How does your AI workflow hold up?
Six questions to ask any AI tool your team uses on site files. Answer for the tool you use today: any “no” is a gap that leaves you at risk.
-
Do answers come from your site's own documents?
Best practice: ensure no interpretation by AI, only answers from the data and documents.
-
Can you open the page behind every citation?
Best practice: a citation that you can verify immediately.
-
Was the tool tested for environmental tasks?
Best practice: AI tested on environmental data review, lab results and site reports.
-
Does it surface the facts and flag data quality issues?
Best practice: it should pull the facts to identify QA/QC errors, data inconsistencies and outliers to allow you to check them.
-
Does it run the statistics, or only describe them?
Best practice: make sure you can see the calculations and double-check the analysis.
-
Does an expert review the AI’s work before it is used?
Best practice: a qualified professional should check the output before it goes into a report.
Generic AI and Statvis, side by side.
Ask the same six questions of any tool, including ours.
| Question | Generic AI | Statvis |
|---|---|---|
| Do answers come from your site's own documents? | The answers come from the model’s training data, plus whatever you pasted in. | The answers come from the site’s own record. Every page of every document is indexed and searchable. |
| Can you open the page behind every citation? | Not always. A citation can be generated text that only looks like a reference. | Each citation links to its source page, paragraph or cell to ensure review. |
| Was the tool tested for environmental tasks? | Tested on generic tasks, not on environmental work. | EnviroBench tests AI tools on environmental tasks before a model is used. |
| Does it surface the facts and flag data quality issues? | Answers the question asked. It doesn’t check the data behind it. | Surfaces the facts and flags QA/QC issues, such as data inconsistencies and outliers, for a person to check. |
| Does it run the statistics, or only describe them? | May describe a method it never ran. | Runs HCA, UMAP and PCA on data to present findings geospatially in 2D and 3D. |
| Does an expert review the AI’s work before it is used? | Not built in. Typically requires additional attention to the reference and rationale. | The data is served up for a qualified person to review the output and citation. |
Webinar with ESAA · Zoom
AI in contaminated sites work, a practical introduction.
A jargon-free introduction for environmental professionals: what AI can and can’t do, how to judge whether a tool is safe to use in professional practice, and where it will change your day-to-day work. The examples come from contaminated sites practice, and no AI background is needed.
- When
- Wednesday, December 9, 2026
1:00 to 3:00 pm Mountain Time (Alberta) - Where
- Online, on Zoom
- Cost
- $29 for ESAA members, $39 for non‑members
Registration and payment are on the ESAA site.
