01 — Sign In to Begin
Reviewer Sign-In
ConfidentialSign in with your HNTB email.
We'll email you a one-time access code. No passwords.
Check your email.
We sent a one-time access code to your inbox. Enter it below to sign in.
Don't see it? Check your Junk or Spam folder. If sign-in codes consistently land there, add the sender to your Outlook Safe Senders list.
01 — Document Intake
Upload Document
Confidential01 — Activity
Welcome back.
Your AI-assisted document reviews and downloads, in one place.
My Past Reviews
About this tool
An AI-assisted first pass for environmental document review.
Drop in an EA, EIS, CatEx, or Section 4(f) evaluation and the tool reads it the way a senior reviewer would — checking citations against the current CFR, validating the math in tables, comparing the document’s structure to WisDOT FDM 20-30, flagging I-794 Style Guide hits, and replying to existing reviewer comments where it can. You get the file back as a Word doc with comments anchored to the source text, ready to backcheck. The intent isn’t to replace reviewers — it’s to handle the mechanical sweep so reviewers can spend their time on judgment.
Built by
HNTB Wisconsin’s AI Cloud Infrastructure Team.
The Environmental QC Tool was designed, built, and deployed in-house by HNTB Wisconsin’s AI Cloud Infrastructure Team — the same group operating IHSDAdam, CapScore, the I-41 MKE Corridor Dashboard, SNAT, and a growing portfolio of secure, document-aware AI workflows that run inside HNTB’s authentication wall and stay inside HNTB’s AWS account. Questions, feedback, or want to onboard your team? Reach out to your HNTB Wisconsin Group Director.
01 How a review works
From upload to annotated Word doc in about two minutes.
Every review follows the same four steps.
1
Upload
Your document goes straight to a private, encrypted folder inside HNTB’s AWS account. It never leaves.
2
Extract
The tool pulls out citations, walks every table, builds a list of headings, runs the I-794 style scan, and lifts any existing reviewer comments.
3
Review
Five focused AI checks run in parallel, each with access to FDM 20-30 and the I-794 Style Guide as reference material.
4
Annotate
Findings come back as native Word comments anchored on the source text. Existing reviewer comments are preserved; AI replies are threaded underneath.
02 The six checks
Six focused checks per review.
Anything that can be verified directly — a citation against the live CFR, a column sum against a stated total, a heading against the FDM 20-30 required-section list — gets verified before the AI sees it. That’s the ground truth the AI works from. Reviewers can turn any of the six off before upload if a check doesn’t apply to the document type.
General Grammar / AP Style
A general copy-edit pass on the prose itself — spelling, comma usage, sentence structure, AP Stylebook conventions (numerals, dates, states, percent). Comments come from a distinct author, HNTB AI Spellcheck/Grammar, so reviewers can color-filter copy edits separately from substantive QC. Where the I-794 Style Guide takes a position, Document Style wins; Grammar fills the gaps.
Document Style & Cross-References
I-794 Style Guide rules — highway naming (“I-XX” not “IH-XX”), acronyms, plain-language preferences, Oxford-comma rule, duplicate stop-words. Every in-text reference to a Table, Figure, Section, or Appendix is checked against the document’s actual headings and captions. Repeated style hits collapse into a single doc-wide comment with the occurrence count.
Table Integrity
For any table with a Total or Subtotal row, the column sum is computed and compared to the stated total within a 1% tolerance for rounding. Mixed units inside a single column, missing captions, footnote callouts without a matching footnote, and empty cells in numeric columns are flagged separately.
FDM Compliance
Each of the 15 sections required by WisDOT FDM 20-30-10.2 is checked for presence and reviewed against the FDM-prescribed content for that section — e.g., Cover Page must include title, route, counties, and an abstract; Purpose and Need must state project objectives. Near-synonym headings count (“Affected Environment” ≈ “Existing Conditions”).
Citation Verification
Every CFR, USC, and Public Law citation in the document is checked against the live eCFR for the cited title’s current snapshot. Verified citations are stamped with the current section heading; rescinded or mistyped citations are flagged.
Reviewer Comment Backcheck
Existing reviewer comments are read in context and get a threaded reply — Agree (concern still applies), Disagree (already addressed in the doc, with a pointer), or Uncertain (needs project knowledge the AI doesn’t have). Every reply is hedged and ends with a recommendation that a human reviewer confirm before resolving.
03 Security & confidentiality
Documents stay inside HNTB.
Reviewers upload draft EAs, EISs, and Section 4(f) evaluations — documents that aren’t yet public, sometimes contain client-privileged content, and reference uncleared third-party data. Everything below is built on the assumption that any one of those leaking is a real problem.
-
HNTB email only, passwordless sign-in
Sign-in is restricted to
@hntb.comaddresses. We email you a one-time code; sessions last 8 hours so you don’t re-auth between morning and afternoon. -
Your reviews are yours
Every upload is stored in a per-user folder. Even with the upload key, no one else can read your reviews — the system checks your sign-in identity against the folder before serving anything back.
-
Encrypted at rest, never public
Storage is fully encrypted and configured to refuse any public-access setting. Versioning is on, so accidental deletes are recoverable.
-
Not used to train AI
The AI services we use run under enterprise terms. Your documents are not retained or used to train any model.
-
Sensible upload limits
Each reviewer can submit up to 60 documents per hour. Plenty of headroom for normal use, with a safety net against runaway scripts.
-
No system internals leak
If something goes wrong in processing, you see a friendly retry message. Diagnostic detail stays internal and is not exposed in the dashboard.
04 Under the hood
Built on AWS, behind HNTB sign-in.
The tool runs entirely inside HNTB’s AWS account — uploads, processing, and annotated outputs all live there and nowhere else. Authentication is HNTB email only, sessions are short-lived, and access is scoped per reviewer. The six AI checks run on a focused workflow with structured outputs and reference material drawn from WisDOT FDM 20-30, the I-794 Lake Interchange Study Style Guide, and a curated AP Stylebook reference. Citations are verified against the live eCFR on every run.