--- title: "AI Drafts Your Q3 Investor Letter | Kolena" url: "/lp/smartbrief-investor-letter/" description: "See a complete Quarterly Budget Variance output — variance schedule, NOI bridge, full-year reforecasts and the drafted investor-letter commentary, every figure cited to source." updated: 2026-09-18T04:39:57.499582+00:00 --- # AI Drafts Your Q3 Investor Letter | Kolena ## Q3 investor letter — hero, letter card, citations, sticky bar Quarterly Budget Variance Agent # AI drafts your Q3 investor letter. Point the agent at the quarter's export and it returns the variance schedule, the NOI bridge, full-year reforecasts and the drafted letter commentary — every figure traceable to the document it came from. Yardi · MRI · RealPage · AppFolio · Entrata Grounded in the CREFC Investor Reporting Package 13 extraction fields **Investor-Letter Variance Commentary** Ashgrove Commerce Center · Q2 2026 NOI variance −$157,900 −9.5% vs budget Verdict Behind plan Outside the 2.0% band Material lines 12 5 favorable, 7 not Source docs 4 Read in one pass For Q2 2026 year-to-date, Ashgrove Commerce Center generated Net Operating Income of $1,507,500 against a budget of $1,665,400, an unfavorable variance of −$157,900 (−9.5%). As the variance exceeds the fund's 2.0% threshold, property performance is classified as behind plan for the period. Effective Gross Income totaled $2,931,400 versus a budget of $2,954,000 (−$22,600 or −0.8%). Unfavorable revenue variances were driven by Base Rental Income (−$44,400), following the lease expiry and vacancy of Suite 210 (11,400 SF) by Kestrel Analytics Group on 2026-04-30, unbudgeted rent concessions on new leases signed in Q1 (−$37,600 in Free Rent and Abatements), and a receivables write-off related to Torrance Freight Systems' Chapter 11 filing (−$19,900 in Vacancy and Credit Loss). These shortfalls were partially offset by favorable variances, including $31,700 from an unbudgeted mezzanine storage licensing program launched in January and $33,200 in higher Real Estate Tax Reimbursements recovered from tenants. Total Operating Expenses were $1,423,900 compared to a budget of $1,288,600, an unfavorable variance of $135,300 (10.5%). Key unfavorable cost drivers are permanent: Real Estate Taxes ran $73,000 over budget due to the 2026 Wake County property reassessment, and Insurance Expense was unfavorable by $18,400 following the annual policy renewal at higher market rates and property values. Repairs and Maintenance exceeded budget by $59,400 (reclassified from timing to permanent), driven by a $34,000 unbudgeted chiller compressor replacement in Q2 and elevated baseline maintenance across the park. Looking ahead, full-year operating performance will reflect the permanent structural revenue reduction from Suite 210 and higher fixed tax and insurance obligations. Key items to watch in Q3 2026 include execution of the deferred marketing and repositioning campaign for Suite 210 (contracted in July at $9,200), monitoring ongoing maintenance run rates, and tracking the Wake County property tax appeal, which is pending decision in Q1 2027. Generated from four source documents. Every figure is checked against the export it came from. Sample package is synthetic — Ashgrove Commerce Center, the fund, the manager and every tenant and balance in it were invented to exercise the agent. The reasoning and the arithmetic are the agent's own. **See the full Quarterly Budget Variance output** Get the workbook, or see it on your own documents. [Get the workbook](#kq3-form) ## Q3 investor letter — what it read, traceability, comparison, field grid What it read ## Four documents, not one page. The letter above was written from these, in a single pass. No renaming, no reformatting, no consistent structure required. 01 #### Budget comparison export Yardi Voyager 7S, quarter and year to date, accrual basis. 02 #### Approved full-year budget The board-adopted plan the variances are struck against. 03 #### Prior-quarter variance report Last quarter's drivers and timing calls, for continuity. 04 #### Property notes The manager's write-up of one-time items and known drivers. Traceability ## Follow any number back to its clause. Two figures from the letter above, traced end to end. Permanent ### A vacancy that reprices the year Source Property notes: Kestrel Analytics Group vacated Suite 210 (11,400 SF) at expiry on **2026-04-30**. Annual base rent **$216,600**, or **$18,050/month**. Read Budget assumed the suite occupied all year. No replacement lease signed or in negotiation. Computed Eight months absent × $18,050 = **$144,400**. Reforecast Full-year base rent $4,952,000 → $4,807,600. Timing ### A favorable variance that isn't a saving Source Marketing and leasing came in **$8,100 under budget**, 22.5% favorable. Read Property notes: the spring campaign was deferred to Q3. Contract signed **2026-07-02 at $9,200**. Judged Classified **timing, not permanent**. The underspend is a deferral, not a saving. Reforecast Full-year budget held flat at $72,000. Delta $0. The second one is the point. A spreadsheet sees a favorable variance. An analyst at 11pm on a close deadline may take the credit for it. The agent reads the note, classifies it as timing, and refuses to reduce the full-year budget — because the money is going out in Q3. The obvious question ## Why not build this in ChatGPT, Claude or Copilot? You can, and you may already have. The materiality rule, the timing calls, the reforecast logic — that all goes in a prompt, a project or a skill, and a good analyst gets there in an afternoon. It holds for one property. Here is what it doesn't survive. 01 #### It breaks at portfolio volume The letter above is one asset. The quarter is the portfolio. Prototypes start erroring out around the twentieth document and skim long files rather than read them. The close date doesn't move. 02 #### It repeats stale answers with no warning Run the next set of properties and the figures can come back identical — the new files never actually read. No error, no flag. In a quarterly cycle that is last quarter's numbers in this quarter's letter. 03 #### It leaves bad scans blank without flagging You tested on a clean export. The portfolio holds a scan of a fax. A prompt-built agent leaves those fields blank and says nothing, so the gap travels all the way to the letter. 04 #### You build and maintain the delivery pipeline Extraction is the easy part. Delivery into your scorecard template, the retries, the monitoring — and a model underneath that drifts, gets deprecated and reprices. You build and staff every bit of it. ### The prompt is about 5% of a working agent. 5%The prompt 95%**Everything that keeps it right** Model evaluationVersion controlGround-truth validation Live monitoringMigration when the model changes Delivery into your template Kolena runs the other 95%. Output lands in your template on every run, cited. [See the full comparison →](/kolena-vs-claude/) What it extracts ## Thirteen fields out of the box. Not a fixed schema. Every prompt is editable, fields can be added or dropped, and the materiality rule, scorecard lines and letter structure follow your policy rather than ours. Form #### Report Overview Property, type, owning entity, period, source system and basis, so a reviewer can orient before reading the variances. Text #### Reporting Period The quarter and year the export covers, normalized so every property's report and letter align. Classification #### Source Accounting System Which PMS produced the export, since each labels budget, actual and variance columns differently. Table #### Line-Item Budget Variance Full GL-line detail for quarter and year to date, variance and percent verified, favorability set by NOI impact, each line flagged against the materiality rule. Table #### Revenue and NOI Variance Bridge Line items rolled into a GPR-to-EGI-to-NOI bridge on a budget-versus-actual basis, with a footing check that each subtotal ties out. Table #### Material Variance Flags The short list of variances that clear the materiality threshold, ranked by dollar magnitude. Table #### Variance Drivers and Explanations For each material variance, the driver category and a grounded explanation, classified timing versus permanent, with a recommended action. Table #### Reforecast Recommendations Full-year guidance for lines whose variance is permanent or a reliable run-rate signal, with the revised budget, the delta and a confidence level. Classification #### Overall Budget Performance A single-label verdict on the property for the period, based on NOI variance and whether revenue and expense are offsetting. Form #### Variance Summary Scorecard The one-look scorecard that populates the property's row in the portfolio roll-up. Text #### Investor-Letter Variance Commentary The drafted narrative for the quarterly investor letter — the deliverable the cycle exists to produce. Table #### Data Quality and Validation Flags The footing, sign, classification and threshold checks an analyst runs before trusting the report, each marked Pass, Fail or Review. Text #### Analyst Notes Classification judgments, undetermined drivers and source-versus-recomputed disagreements, so a reviewer knows where to double-check. Yours #### Whatever else the quarter needs New extractions, your own prompts, your own output schema — configured to how your team actually reports.