AI Soil Evaluation Report for Septic Designers: What It Can Do

An AI soil evaluation report is a drafting and checking layer built around the evaluator's own field data. The agent turns notes, photos and boring logs into the report structure and flags gaps or contradictions, while the site evaluator or licensed professional classifies the soil, reaches the conclusions and signs. The time you save is in the typing and the cross-checking, never in the judgment.

Key takeaways

• The evaluator observes, classifies and decides. An agent only organizes, formats and checks what the evaluator recorded.
• The biggest saving comes from turning field notes, photos and boring logs into the report form and narrative, not from interpreting the soil.
• Automated cross-checks catch internal contradictions (depth continuity, texture against the stated class, restrictive feature against the selected system) before a reviewer does.
• Web Soil Survey is a useful desktop comparison and never a replacement for an actual hole in the ground.
• Every output is a draft with visible source notes, and your state rules decide what the report must contain.

What does an AI soil evaluation report actually do?

Think of the work in two halves. The first half happens in the field: you dig or bore, look at color, texture, structure and any redoximorphic features, note the depth of anything restrictive, and photograph the profile. That half is yours, and it needs your eyes and your license.

The second half happens at the desk: the notes become a form, the form becomes a narrative, the narrative has to agree with the site plan and the system you selected, and the whole thing becomes a PDF in the right order for the permit portal. This half is where an agent helps. It reads your handwritten or typed notes, places each value in the right field, and drafts the observation language in your voice from your own wording.

In Texas the site evaluation requirements sit in 30 TAC 285.30, and the planning materials a permit authority expects sit in 30 TAC 285.5. An agent can be given the headings and fields your office already uses for those, so the draft arrives in the shape you submit, not in a generic one.

What stays with the evaluator?

Some things must not move, however good the draft looks.

• Soil classification and the decision on what each horizon is.
• The judgment on whether a feature is restrictive, and what it means for the site.
• The conclusion about which systems the site can support.
• Any statement that certifies the evaluation, plus the signature and seal where your rules require them.
• The choice to go back to the field when something in the draft does not fit what you saw.

The agent has no standing to decide these, and a good setup makes that visible. Every draft line should show which of your notes it came from. If a line has no source, it is flagged as a gap instead of being filled in with something plausible. Your state rules set the required content and the thresholds. The agent does not supply them from memory. For Texas, read the rule text directly at the source, and treat the agent as the assistant that keeps your own template filled and consistent.

From field notes to a draft report, step by step

This is a workflow an office can adopt without changing how it digs.

• Capture the notes the same way every time: a short template for each hole with depth ranges, texture, color, structure, features, water observations and a free comment line.
• Give the agent the notes, the photos with hole labels, the site plan and your blank report template.
• Ask for a first pass that fills the template only from the notes, and lists every field it could not fill.
• Review the list of gaps and answer each one yourself, from the field or from your memory of the site.
• Ask for the narrative sections drafted from the filled fields, in your usual phrasing, with the source field named beside each statement.
• Run the cross-checks described in the next section and resolve every flag.
• Read the whole report as the person who signs it. Edit freely. Anything you change is final, and the agent never overwrites your edit on a later pass.
• Export the package in the order your permit authority wants and keep the notes with the file.

Cross-checks worth automating

A tired reviewer misses small contradictions. A checker does not tire. These are the checks that tend to pay off, all of them compared against your own entries and never against invented values.

• Depth ranges for each hole run without gaps or overlaps from the surface down.
• The texture written in the description agrees with the class name you assigned to that horizon.
• A restrictive feature noted in the field appears in the narrative and in the system selection reasoning.
• Hole labels in the report match the labels on the site plan and the photo files.
• Dates, evaluator name and license field are present and consistent from the first page to the last.
• The system type named in the report matches the one on the plan sheet.
• Any field left blank is listed, not silently skipped.

When the checker flags something, it describes the inconsistency and points at both places. It does not propose that one of your observations was wrong. That conversation is between you and the soil.

An illustrative scenario

This is an illustrative example, not a real project. An evaluator spends the morning on two sites and writes terse notes on a tablet between holes. By afternoon the notes are good but uneven, one hole recorded depths in inches and another in feet, and a photo label does not match the plan.

The agent produces a filled report for each site and a list of what it cannot place: the unit mismatch, the unmatched photo, and a hole with no water observation. The evaluator answers the three questions in a couple of minutes, since she remembers the sites. The narrative draft names which note each sentence came from. She rewrites one sentence she finds too confident about a feature, and the agent keeps her version.

During the cross-check, the checker notes that a shallow restrictive layer appears in the notes for one hole but is missing from the paragraph on system suitability. She adds it herself and rethinks the paragraph. That catch is the whole point: the draft did not decide anything, it made the gap visible.

How Be Human handles this as a managed service

Be Human designs, implements and operates personalized AI agents for small businesses, and a soil evaluation agent is set up around your own template, your own notes style and your rules. We keep the agent's permissions narrow: it reads what you give it and writes drafts, nothing more. A person at Be Human monitors the agent, adjusts it when your forms change, and you stay the reviewer of record. For the wider picture of how an agent fits a design office, see AI for septic system design, and for how the finished report fits the submittal, see automating the OSSF permit package. The general service is described on the managed AI agents page.

It also helps to keep the field-to-office link short. If the soil report and the system selection are drafted from the same notes, the choice you make in the aerobic, drip, spray or conventional checklist lines up with what the report says.

Where to start

Pick one report type you write most often and time yourself once, from the last hole to the finished PDF. The part that felt slowest is almost always the transcription and the consistency check, which is exactly where an agent earns its place. Take the free Be Human AI diagnostic to find the first task in your office worth handing to an agent.

Frequently asked questions

Can an AI classify the soil for me? No. Classification comes from what you observed in the profile, and it is a professional judgment that your state rules assign to a qualified evaluator. An agent can check that your description and your class name do not contradict each other, and it can tell you when a field is empty.

Can I use Web Soil Survey instead of digging? Web Soil Survey is a good way to get background on the mapped soil units before a site visit, and it can help you notice surprises. It is a broad map and it does not replace the site evaluation your rules require, so treat it as context and let the hole in the ground decide.

What should I check in the draft before signing? Read it as if you had written it from scratch. Confirm every value against your notes, look at each flagged gap and each cross-check result, and check that the report and the site plan describe the same system. If anything feels off, go back to your field data or to the site.

Does the agent store or share my site data? That depends on how the service is configured, so ask before you start. A managed setup should limit the agent to the files you give it, keep a record of what it read and wrote, and keep your client data out of anything shared. Authorization and data handling are worth settling in writing, and the NIST AI Risk Management Framework is a sensible checklist for those questions.

Be Human editorial guidance. Examples are illustrative unless explicitly attributed. Sources provide background; implementation choices depend on your workflow.