# Geotech Engineering Services + AI: A Partner Guide

**By Dan Cumberland** · Published May 13, 2026 · Categories: AI Strategy

> Geotech engineering services carry a kind of liability most other AEC partners don't: a personal seal that no AI vendor will ever stand behind, plus a 2026...

## Why Your Geotech Is Reacting The Way They Are

Geotech engineering services carry a kind of liability most other AEC partners don't: a personal seal that no AI vendor will ever stand behind, plus a 2026 wave of professional indemnity exclusions that have shifted the cost of getting AI wrong directly onto the engineer\.

According to Risk Specialty Group[3](/blog/blog-geotech-engineering-services#ref-3), "professional judgment, responsibility, and liability remain firmly with the engineer" regardless of whether AI touched the deliverable\.  The vendor accepts none of it\.  And the carrier sitting behind the engineer is now actively writing AI out of coverage: Verisk's AI exclusion endorsement forms went into effect January 1, 2026[3](/blog/blog-geotech-engineering-services#ref-3)\.  Howden's 2026 claims\-risk report[4](/blog/blog-geotech-engineering-services#ref-4) confirms the pattern globally, noting professional indemnity insurers "are cautious about what the longer\-term impacts will be on risk and liability\."

That's the business reality\.  The professional reality is older and runs deeper\.  ASCE's Code of Ethics[5](/blog/blog-geotech-engineering-services#ref-5) requires civil engineers to "hold paramount the safety, health, and welfare of the public" and to "perform services only in areas of their competence\."  An engineer who can't yet articulate where AI fits inside that competence boundary has a duty to be careful\.  That's not Luddism\.  That's the seal doing its job\.

**What your geotech is actually carrying:**

- A personal seal with personal professional liability — the engineer signs, the engineer owns it
- An ASCE ethical obligation to public safety and within\-competence practice
- Hallucination risk that's already material — Construction Dive reports[6](/blog/blog-geotech-engineering-services#ref-6) that "chatbots can create authoritative\-sounding reports based on flawed inferences, which can lead to liabilities that outlive a project," and AI\-generated fake geotechnical research papers are now appearing in academic search engines
- Direct PI exposure from carriers writing AI exclusions in 2026

Professional judgment, responsibility, and liability for AI\-influenced engineering work remain firmly with the engineer who signs and seals the deliverable\.  That sentence is the whole reason the conversation matters\.

That caution is real\.  It also doesn't mean nothing's happening — because something almost certainly is\.

## What Your Geotech Engineering Services Partner Is Probably Already Doing

Roughly half of geotechnical engineering professionals reported using AI in their work in 2025, and the use cases span from low\-risk data extraction to LLM\-drafted specifications that get signed and sealed\.

Ground Engineering's 2025 international survey of 1,000\+ geotechnical professionals[7](/blog/blog-geotech-engineering-services#ref-7) found that 51% reported AI use on the job — up from 19% in 2020\.  Adoption has nearly tripled in five years\.  The same coverage[7](/blog/blog-geotech-engineering-services#ref-7) notes that LLMs like ChatGPT, Claude, and Microsoft Copilot are now used routinely "to automate workflows, analyse large datasets, build predictive models and improve quality control processes\."

The question is no longer whether your geotech is using AI on your project\.  The question is which of three places they're using it, and whether you've agreed on the third\.

**Where AI Is Showing Up in Geotech Engineering Services:**

```html-table
<table><thead><tr><th>Risk Level</th><th>Example Use Cases</th><th>Where It Lands</th></tr></thead><tbody><tr><td><strong>Low</strong></td><td>Borehole-log data extraction; fusion of borings, CPTs, seismic, resistivity, and GPR data into continuous subsurface models<sup><a href="#ref-8" class="footnote-ref">8</a></sup></td><td>Pre-deliverable data prep — far from the seal</td></tr><tr><td><strong>Medium</strong></td><td>Self-learning soil and rock testing systems that adapt protocols in real time based on incoming measurements<sup><a href="#ref-9" class="footnote-ref">9</a></sup></td><td>Lab and field — process improvements with QA loops</td></tr><tr><td><strong>High</strong></td><td>Drafting technical specifications, test methods, and report sections that flow into signed-and-sealed work product<sup><a href="#ref-10" class="footnote-ref">10</a></sup></td><td>At the seal — where Layfield Group's warning lands</td></tr></tbody></table>
```

Layfield Group puts the high\-risk line bluntly[10](/blog/blog-geotech-engineering-services#ref-10): "AI should never be used to generate technical specifications, test methods, test reports, technical articles, or research papers without the author having a thorough understanding of the subject\."  That sentence is the whole basis for why a competent geotech might use AI for borehole\-log extraction in the morning and refuse to use it for spec drafting in the afternoon\.  It's the same engineer\.  It's a different category of work\.

Once you accept that something is already happening, the next move is the conversation itself — and there's a way to open it that doesn't land as an audit\.

## How To Open The Conversation: The Four Questions

The conversation works when you open it as a question about shared protection — yours and theirs — rather than a policy demand\.  Four questions, asked in this order, get the alignment without the friction\.

Ask, don't tell\.  Engineers experience top\-down AI mandates as attacks on their craft; they experience honest questions as collaboration\.  This is a pattern we've documented before in our work on [why engineering firms struggle with AI adoption](https://dancumberlandlabs.com/blog/engineering-firms-ai-adoption/) — the firms that move fastest on AI are the ones whose leaders asked first instead of mandating first\.

Sequence matters\.  Start with what they're already comfortable doing, before you ask what they're not\.  Trust gets built on the upside before you negotiate the protection\.

**The four questions:**

1. **Which AI tools are you currently comfortable using on our work?** — lets them lead with what's already working and gives the engineer a chance to share craft, not defend it\.
2. **What types of deliverables do you keep AI out of entirely?** — surfaces their professional judgment about where the seal can't go near AI, and this is where you'll learn the most\.
3. **When AI touches a deliverable, who is the human in the loop, and where does the seal sit?** — gets at the chain of review and the liability anchor; the answer should be specific\.
4. **What does your PI carrier currently say about AI use, and has anything changed in 2026?** — names the Verisk wave[3](/blog/blog-geotech-engineering-services#ref-3) and Howden's broader caution[4](/blog/blog-geotech-engineering-services#ref-4) without you having to lecture about it\.

> **Before you open this conversation, call your own PI broker first\.**  Making AI a formal policy issue with your subs can itself trigger a coverage question on your own policy\.  Get your own carrier's posture clear before you start the call\.

The four questions belong on a sticky note next to the phone\.  A founder of a $20M–$100M prime should be able to ask all four in fifteen minutes and leave the call with a clearer read on the relationship than two years of project work has provided\.

Conversation establishes alignment\.  Contracts make it survive personnel changes\.

## What Goes In Writing: The Contractual Moves

Three contract moves turn the conversation into something durable: an AI disclosure clause in the subconsultant agreement, a flow\-down provision that follows AI obligations across the chain, and an indemnity that contemplates AI\-derived error\.  The federal precedent already exists; private contracts are following\.

**1\. AI disclosure language\.**  Specify what AI tools may or may not be used on which categories of work product, and what level of human review is required before anything bearing a seal goes out the door\.  Don't ban; scope\.  A blanket ban is rarely enforceable in practice and ignores the low\-risk uses that are saving everyone time\.

**2\. Flow\-down clauses that name AI\.**  AXA XL's contract guidance[11](/blog/blog-geotech-engineering-services#ref-11) reminds prime consultants that "a flow\-down, or pass\-through, clause requires the obligations of the Prime Agreement to also apply to the subconsultant\."  The 2026 question is whether yours name AI specifically, or assume it doesn't matter\.  A 2024 flow\-down clause silent on AI is a 2026 liability question with no answer\.  Our work on [AI strategy for professional services firms](https://dancumberlandlabs.com/services/ai-strategy/) covers the contractual posture in more detail\.

**3\. AI\-aware indemnity\.**  Spell out who covers an AI\-derived error in a sealed deliverable, and how that indemnity interacts with the subconsultant's PI policy as those policies start excluding AI work[3](/blog/blog-geotech-engineering-services#ref-3)\.  This is where the engineer's seal[3](/blog/blog-geotech-engineering-services#ref-3), the carrier's exclusion endorsement[3](/blog/blog-geotech-engineering-services#ref-3), and the prime's flow\-down obligation[11](/blog/blog-geotech-engineering-services#ref-11) all collide\.

The federal precedent points where private contracts are heading\.  Under the GSA's proposed AI clause, per Crowell & Moring[12](/blog/blog-geotech-engineering-services#ref-12), "the prime contractor is responsible for all AI service providers" — including subcontractors and commercial vendors — "even when the contract has nothing to do with delivering AI\."  Standardized AI clauses for AEC sub agreements don't exist yet\.  But the direction is set\.

**How AI liability flows in geotech engineering services:**

> **Prime consultant** \(architect / civil / developer\) → **Geotechnical subconsultant** \(under flow\-down agreement\) → **AI tool vendor** \(accepts no liability for output\) → **Engineer's seal** \(signs, owns it personally\) → **PI carrier** \(now writing AI exclusion endorsements as of 2026\-01\-01\)

Read that chain carefully\.  The liability lands on the engineer\.  The vendor is offshore from the conversation\.  The carrier is moving away\.  That's the asymmetry — and it's why the conversation in Section 4 isn't optional\.

Conversation and contract handle the relationships you want to keep\.  There's a separate question about the ones that may not be working\.

## The Line Between Caution And Resistance

Caution and resistance look identical from the outside\.  The diagnostic is whether your geotech can articulate *what* they avoid and *why* — caution is specific, resistance is blanket\.

Caution sounds like: "I'll use AI for borehole\-log extraction, not for sealed engineering judgment\."  Resistance sounds like: "We don't touch any of it, and we won't talk about why\."  The first is anchored in [how to think about AI implementation](https://dancumberlandlabs.com/services/ai-implementation/) at a scope level\.  The second is a closed door with no handle\.

```html-table
<table><thead><tr><th></th><th><strong>Rational Caution</strong></th><th><strong>Outdated Resistance</strong></th></tr></thead><tbody><tr><td><strong>Specificity</strong></td><td>Names what's in scope and what isn't</td><td>Blanket; can't articulate the line</td></tr><tr><td><strong>Willingness to discuss</strong></td><td>Engages the four questions in detail</td><td>Won't engage; treats the questions as an attack</td></tr><tr><td><strong>Openness to update</strong></td><td>Will revisit as tools, policies, and carrier guidance evolve</td><td>Position is static, immune to new information</td></tr></tbody></table>
```

This is a diagnostic for the prime, not a verdict on the geotech\.  Use it as judgment, not weapon\.  The conversation in Section 4 is also the diagnostic — if a firm can't engage with the four questions at all, that's the data\.

What you do next depends on whether the relationship is salvageable, scope\-able, or replaceable\.  That's not a contract question — it's a leadership one\.

## FAQ: The Questions Prime Consultants Are Actually Asking

Five questions come up in nearly every prime/geotech AI conversation\.  Here are the direct, sourced answers\.

**Are geotechnical engineers using AI?**  Yes\.  51% of geotechnical professionals reported AI use in 2025, up from 19% in 2020[7](/blog/blog-geotech-engineering-services#ref-7), primarily for data extraction, dataset analysis, predictive modeling, and quality control\.  LLMs like ChatGPT, Claude, and Copilot are now routine in office workflows[7](/blog/blog-geotech-engineering-services#ref-7)\.

**Who is liable if a geotechnical report contains an AI hallucination?**  The engineer who signs and seals the report\.  AI vendors accept no liability for output, and PI carriers are increasingly writing AI exclusions — Verisk's endorsement forms went into effect January 1, 2026[3](/blog/blog-geotech-engineering-services#ref-3)\.  The vendor accepts none of it\.  The engineer accepts all of it\.  That asymmetry is the whole reason the conversation matters\.

**Can I require my geotechnical subconsultant to disclose AI use?**  Yes — through subconsultant agreement disclosure language and flow\-down clauses[11](/blog/blog-geotech-engineering-services#ref-11)\.  The GSA's proposed AI clause[12](/blog/blog-geotech-engineering-services#ref-12) is the federal precedent that private contracts are likely to follow\.

**Should I ban AI use by my subconsultants?**  A blanket ban is rarely workable and may not be enforceable in practice\.  Disclosure plus scope limits — what AI may be used for, on which deliverables, with what human review — is the more defensible posture[11](/blog/blog-geotech-engineering-services#ref-11)[12](/blog/blog-geotech-engineering-services#ref-12)\.

**What does ASCE say about AI in engineering?**  ASCE's Code of Ethics[5](/blog/blog-geotech-engineering-services#ref-5) requires engineers to perform services only in areas of their competence and to hold paramount public safety\.  Both apply directly to AI use\.  ASCE's own ethics scholarship[13](/blog/blog-geotech-engineering-services#ref-13) adds that engineers have a duty to ensure unbiased AI training data — the example given is that "geotechnical reports from a previous project in coastal Tampa, Florida" should not be used to train a model for use on a project in rocky Denver, Colorado\.

These are the questions\.  The harder thing — the part most prime consultants don't have a partner for — is mapping the answers to your specific contracts, projects, and risk posture\.

## The Geotech Engineering Services Conversation Is Partnership, Not Audit

The AI conversation with your geotech engineering services partner isn't an audit you run on them\.  It's a protection you build with them\.  Both parties have something to lose by silence, and something to gain by clarity\.

If you do one thing this week: call your own PI broker, then call the principal engineer at your most\-relied\-on geotech sub and ask the four questions\.  Fifteen minutes on each call\.  You'll know more about the relationship than two years of project work has told you\.

Move closer to the fire\.  The geotech relationships worth keeping are the ones that can survive the conversation; the contracts worth signing are the ones that name the risk you're both already managing\.  The engineer's professional caution is rational, and the prime's need to keep work moving is real\.  Neither side has to lose for the project to win\.

If mapping the four questions, the contractual moves, and the caution\-vs\-resistance diagnostic to your specific portfolio of geotech subs feels like the kind of work that deserves a partner, our [AI implementation services for AEC firms](https://dancumberlandlabs.com/service/) are built for exactly this — peer\-to\-peer, no hype, mapped to your contracts and your carriers, not a vendor's roadmap\.

## References

1. Bluebeam, "New Bluebeam Report Shows Early AI Adopters in AEC Seeing Significant ROI Despite Uneven Adoption" \(October 2025\) — [https://press\.bluebeam\.com/2025/10/new\-bluebeam\-report\-shows\-early\-ai\-adopters\-in\-aec\-seeing\-significant\-roi\-despite\-uneven\-adoption/](https://press.bluebeam.com/2025/10/new-bluebeam-report-shows-early-ai-adopters-in-aec-seeing-significant-roi-despite-uneven-adoption/)
2. American Society of Civil Engineers, "Architecture, engineering, construction sector slow to adopt AI, survey shows" \(December 18, 2025\) — [https://www\.asce\.org/publications\-and\-news/civil\-engineering\-source/article/2025/12/18/architecture\-engineering\-construction\-sector\-slow\-to\-adapt\-ai\-survey\-shows](https://www.asce.org/publications-and-news/civil-engineering-source/article/2025/12/18/architecture-engineering-construction-sector-slow-to-adapt-ai-survey-shows)
3. Risk Specialty Group, "AI Liability Insurance For Architects 2026" \(2026\) — [https://riskspecialtygroup\.com/ai\-liability\-insurance\-architects\-2026/](https://riskspecialtygroup.com/ai-liability-insurance-architects-2026/)
4. Howden Group, "Top 5 claims risks facing engineers in 2026" \(2026\) — [https://www\.howdengroup\.com/uk\-en/top\-5\-claims\-risks\-facing\-engineers\-2026](https://www.howdengroup.com/uk-en/top-5-claims-risks-facing-engineers-2026)
5. American Society of Civil Engineers, "ASCE Code of Ethics" \(2020\) — [https://www\.asce\.org/\-/media/asce\-images\-and\-files/career\-and\-growth/ethics/documents/asce\-code\-ethics\.pdf](https://www.asce.org/-/media/asce-images-and-files/career-and-growth/ethics/documents/asce-code-ethics.pdf)
6. Construction Dive, "When AI chatbots hallucinate, infrastructure pays" \(2025\) — [https://www\.constructiondive\.com/news/ai\-hallucinate\-construction\-data\-infrastructure\-technology/814966/](https://www.constructiondive.com/news/ai-hallucinate-construction-data-infrastructure-technology/814966/)
7. Ground Engineering, "How artificial intelligence is reshaping geotechnical engineering skills" \(February 20, 2026\) — [https://www\.geplus\.co\.uk/features/how\-artificial\-intelligence\-is\-reshaping\-geotechnical\-engineering\-skills\-20\-02\-2026/](https://www.geplus.co.uk/features/how-artificial-intelligence-is-reshaping-geotechnical-engineering-skills-20-02-2026/)
8. TRC Companies, "How AI and Machine Learning are Transforming the Geotechnical Lifecycle" \(2025\) — [https://www\.trccompanies\.com/insights/how\-ai\-and\-machine\-learning\-are\-transforming\-the\-geotechnical\-lifecycle/](https://www.trccompanies.com/insights/how-ai-and-machine-learning-are-transforming-the-geotechnical-lifecycle/)
9. Fugro, "Transforming Soil Lab Testing with AI Technology" \(2025\) — [https://www\.fugro\.com/news/long\-reads/2025/transforming\-soil\-lab\-testing\-with\-ai\-technology](https://www.fugro.com/news/long-reads/2025/transforming-soil-lab-testing-with-ai-technology)
10. Layfield Group, "AI: The Next Big Issue in Geotechnical Design & Construction" \(2025\) — [https://www\.layfieldgroup\.com/geosynthetics/resources/knowledge\-center/ai\-the\-next\-big\-issue\-in\-geotechnical\-design\-construction/](https://www.layfieldgroup.com/geosynthetics/resources/knowledge-center/ai-the-next-big-issue-in-geotechnical-design-construction/)
11. AXA XL, "8 design contract clauses every subconsultant should know" \(2024\) — [https://axaxl\.com/fast\-fast\-forward/articles/8\-design\-contract\-clauses\-every\-subconsultant\-should\-know](https://axaxl.com/fast-fast-forward/articles/8-design-contract-clauses-every-subconsultant-should-know)
12. Crowell & Moring LLP, "AI for Government: 7 Days for Contractor Comments on GSA Proposed Contract Clause for AI Systems" \(March 2026\) — [https://www\.crowell\.com/en/insights/client\-alerts/ai\-for\-government\-7\-days\-for\-contractor\-comments\-on\-gsa\-proposed\-contract\-clause\-for\-ai\-systems](https://www.crowell.com/en/insights/client-alerts/ai-for-government-7-days-for-contractor-comments-on-gsa-proposed-contract-clause-for-ai-systems)
13. American Society of Civil Engineers, "Balancing Breakthrough & Ethics: Artificial Intelligence in Civil Engineering" \(2024 Mead Prize\) — [https://www\.asce\.org/\-/media/asce\-images\-and\-files/communities/students\-and\-younger\-members/documents/2024\-mead\-prize\-for\-students\-winning\-paper\.pdf](https://www.asce.org/-/media/asce-images-and-files/communities/students-and-younger-members/documents/2024-mead-prize-for-students-winning-paper.pdf)


---

Source: https://dancumberlandlabs.com/blog/geotech-engineering-services/