)

Feedback to rep +

Actionable list of what's missing so the rep can fix it before resubmitting

Pre-screen attached to production manager notification +

Slack message or Housecall Pro note showing agent verdict, any force-overrides, and confidence score

Submission record +

Logged in your dashboard with agent verdict, rep response, and production manager's final feedback for metrics and future training

draft ready for estimator review _
tech used
JotForm webhook integrationOpenRouter vision LLMSlack APISQLite
tool alternatives
Any form platform with webhook support (Typeform, Gravity Forms) in place of JotFormAny LLM with vision capability (Claude, GPT-4V, Gemini) via OpenRouter
honest qualification

Is this for you?

built for you if
  • + Landscape contractors with a single QA reviewer - If one person is the bottleneck on design approval, this agent moves the routine checks upstream so that person focuses on edge cases.
  • + Teams with hand-drawn or paper-based design workflows - The agent reads sketch photos and form fields, not CAD files or digital design tools. Works with your current process.
  • + Franchises or multi-location operators - Rules are maintained in plain language and applied consistently across all locations. Corpus grows as you accumulate approvals.
  • + Companies ready to measure QA throughput - The agent logs every submission, verdict, override, and production manager correction so you can track where the bottleneck moves next.
not for you if
  • - Teams without a clear QA checklist - The agent works best when you can articulate what 'good' looks like in rules or examples. If QA is entirely subjective or changes per project, the agent will struggle.
  • - Fully digital design workflows - If your reps submit CAD files or use design software, this agent is built for photo-based sketches. A different tool may fit better.
  • - Companies that want to remove the production manager entirely - This agent pre-screens submissions and flags issues. It does not replace human judgment. The production manager stays in the loop for v1.
pricing

Scoped build plus usage-based runs.

to build

The agent is built to your specifications and deployed on Railway. You pay for the initial build and setup. After launch, you pay per submission the agent reviews.

then
  • Initial build includes prototype, corpus upload, rules editor, and admin dashboard.
  • Usage pricing covers the vision AI calls to OpenRouter and the agent's reasoning per submission.
  • No per-user or per-rep fees. Pricing scales with submission volume.
FAQ
How fast does the agent review a design submission?

The agent screens each submission in under 2 minutes. The rep uploads the sketch photo and form fields through JotForm, and within 1-2 minutes receives feedback on what's missing or flagged against your QA checklist and past approvals. This replaces the current 3-6 day wait for the production manager to review.

What does the agent actually check on each design?

The agent verifies your QA checklist items: dimensions labeled, material callouts present, measurements readable, ingress and egress marked, and any other rules your team defines. It reads the hand-drawn sketch photo using vision AI and compares it against your past-approved designs to ground its judgment in your team's actual standards, not generic rules.

What happens if the rep disagrees with the agent's feedback?

The rep can force-submit the design anyway. The agent's feedback and the override are both logged and attached to the production manager's notification, so the production manager sees exactly what the agent flagged and that the rep chose to proceed. This keeps the rep from getting stuck while building the dataset to improve the agent over time.

Does the agent replace the production manager's review?

No. The agent pre-screens submissions and flags issues before they reach the production manager. The production manager still reviews every design, but now sees only pre-screened work and can focus on edge cases and judgment calls instead of routine checks. This frees the production manager to handle the harder decisions while reps fix obvious issues upstream.

How does the agent learn what good design looks like for your company?

You provide 20-30 example chains of past submissions, the production manager's feedback, and the final approved designs. The agent uses these examples as few-shot context so its judgment matches your team's actual standards. As you accumulate more submissions and feedback, the agent grounds itself in your specific business practices, not generic rules.

Can you change the QA rules without a developer?

Yes. You maintain the QA checklist in plain language through the admin dashboard. When you update the rules, they become active immediately for the next submission. No code changes or developer involvement needed. The system tracks which version of the rules was active for each submission so you can measure whether rule changes improve outcomes.

What data does the system track to measure if it's working?

The system logs every submission, the agent's verdict, whether the rep force-submitted, and where the production manager's final feedback differed from the agent's flags. You can see submission volume by week, agent verdict distribution, how long designs wait before review, and the override rate. This data shows where the bottleneck moves next and when the agent's confidence is high enough to consider auto-approval.

What file formats does the agent accept for design submissions?

The agent reads hand-drawn design sketches submitted as photos (graph paper with color-coding, dimensions, and material callouts) plus the form fields from your submission form (job ID, project type, square footage, materials, notes). It is built for photo-based sketches, not CAD files or digital design software exports.

related builds
roofquote-ai-92862583self-storage-sourcing-agent-prd-v2-0-6e730be6rent-roll-parser-agent-prd-v1-0-a6644eb0deckclose-21664542bidscout-3dfbd88a
next step

Stop waiting 3-6 days for design QA.

The agent screens every submission in under 2 minutes against your rules and past approvals. Your production manager sees only pre-screened work. Your reps fix issues before the queue. Scheduling starts without waiting on one human.