Update help docs and AI knowledge base for 4 new AI bookkeeping features

- Reports.cshtml: added AI Payment Risk Prediction, Ask Your Financials, and Bank Rec Auto-Match subsections under AI-Powered Reports; updated on-this-page nav with sub-links for all three - AccountsPayable.cshtml: added Recurring Bill Detection section explaining pattern cards, frequency types, confidence badges, next expected date, and the 2-occurrence minimum - HelpKnowledgeBase.cs: added Recurring Bill Detection to BILLS section; added AI Payment Risk Prediction and Ask Your Financials to REPORTS available-reports list; added features 12–15 to AI FEATURES section (Recurring Detection, Payment Risk, Financial Query, Bank Rec Auto-Match) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-10 19:30:39 -04:00
parent 959e323f3a
commit 59beba2e15
3 changed files with 103 additions and 0 deletions
@@ -511,6 +511,8 @@ public static class HelpKnowledgeBase

        **AI Receipt Scanning:** Bills → "Scan Receipt" — upload a photo or PDF of a vendor receipt/invoice and AI will pre-fill the vendor, date, amount, and line items.

+        **Recurring Bill Detection:** [Bills](/Bills) → "Detect Recurring Bills" button (top-left of the bills list, or navigate to /Bills/RecurringDetection). AI scans the last 12 months of bills to identify vendors you pay on a regular schedule. Each pattern card shows the vendor name, frequency (Monthly / Bi-Monthly / Quarterly / Annual / Irregular), typical amount, confidence level (High / Medium / Low), estimated next expected date, and a suggested action (e.g., "Set up auto-pay" or "Budget monthly"). Useful for cash flow planning — knowing what's coming reduces surprises. At least 2 occurrences are needed before the AI can establish a pattern; one-time bills are filtered out automatically.
+
        **Accounts:** [/Accounts](/Accounts) — The chart of accounts (Assets, Liabilities, Equity, Revenue, COGS, Expenses). Accounts are assigned to bill line items for financial reporting.

        **Accounting Export:** [/AccountingExport](/AccountingExport) — Export financial data to accounting software.
@@ -577,6 +579,8 @@ public static class HelpKnowledgeBase
        - *Cash Flow Forecast* — 30/60/90-day projection based on open AR, AP, and job pipeline
        - *Anomaly Detection* — AI scans for duplicate bills, amount spikes, unusual vendors
        - *AR Follow-Up Emails* — AI drafts collection emails for overdue invoices (from AR Aging report)
+        - *AI Payment Risk Prediction* — on the AR Aging report, click "Predict Payment Risk" to get an AI assessment for each outstanding customer: High / Medium / Low risk with reasoning and recommended action (call now, send reminder, standard follow-up). Powered by each customer's payment history, current balance, and days outstanding.
+        - *Ask Your Financials* — [/Reports/FinancialQuery](/Reports/FinancialQuery) — natural language query interface. Type any financial question ("What were my top expenses last quarter?", "Which customers owe the most?") and the AI answers using your real data. Includes suggestion chips, follow-up prompts, supporting facts, and session history. The right panel shows a YTD financial snapshot (revenue, expenses, net income, open AR, open AP).
        - *Powder Usage Report* — powder consumption by item/job
        - *Job Cycle Time Report* — how long jobs spend in each status

@@ -1239,6 +1243,14 @@ public static class HelpKnowledgeBase

        11. **AI Quick Quote** — A floating button (visible on every page) that lets you get an instant rough estimate from a verbal description — ideal for phone calls and walk-in customers. Type a description such as "4 wheels, gloss black, need sandblasting", enter quantity and coat count, and the AI returns a price estimate with a confidence score. Detected color names are matched against your inventory so you can see at a glance whether you have the powder in stock. You can then save the quote under a "Walk-In / Phone" customer with one click and reassign it to the real customer record later. Access via the **dark-blue floating button** in the bottom-right corner, just above the AI Help button.

+        12. **Recurring Bill Detection** — AI scans the last 12 months of vendor bills to identify recurring payment patterns. Access via [Bills](/Bills) → "Detect Recurring Bills." See the BILLS section above for full details.
+
+        13. **AI Payment Risk Prediction** — On the [AR Aging](/Reports/ArAging) report, click "Predict Payment Risk" to get a risk assessment (High / Medium / Low) for each outstanding customer with reasoning and a recommended action. Powered by payment history, current balance, and days outstanding.
+
+        14. **Ask Your Financials** — Natural language financial queries at [/Reports/FinancialQuery](/Reports/FinancialQuery). Ask questions in plain English and get answers drawn from your real financial data, with supporting facts and follow-up prompts. See the REPORTS section above for full details.
+
+        15. **Bank Rec Auto-Match** — On the [Bank Reconciliations](/BankReconciliations) Reconcile page, click "AI Suggest Matches" to have AI review your uncleared transactions and suggest which ones to mark as cleared to reach your target balance. Each suggestion includes a confidence percentage and a reason. Click "Apply All Suggestions" to mark all recommended items cleared in one step. You still control the final reconciliation — the AI suggestions are a starting point, not a commitment.
+
        **Plan availability:** AI Photo Quotes and AI Inventory Assist are enabled at the subscription plan level. If you do not see the AI Photo Quote option in the quote wizard or the AI lookup button on inventory items, the feature may not be included in your current plan. Contact your administrator or check [Billing](/Billing) to see your plan details.

        The AI Profile (in Company Settings) lets you describe your shop's specialties to improve AI quote estimates. This tab only appears when AI Photo Quotes are enabled for your account.
@@ -229,6 +229,43 @@
            </p>
        </section>

+        <section id="recurring-detection" class="mb-5">
+            <h2 class="h4 fw-bold border-bottom pb-2 mb-3">
+                <i class="bi bi-arrow-repeat text-primary me-2"></i>Recurring Bill Detection
+            </h2>
+            <p>
+                The <strong>Detect Recurring Bills</strong> tool is accessible from the Bills list via the
+                button in the top-left of the page, or directly at
+                <a asp-controller="Bills" asp-action="RecurringDetection">/Bills/RecurringDetection</a>.
+                Click <strong>"Detect Recurring Bills"</strong> and Claude analyzes the last 12 months of your
+                bill history to find vendors you pay on a regular schedule.
+            </p>
+            <p>
+                Each detected pattern is shown as a card with:
+            </p>
+            <ul class="mb-3">
+                <li class="mb-1"><strong>Vendor name</strong> and detected frequency (monthly, quarterly, biannual, annual).</li>
+                <li class="mb-1"><strong>Typical amount</strong> — the usual charge from that vendor.</li>
+                <li class="mb-1"><strong>Next expected date</strong> — Claude's estimate of when the next bill is likely to arrive.</li>
+                <li class="mb-1"><strong>Confidence badge</strong> — High (4+ consistent occurrences), Medium (2–3 occurrences or variable timing), Low (weak pattern, worth monitoring).</li>
+                <li class="mb-1"><strong>Suggested action</strong> — for example, "Set a monthly reminder for this bill."</li>
+            </ul>
+            <p>
+                This is useful for cash flow planning — knowing that a $1,200 electricity bill arrives on the
+                15th every month, or that your insurance renews every January, lets you reserve funds in advance
+                and avoid surprises. High-confidence patterns are reliable enough to act on; Low-confidence
+                patterns are worth keeping an eye on but should not be treated as certain.
+            </p>
+            <div class="alert alert-permanent alert-info d-flex gap-2 mb-0" role="alert">
+                <i class="bi bi-lightbulb-fill flex-shrink-0 mt-1"></i>
+                <div>
+                    Recurring bill detection requires at least 2 occurrences of a vendor bill at a similar
+                    interval to detect a pattern. Shops with less than 2 months of history will see few or no
+                    results. The scan covers bills only — direct expenses are not included.
+                </div>
+            </div>
+        </section>
+
        <section id="expense-accounts" class="mb-5">
            <h2 class="h4 fw-bold border-bottom pb-2 mb-3">
                <i class="bi bi-folder2-open text-primary me-2"></i>Expense Accounts
@@ -251,6 +251,57 @@
                You can edit the draft before sending.
            </p>

+            <h3 class="h6 fw-semibold mt-3 mb-2">AI Payment Risk Prediction</h3>
+            <p>
+                Available inside the <strong>AR Aging</strong> report (<a asp-controller="Reports" asp-action="ArAging">/Reports/ArAging</a>).
+                Click <strong>"Predict Payment Risk"</strong> at the bottom of the page to have Claude analyze
+                each open AR customer and assign a risk level:
+            </p>
+            <ul class="mb-3">
+                <li class="mb-1"><strong>High</strong> — customer has a history of late payment and is already overdue; prioritize a phone call today.</li>
+                <li class="mb-1"><strong>Medium</strong> — overdue but reasonable history, or current but spotty past performance.</li>
+                <li class="mb-1"><strong>Low</strong> — typically pays on time; no immediate follow-up needed.</li>
+            </ul>
+            <p>
+                Each prediction includes an estimated number of additional days to collection and a one-sentence
+                explanation of the scoring. Use this to triage your collection calls — start with High-risk
+                customers first.
+            </p>
+
+            <h3 class="h6 fw-semibold mt-3 mb-2">Ask Your Financials (Natural Language Queries)</h3>
+            <p>
+                A conversational AI tool at <a asp-controller="Reports" asp-action="FinancialQuery">/Reports/FinancialQuery</a>
+                that lets you ask plain-English questions about your business finances and get direct answers
+                grounded in your actual data. Example questions:
+            </p>
+            <ul class="mb-3">
+                <li class="mb-1">"What was our revenue this year?"</li>
+                <li class="mb-1">"What are our biggest expenses?"</li>
+                <li class="mb-1">"Which month had the highest revenue?"</li>
+                <li class="mb-1">"How much do customers owe us?"</li>
+            </ul>
+            <p>
+                Each answer includes <strong>supporting facts</strong> pulled directly from your data so you can
+                verify the figures, and a <strong>follow-up suggestion</strong> for what to ask next. Claude
+                will not invent numbers — if the data is not available in the snapshot, it says so. The page
+                also shows clickable example chips and remembers your last 5 questions during the session.
+            </p>
+
+            <h3 class="h6 fw-semibold mt-3 mb-2">Bank Rec AI Auto-Match</h3>
+            <p>
+                Inside <strong>Bank Reconciliation</strong> (<a asp-controller="BankReconciliations" asp-action="Index">/BankReconciliations</a>),
+                the <strong>Reconcile</strong> working view includes an <strong>AI Auto-Match</strong> panel.
+                Click <strong>"Suggest Matches"</strong> and Claude analyzes all uncleared transactions against
+                your statement ending balance, then suggests which items to mark as cleared — sorted by
+                confidence score with a one-sentence reason for each.
+            </p>
+            <p>
+                Click <strong>"Apply All Suggestions"</strong> to accept them in bulk; the checkboxes are marked
+                and persisted automatically. Review the highlighted rows (shown in blue) before applying if you
+                want to verify each one individually. Auto-match does not complete the reconciliation — you
+                still click "Complete Reconciliation" yourself once the difference reaches $0.00.
+            </p>
+
            <h3 class="h6 fw-semibold mt-3 mb-2">Powder Insights</h3>
            <p>
                An AI-powered analysis of your powder usage patterns, efficiency, and cost optimization,
@@ -304,6 +355,9 @@
                    <a class="nav-link py-1 px-3 small text-body ps-4" href="#financial-reports" style="font-size:.75rem">Sales Tax Report</a>
                    <a class="nav-link py-1 px-3 small text-body" href="#operations-reports">Operations Reports</a>
                    <a class="nav-link py-1 px-3 small text-body" href="#ai-reports">AI-Powered Reports</a>
+                    <a class="nav-link py-1 px-3 small text-body ps-4" href="#ai-reports" style="font-size:.75rem">Payment Risk Prediction</a>
+                    <a class="nav-link py-1 px-3 small text-body ps-4" href="#ai-reports" style="font-size:.75rem">Ask Your Financials</a>
+                    <a class="nav-link py-1 px-3 small text-body ps-4" href="#ai-reports" style="font-size:.75rem">Bank Rec Auto-Match</a>
                    <a class="nav-link py-1 px-3 small text-body" href="#pdf-export">PDF &amp; CSV Export</a>
                </nav>
            </div>