REPSShield AI Advisor: How It Works

Most tax software uses AI as a label for rule-based autocomplete. REPSShield’s AI Advisor is something different: an agentic system with three specialized operational modes, a curated knowledge base of 282 IRS and Tax Court documents, tool-calling capabilities, and a 12-check compliance engine that evaluates your REP status documentation against the same framework an IRS examiner uses.

This post describes how each component works and why the architecture matters for a taxpayer whose REP status may be audited.

TL;DR: REPSShield’s AI Advisor operates in three modes — Quick (conversational time entry), Advisor (citation-backed tax guidance from 282 IRS and Tax Court documents), and Compliance (a 12-check audit engine that evaluates your log the way an IRS examiner would). Unlike a generic chatbot, it reads your actual time log, cites specific regulations and case law, and identifies documentation gaps before you file.

Beyond a Chatbot

A generic AI chatbot answers questions from training data with no reference to your specific situation, no access to current IRS guidance, and no ability to take action. It cannot read your time log, assess your documentation gaps, or tell you that the IRS Passive Activity Loss Audit Technique Guide specifically instructs examiners to look for rounded hour entries in logs like yours.

REPSShield’s AI Advisor operates with four properties that distinguish it from a generic interface:

Specialized knowledge base — 282 curated documents including IRS publications, Treasury regulations, Tax Court cases through early 2026, and IRC sections relevant to Section 469 and material participation. Answers are grounded in this corpus, not general training data.
Tool use — The AI can call internal tools: reading your time entries, running your hours against the material participation tests, triggering the compliance engine, creating structured time entries on your behalf, and retrieving property-specific data.
Three distinct operational modes — Quick mode for time entry, Advisor mode for guidance, and Compliance mode for audit preparation. Each mode uses different retrieval strategies, different system prompts, and different output formats.
Session memory — Conversations persist across sessions. The AI knows what you discussed last week, what your filing situation is, and which properties you have been asking about.

Quick Mode: Conversational Time Entry

The most friction-laden part of REP documentation is the moment immediately after a qualifying activity — when you need to create a log entry while the details are fresh, but opening a form and filling in five fields is enough friction to make you defer it until later, when “contemporaneous” no longer applies.

Quick mode addresses this with conversational entry creation. You describe what you did in plain language:

“I spent about two hours at the Oak Street property this morning dealing with the furnace issue — the HVAC contractor came out and I was there for the inspection.”

The AI parses this into a structured time entry:

Date: Today’s date (inferred from session context, correctable)
Property: Oak Street property (matched against your property list)
Activity type: Construction/improvement oversight (selected from 30 recognized types)
Hours: 2.0
Description: “HVAC contractor furnace inspection — on-site for duration of inspection.”

Before the entry is committed, you see an interactive preview card that shows all five fields and lets you adjust any of them. You confirm, and the entry is created with a server-side timestamp that establishes contemporaneous creation.

The AI is not guessing at activity types from a list of 14. REPSShield recognizes 30 qualifying activity types derived from the IRS regulations and audit guidance — including distinctions like “contractor oversight” versus “direct construction work,” “tenant acquisition” versus “tenant management,” and “property acquisition due diligence” versus “investment research” (which does not qualify). The correct classification matters because an IRS examiner reviewing your log will question whether each entry’s activity type is genuinely qualifying.

Interactive form cards. For more complex situations — multiple properties in one session, a partially remembered activity, or a time entry that needs supporting evidence attached — Quick mode renders interactive form cards within the chat interface. A SelectFormCard presents a property picker or activity type selector inline. A QuestionnaireCard walks through the five required fields in a guided sequence when the initial input was ambiguous. These are not pop-ups; they render as conversational elements so the entry creation flow does not require switching to a different part of the application.

Advisor Mode: Citation-Backed Tax Guidance

The question “does time spent coordinating with my property manager count toward my 750-hour total?” has a specific answer grounded in Treasury Regulation 1.469-9(b)(4) and a body of Tax Court cases. A generic chatbot gives you a hedged, general answer. Advisor mode gives you the regulation citation, the relevant case law, and an explanation of how the rule applies to your specific situation.

The Knowledge Base

REPSShield’s knowledge base contains 282 documents assembled specifically for Section 469, real estate professional status, and material participation analysis:

IRS Publications 527, 925, and 946
Treasury Regulations 1.469-1 through 1.469-11 and temporary regulations
IRC Sections 469, 168, 1231, and related provisions
Tax Court memoranda and opinions through early 2026, including Mirch v. Commissioner (2025), which established updated standards for documentation quality in REP audits
IRS Passive Activity Loss Audit Technique Guide
Chief Counsel Advice memoranda on material participation and grouping elections
Revenue Procedures relevant to late elections and corrections

This corpus is maintained by REPSShield’s tax research team and updated as new cases and guidance are published. The AI does not answer REP questions from general training data — it retrieves from this corpus first.

Finding the Right Answer in a Large Document Set

282 documents is a lot of material. When you ask a question, the AI does not simply scan for matching words — it reasons about what you meant and looks for relevant guidance even when the exact terminology differs between your question and the source document.

A question like “does travel time count toward my hours?” will surface IRS guidance on transportation to and from rental activities even if the regulation uses different phrasing. This matters because the IRS, Treasury, and Tax Court all use different language for the same concepts — and an answer built only on exact keyword matching would miss large portions of the available guidance.

Current events and recent cases. The knowledge base is updated regularly, but for very recent Tax Court decisions or IRS announcements — something decided last month — the AI supplements its research with targeted searches of IRS.gov and Tax Court publications. When an answer draws on a recent web source rather than the reviewed knowledge base, this is clearly labeled so you know the confidence level of what you are reading.

Citation Chips

Every Advisor mode answer that draws on a specific document includes citation chips — inline markers that identify the source and its trust level. A chip might read “Treas. Reg. 1.469-9(b)(4) · Knowledge Base” or “Mirch v. Commissioner, T.C. Memo 2025-18 · Knowledge Base” or “IRS.gov · Trusted Web.”

The trust level distinction matters. Knowledge base documents have been reviewed and curated by the REPSShield tax research team. Web results are authoritative but unreviewed. The distinction tells you how much weight to place on the citation when you take it to your accountant or tax attorney.

Compliance Mode: See Through an IRS Examiner’s Eyes

Advisor mode answers your questions. Compliance mode inverts the perspective — it evaluates your documentation the way an IRS examiner would, before you are ever contacted.

Compliance mode follows the IRS Passive Activity Loss Audit Technique Guide, which is the actual internal document that IRS examiners use when examining REP status claims. This is a publicly available document that most REP investors have never read. The Audit Technique Guide tells examiners to look for specific patterns: rounded hours, implausibly high totals, logs that appear created after the fact, activities that sound qualifying but are not, and material participation claims that rely on the wrong test for the property type.

The 12-Check Audit Engine

Compliance mode runs 12 discrete checks against your time log and tax profile, generating a compliance report you can download as a PDF:

1. Contemporaneous logging pattern. Analyzes the distribution of entry creation timestamps relative to activity dates. Entries created within 24 hours of the activity date are strongest; entries created days or weeks later are flagged with the number of days of lag. A pattern of same-day logging is a significant audit defense.

2. Rounded hours detection. Flags entries where hours are suspiciously round — every entry exactly 1, 2, 3, or 4 hours — which is a documented IRS red flag. The check looks at both individual entries and statistical patterns across your full log.

3. Duration plausibility. Cross-references logged duration against the activity type and property. An entry logging 8 hours of “tenant screening” for a single-unit property in one day is flagged as implausible. An entry logging 0.25 hours of “construction oversight” is flagged as unusually short for a site visit.

4. Activity eligibility verification. Each entry’s activity type is checked against the IRS qualifying activity taxonomy. Entries categorized as “investment analysis,” “education,” or “general research” are flagged as non-qualifying and excluded from your qualifying hour total.

5. Evidence presence. Entries without attached evidence are identified. The check calculates what percentage of your logged hours have corroborating documentation and flags high-value entries (large hour blocks, entries in heavily audited activity types) that lack evidence.

6. Material participation test application. Verifies that you are relying on the correct material participation test for each property. An investor with a full-service property manager who claims Test 3 (100 hours, more than anyone else) is flagged — the manager’s hours likely exceed the investor’s.

7. On-call time analysis. Flags entries that appear to claim on-call time as qualifying hours. The IRS and Tax Court have consistently held that time spent “on call” or “available” does not constitute participation unless actual services were rendered.

8. Travel time attribution. Verifies that travel time entries are attached to a destination property and a qualifying activity. Travel time claimed without a linked activity at the destination is a common audit adjustment.

9. Pattern consistency across tax year. Looks for discontinuities — months with zero entries, months with unusually high entries, large gaps followed by catch-up logging. A consistent pattern across the year is more credible than an erratic one.

10. Short-term rental path analysis. For STR properties in your account, applies the 500-hour and 100-hour material participation tests separately from the REP analysis, and flags properties where claimed losses may not be supportable under either path. See the getting started guide for the STR threshold details.

11. 50% test verification. Calculates your total qualifying real estate hours against your documented non-real-estate professional time. Flags if the more-than-half-time test appears at risk based on your logged hours and your stated W-2 or business hours.

12. Grouping election consistency. If you have made a grouping election, verifies that your material participation analysis is being applied to the grouped activity rather than property-by-property, and flags any properties that appear to have been added after the election without an election amendment.

The compliance report presents each check with a status (pass, warning, or flag), the specific issue identified, and a recommended remediation action. The downloadable PDF is formatted for use as an internal review document or as a reference when discussing your position with a tax advisor.

The Mirch doctrine. The 2025 Mirch decision added specificity to what “contemporaneous” means in the context of electronic records. Compliance mode applies the Mirch documentation quality standard directly — flagging entries that would not satisfy the Mirch standard based on their creation timestamp, content specificity, and evidence attachment status.

Session Memory

REPSShield stores your conversations with the AI Advisor in a persistent, searchable session history. Conversations are retained for 12 months and are searchable by keyword.

AI-generated titles. Each conversation is automatically titled based on its content — “Material participation test for Oak Street STR,” “Evidence requirements for contractor oversight hours” — so you can navigate your history without reading through every thread.

Rolling summarization. Long conversations — a 40-message thread working through a complex grouping election question — are automatically summarized as they grow beyond the context window. The AI maintains a rolling summary of earlier turns that provides continuity without requiring you to re-explain your situation at the start of every session.

Cross-session context. When you start a new conversation, the AI has access to your 19-step tax profile (see below), your current-year hour totals by property, and a summary of recent conversation topics. You do not need to re-establish your situation in every session.

The 19-Step Tax Profile

Every recommendation from the AI Advisor is shaped by a tax profile built through an interactive questionnaire — 19 questions that establish the parameters of your filing situation, investment structure, and personal involvement style.

The profile covers:

Filing status (single, married filing jointly, married filing separately, head of household)
Employment type (W-2 employee, self-employed, business owner, retired, or combinations)
Approximate AGI range (which affects passive loss deduction phaseouts)
Number and type of rental properties
Whether you use a property manager (and at what level of involvement)
Your primary role in property operations (hands-on, supervisory, or strategic)
Whether you have made a grouping election
Whether you have short-term rentals, and their average rental period
Your approach to evidence collection

The questionnaire renders as an interactive QuestionnaireCard within the chat interface — not a separate form. Each question appears in sequence, with the AI providing context for why the question matters and how the answer will shape its recommendations.

Once complete, the profile is stored and informs every subsequent Advisor and Compliance mode interaction. An investor filing jointly with a spouse who is the designated REP receives different analysis than a single filer targeting REP status personally. An investor with a full-service property manager is flagged on Test 3 material participation. An investor with a high W-2 hour count sees their more-than-half-time test risk quantified in every compliance check.

The profile can be updated at any time as your situation changes — a new property, a job change, a property manager transition. Updating the profile immediately adjusts the AI’s analysis.

Pushback Mode

The default AI Advisor configuration is informative and constructive. It explains requirements, flags gaps, and suggests improvements without challenging your judgment on entries you have already logged.

Pushback mode inverts this posture. When enabled, the AI acts as a constructive adversary — reviewing your entries and your reasoning and raising the objections an IRS examiner or opposing Tax Court counsel might raise.

For an entry that reads “Property management — 3 hours — general oversight of Oak Street property,” Pushback mode responds:

“This entry is vulnerable. ‘General oversight’ is not a recognized activity type, and the description does not identify a specific task performed. An IRS examiner reviewing this entry will ask what specific services were rendered. Consider revising the description to specify what oversight consisted of — a site visit, a tenant call, a review of maintenance work — and update the activity type to match.”

Pushback mode is opt-in and session-level. You turn it on when you want a stress test of your documentation before year-end or before filing. You turn it off for standard advisory conversations. It is designed for investors who want to find their own weaknesses before an examiner does.

Smart Query Routing

Not every question requires the same level of computation. “What is the 750-hour rule?” is a retrieval question that a fast, lightweight model can answer accurately from the knowledge base. “My wife works 1,400 hours per year at her W-2 job, I have three LTRs and two STRs, I used a property manager for six months at Oak Street, and I’m trying to figure out whether I can group my LTRs to hit the 500-hour material participation test — what are the risks?” is a multi-property, multi-test reasoning problem that benefits from a full-powered model with extended context.

REPSShield routes queries automatically based on complexity signals: length, the number of properties referenced, whether the question involves multi-test analysis or cross-property aggregation, and whether the question references your specific log data. Simple questions return answers in under two seconds. Complex multi-property questions take longer and surface the routing decision in the UI — “Using enhanced analysis for this question” — so you know you are getting the deeper treatment.

This routing is transparent and overridable. If you want enhanced analysis on a simple question, you can request it. If you want a faster answer on a complex question and are willing to trade depth for speed, you can specify that.

REPSShield’s AI Advisor is available in full during the 14-day free trial — all three modes, the complete 12-check compliance engine, full session history, and the 19-step tax profile. No credit card required to start.

The IRS does not give you a warning before it selects your return for examination. The time to run the compliance checks is before you file, not after you receive the IDR. Start your free trial and run the compliance engine against this year’s log before the filing deadline.

Related reading:

Getting Started with REP Hour Tracking — the documentation standards that the AI compliance engine measures your log against.
Connect Your Tools to REPSShield — how REPSShield syncs with the property management software and apps you already use, so hours log themselves.