The State of AI in Environmental Consulting
How environmental consultants are putting AI to work, and what separates the leaders from everyone else. A field-wide benchmark: real adoption numbers, the workflows being rewritten, the liability stalling firms, and the maturity model that puts every firm on the curve.
Built by a team from
Built by a team from: Stanford Doerr School of Sustainability, Stanford School of Earth, Energy & Environmental Sciences, Harvard University, Morgan Stanley
Find us at
Find us at: Battelle Chlorinated, NSCW, A&WMA ACE, BOMA, NAGGL, EPSS, EBA, Texas Environmental Superconference, America East SBA, GEC, GRA Western Groundwater, RemTEC, AEHS East, EBI Pac NW Summit, SERDP & ESTCP, NEBS
AI has arrived in environmental consulting. Most firms just haven't organized around it yet.
Adoption is real and the returns are documented, but it's uneven, ungoverned, and concentrated in a handful of firms. The biggest are already building AI in-house. For everyone else, the edge is a partner they can trust. Start with four numbers.
time to draft a proposal at firms running their own AI tools, down to a tenth of what it used to take.
Global consultancy benchmarks, 2026
of a typical $10,000 Phase I ESA is labor. That is the cost AI is now pushing toward roughly $1,000 a site.
Industry expert interviews, 2026
recovered when AI read 300 old site reports and flagged a contamination release everyone had missed.
Atlas field research, 2026
of firms qualify as “skilled practitioners,” with AI built into the daily work. Adoption is wide but shallow.
AEC industry surveys, 2026
Where AI lands first
This is already happening. These are the environmental workflows AI is rewriting now, where the time savings are documented and the EP still controls the final product.
Phase I ESA & impact-assessment drafting
AI works through the site data (flora, fauna, wind velocity) and drafts the report. The Environmental Professional reviews and signs.
Records review & database search
AI reads tens of thousands of pages of data-room documents into validated assumption sheets and red-flag lists.
Proposal generation & SOWs
Firms running dozens of in-house AI tools have cut proposal time to a fraction of what it took.
Data analysis & scientific claims
It works through tens of millions of data points at once, for faster, more defensible regulated reporting.
Historical document review
Pull site history, ownership, and prior uses out of decades of scanned records, maps, and reports.
Field notes & meeting capture
Capture site-walk and kickoff notes, then turn them into agendas, summaries, and action items.
Time-savings figures: generAIt Solutions, Build.inc, and vendor benchmarks for AI-assisted Phase I ESA workflows.
The AI-Native Maturity Model
Every firm sits somewhere on the curve from AI-curious to AI-Native, measured across five dimensions. Click a stage to see what it looks like in practice.
Where firms are today
Atlas synthesis · 2026 industry dataAnchored to 2026 industry data, ~37% piloting, ~38% scaling, under 5% “skilled practitioners.” The curve's right side is where the competitive separation is happening.
Stage 3: Operationalizing
AI in real workflows, with the first guardrails in place.Workflow Integration
AI is the standard first pass on a core deliverable, with a human finishing it.
Tooling & Infrastructure
Sanctioned firm accounts; some tools connected to firm data.
Governance, QA & Liability
A written AI-use policy; defined human review; EP sign-off preserved.
Talent & Culture
A named owner; basic training; leadership endorses it.
Client & Commercial Model
Some fixed-fee or productized work; AI capability mentioned in pitches.
Where does your firm actually land? Score yourself across all five dimensions and see your stage on the curve.
Take the assessment →Two forces keep good firms stuck.
The technology works. What stalls adoption is the economics and the liability, and both are specific to this field in ways most AI commentary misses.
Professional liability & defensibility
Deliverables are EP-signed and tied to ASTM E1527-21. AI can't carry that liability, and malpractice claims involving AI just hit a five-year high.
The billable-hour disincentive
Bill faster and you bill less. Leaders are moving to fixed-fee and value-based pricing, one global firm is taking fixed-price work from 50% to 60% of its mix, so efficiency turns into margin instead of lost revenue.
Accuracy, hallucination & “AI washing”
A wrong REC or a single unit slip carries real legal weight. Recent filings show automated errors off by 1,000× (a mercury threshold) and 1,000,000× (a carbon figure), the kind of mistake that forces a refiling, not a shrug.
The data-quality problem
About 75% of firms still keep core data in spreadsheets, and AI can't work on what it can't read. The ones getting results cleaned up their data first.
Vague insurance & client acceptance
Most liability policies say nothing about AI, and the norms for what clients will accept, and what you have to disclose, are still being written.
Data security & confidentiality
Sensitive site and client data can't leak into consumer tools.
Skills, time & buy-in
No one owns it, no one is trained, and leadership hasn't committed.
What AI-Native firms do differently
The firms in front aren't using better chatbots. They've changed how the work, the pricing, and the team are organized.
They redesign the work instead of bolting AI on
They rebuild the workflow around AI from the ground up. People move up to judgment, QA, and the client relationship.
They make governance a product
Traceability, audit trails, and defensibility become something clients pay for, not a compliance afterthought.
They break the billable-hour trap
Pricing moves to fixed-fee, value, and subscription, so efficiency becomes margin instead of lost revenue.
They turn their archive into an edge
Years of reports and site data become an advantage no competitor or off-the-shelf tool can copy.
They make AI fluency the baseline
Role-based training and AI-literate hiring make the skill part of the firm, not a few enthusiasts.
What the field's leaders are saying
Anonymized public statements from executives at leading North American environmental, engineering, and infrastructure firms, plus practitioners we spoke with about how AI is moving through the field.
“AI and the software enablement that we have seen is allowing us to do more, not less, with more people as well, with the use of AI.”
“I wouldn't characterize this as a renegotiation. I'd characterize it as our clients trying to seek out ways that they can employ us to deploy something that is more valuable.”
“We have over 600 data engineers working on specific AI agents, deploying AI to influence commercial decisions early in the process.”
“Our specialists bring decades of field insight and AI amplifies it, turning complex datasets into clear, actionable intelligence.”
“We continued to scale our AI capabilities, providing access to AI agentic tools and models to all employees.”
“Technology isn't the risk, what we ask it to do is the risk.”
“AI can make garbage data look compelling, beautiful visualizations and seemingly logical outputs over bad data.”
“We're inviting environmental consulting leaders to add their voice to the next edition. Get in touch to take part.”
Find out where your firm stands, then move up the curve.
Take the 2-minute assessment for your firm's stage, where you stand against the field, and the three moves to the next level. Atlas builds the governed AI agents, from Phase I ESAs to records review and reporting, that get you there.