EVS valuation app: vibe coding and the biggest transformation since Excel

On 23 October 2026, at the TEGoVA conference in Vienna, we will build a valuation app according to the European Valuation Standards (EVS) live on stage – in a few minutes and without a development team. What makes this possible is a way of working that has had a name since 2025: vibe coding.

TEGoVA 2026 · Take-away

The master prompt from the talk

This is the prompt that builds the EVS valuation app live on stage. Open a new chat in Claude, paste the prompt in full and send it. After about seven to ten minutes the MVP is ready; in the app, simply click “Load demo case”.

Download prompt (.txt)

View prompt

You are building a working MVP for a live demo in front of professional property valuers at a TEGoVA conference. Build it as ONE self-contained HTML artifact (vanilla JavaScript, no external dependencies except Google Fonts). It must work flawlessly on first render.

PRODUCT
"EVS Valuation Workbench" – guides a qualified valuer through the Market Value valuation of a small multi-family residential building, structured according to the European Valuation Standards 2025 (EVS 2025, TEGoVA). The tool does NOT replace the valuer: every judgement (adjustments, rates, weights, final value, descriptions) is entered or confirmed by the valuer. Footer line: "Decision support only – the qualified valuer (EVS 3) remains responsible for the valuation."

LAYOUT
- Header: app name, DE/EN language toggle (all labels, messages and the report switch language), "Load demo case", "Clear", "Generate EVS report".
- Left: workflow in 8 numbered steps (stacked cards). Right: sticky sidebar with (a) live result card (Market Value, €/m², income, DCF, comparison value, deviation, gross initial yield, range) and (b) "EVS Compliance Check".
- Professional, high contrast, large font (projected). Light theme. Tables scroll horizontally inside their own container.

STEP 1 – ENGAGEMENT (EVS 4 / EVS 5)
Client, purpose, basis of value (default Market Value, EVS 1), valuation date, inspection date and extent, valuer name and qualification (REV/TRV), conflict-of-interest declaration (checkbox + text), assumptions, special assumptions.

STEP 2 – SUBJECT PROPERTY + AI DESCRIPTION FROM PHOTOS
Fields: property type, address/location, site area (m²), year built, condition (select), rent regulation status, HWB (kWh/m²a), fGEE, energy class (A–G), heating.
Below, a highlighted panel "Location and property description from photos" (label "AI draft"):
- Photo upload (file input, multiple, JPEG/PNG/WebP, drag and drop) with thumbnails and a remove button per photo.
- Button "Generate description": sends the photos plus the property data above to Claude, using the AI capability available to artifacts in this environment, and asks for JSON {"location": string, "property": string, "to_verify": [string]}. Instruction to Claude: describe only what is visible or given; no invented figures, brands, distances or defects; macro-location from the address only and marked "to be verified"; neutral report style; 120–180 words each; answer in the current UI language.
- Show "Claude is analysing the photos …" while running, a Cancel button, and a clear error message per failure.
- If the AI capability or image analysis is not available, show a clear message and keep the feature usable via a "Load sample text (backup)" button that fills both fields with a prepared fictitious text.
- Two editable textareas "Location description" and "Property description", each marked "AI draft – unchecked" until confirmed; a list "For the valuer to verify" from to_verify.
- Checkbox "I have checked these descriptions and take responsibility for them (EVS 3)". Regenerating resets it.

STEP 3 – RENT ROLL
Editable table (add/remove rows): unit, floor, area m², tenant, lease start, fixed term until (empty = open-ended), vacant (checkbox), contract rent €/m²/month (disabled when vacant), market rent €/m²/month. Computed per row: monthly rent (vacant units at market rent), deviation contract vs. market (green over-rented, red under-rented). Footer totals. Tiles: number of units, net floor area (sum of units), vacancy rate by area, average contract rent, contract vs. market rent, gross annual rent.

STEP 4 – INCOME APPROACH (Austrian method)
Inputs: rent loss risk % of gross rent, maintenance €/m² p.a., capitalisation rate (Liegenschaftszinssatz) %, total economic life (years), land value €/m², vacancy until let (months).
gross annual rent = Σ monthly rents × 12; operating costs = gross × risk% + maintenance × area; net income = gross − costs; land value = site area × land €/m²; land value interest = land value × i; building net income = net income − land value interest; remaining life = total life − (valuation year − year built); multiplier = (1 − (1 + i)^−n) / i; vacancy deduction = Σ (vacant area × market rent) × months; income value = building net income × multiplier + land value − vacancy deduction. Show each line as a tile.

STEP 4b – DCF APPROACH WITH WATERFALL
Inputs: indexation of contract rents % p.a., market rent growth % p.a., cost inflation % p.a., discount rate %, exit capitalisation rate %, exit sale costs %. Reuse rent loss risk, maintenance and void months from Step 4.
Monthly simulation over 11 years from the valuation date, per unit: until lease expiry (month after "fixed term until"; open-ended = never) contract rent indexed annually; after expiry the void months with zero income, then market rent indexed annually. Vacant units start with the void. Potential rent = what each month would earn without void; void loss = potential − received.
Per year t = 1..11: operating costs = received × risk% + maintenance × area × (1 + cost inflation)^(t−1); net income = received − costs; discount factor = 1 / (1 + r)^t (end of year).
Exit value gross = year-11 net income / exit cap rate; net = gross × (1 − sale costs); PV exit = net × discount factor of year 10. DCF value = Σ PV net income years 1–10 + PV exit.
Show: (a) a WATERFALL chart (inline SVG, no library) bridging PV potential rent → − PV void loss → − PV operating costs → = PV net income (subtotal) → + PV exit → = DCF value (total), positive bars in the accent colour, negative in red, subtotal grey, total dark, value labels on the bars, dashed connectors; (b) tiles incl. exit share of value and deviation DCF vs. income value, plus the sentence "x% of the DCF value comes from the exit"; (c) the yearly cash flow table plus an exit row and the DCF total; (d) a 3×3 sensitivity grid: discount rate ±0.5 pp × exit cap rate ±0.25 pp, base cell highlighted.

STEP 5 – SALES COMPARISON APPROACH
Editable table of comparable multi-family house sales (add/remove): ID, price, net floor area, sale date, energy class, % adjustments for location, condition, features, size, age.
price per m²; time adjustment = days from sale to valuation date / 365 × market index (default 3.0% p.a.); energy adjustment from Step 6; total = sum; adjusted €/m² = €/m² × (1 + total); gross adjustment = sum of absolute values; weight = 1/gross, normalised. Results: mean, weighted mean, standard deviation, coefficient of variation; comparison value = weighted mean × net floor area of the rent roll. Rows with gross adjustment > 15% in amber.

STEP 6 – ENERGY EFFICIENCY (EVS 6)
Adjustment per class step (default 2.0%, labelled "illustrative – replace with your own market evidence"). Energy adjustment = (comparable class index − subject class index) × step, A..G = 0..6. Mini table per comparable.

STEP 7 – RECONCILIATION, UNCERTAINTY, REPORT
- Weights for income value (default 70%), comparison value (30%) and DCF value (0%, plausibility check), normalised to 100%. Show all three values in the sidebar. Deviation income vs. comparison value with traffic light (green ≤ 10%, amber ≤ 20%, red > 20%).
- Market value = weighted result rounded to the nearest 10,000 €. Range = market value × (1 ± coefficient of variation). Uncertainty comment field.
- "Generate EVS report": a report view inside the page (no window.print – the artifact frame blocks it; add "Copy report text"). Sections: 1 Engagement and purpose, 2 Basis of value and valuation date, 3 Valuer and independence, 4 Location (AI-assisted text), 5 Property description (photo thumbnails + text + note whether the AI text was confirmed by the valuer), 6 Energy performance (EVS 6), 7 Tenancy situation (rent roll table), 8 Income approach (full derivation), 9 DCF approach (parameters, waterfall, key results), 10 Sales comparison approach (table), 11 Reconciliation and Market Value, 12 Valuation uncertainty, 13 Assumptions, 14 Signature. Marked "DRAFT" while compliance items are missing.

EVS COMPLIANCE CHECK (sidebar, live)
14 items: client, purpose, basis of value, valuation date, inspection, valuer qualification, conflict-of-interest declaration, assumptions, location and property description present, description checked by valuer, energy class, rent roll entered, at least 3 comparables, uncertainty comment. Green check / red cross, progress bar "x / 14".

DEMO CASE (fictitious – embed and load with "Load demo case"; leave conflict-of-interest declaration, uncertainty comment and both descriptions EMPTY so the check starts at 10/14)
Client Musterbank AG, Vienna (fictitious); purpose secured lending; Market Value; valuation date 2026-10-23; inspection 2026-10-20 internal and external, all units except Top 5; valuer Dr. Maria Muster (fictitious), REV; assumptions "No contamination; title free of value-affecting encumbrances; areas per plans; leases as provided"; special assumptions "None".
Subject: multi-family house, 6 units; 1210 Wien-Floridsdorf, quiet residential street (fictitious); site 700 m²; built 2005; condition good; partial application of MRG, free rent setting; HWB 55; fGEE 0.85; class C; gas condensing central heating.
Rent roll (unit | floor | m² | tenant | start | fixed term until | contract €/m² | market €/m²):
Top 1 | EG | 62 | Tenant A | 2019-03 | open-ended | 13.80 | 14.50
Top 2 | EG | 58 | Tenant B | 2022-09 | 2027-08 | 14.50 | 14.50
Top 3 | 1. OG | 75 | Tenant C | 2016-05 | open-ended | 12.90 | 14.20
Top 4 | 1. OG | 68 | Tenant D | 2024-02 | 2029-01 | 15.20 | 14.80
Top 5 | DG | 82 | VACANT | – | – | – | 16.50
Top 6 | DG | 71 | Tenant E | 2023-06 | 2028-05 | 15.80 | 15.50
Income parameters: rent loss risk 2%; maintenance 10 €/m² p.a.; capitalisation rate 3.0%; total economic life 80; land value 800 €/m²; vacancy 6 months.
Comparables (price | m² | date | class | location | condition | features | size | age):
H1 1980000 | 410 | 2026-02-10 | C | 0 | 0 | 0 | 0 | -2
H2 2350000 | 520 | 2025-12-05 | B | +3 | -2 | 0 | +2 | -3
H3 1720000 | 380 | 2026-05-20 | D | 0 | +4 | 0 | -1 | +2
H4 2150000 | 445 | 2026-07-01 | C | -3 | 0 | -2 | 0 | 0
H5 1880000 | 400 | 2026-04-15 | E | 0 | +3 | +2 | 0 | +5
DCF parameters: indexation 2.0%, market rent growth 2.0%, cost inflation 2.0%, discount rate 5.0%, exit cap rate 3.5%, sale costs 2.0%.
Market index 3.0% p.a.; energy step 2.0%; weights income 70% / comparison 30% / DCF 0%.
Expected results for checking: net floor area 416 m²; gross annual rent 74,070 €; income value ≈ 1,972,000 €; DCF: year-1 net income ≈ 59,650 €, PV net income ≈ 546,900 €, PV exit ≈ 1,421,900 €, DCF value ≈ 1,968,800 € (exit share 72%); weighted mean ≈ 4,821 €/m²; comparison value ≈ 2,005,700 €; deviation −1.7%; Market Value 1,980,000 €; H5 flagged amber.

QUALITY
All calculations recompute live on every change. German number style in DE (1.980.000 €), English style in EN (€1,980,000). Banner "Demo data – fictitious" when the demo case is loaded. No placeholder text, no "TODO", no console errors.

In English, approx. 1,850 words. All demo data are fictitious. The app is decision support – responsibility remains with the qualified valuer (EVS 3).

What is vibe coding?

In vibe coding you describe in natural language what an application should do. An AI writes the code, you test the result, give feedback and improve it step by step. Instead of programming line by line, you have a conversation about the desired outcome. The professional question moves to the centre; syntax fades into the background.

Andrej Karpathy and the term

The term was coined by Andrej Karpathy, one of the best-known AI researchers in the world. He was a founding member of OpenAI, led AI development at Tesla and later founded the education start-up Eureka Labs. In early 2025 he described how he builds software by giving in completely to the AI and barely reading the code himself. The term spread rapidly.

Why we will build our own apps from now on

Until now, software for professionals was a compromise: you bought a standard product and adapted your way of working to it. When an app takes hours instead of months to build, that reverses. A valuer shapes the tool exactly around their own method, templates and quality standards – and can adapt it at any time.

The end of SaaS?

Software as a service – rented software from the cloud – was the dominant business model for twenty years. Its promise: nobody has to develop anything themselves. That promise loses value when development costs almost nothing. SaaS will not disappear overnight; infrastructure, data storage and security remain important. But many specialist tools people pay for monthly today, they will build themselves tomorrow.

Everyone their own startup

Anyone who can describe a professional question precisely can build a product today. Decades of practical knowledge become a tool of your own that you can use, share or sell. The bottleneck is no longer the code but the expertise.

The biggest transformation since Excel

With the spreadsheet, in the late 1980s, suddenly anyone could calculate valuations, cash flows and scenarios themselves. Vibe coding is the next step: the spreadsheet becomes the application, the formula becomes the tool. Real estate is facing its biggest transformation since Excel.

Responsibility stays with people

A prototype replaces neither professional review nor the valuer’s responsibility. Before productive use it needs testing, verified calculation logic and documentation. AI speeds up the build; the judgement stays with the valuer.