Back-office audit: a bureau of technical inventory

This is the 5,000 EUR AI Profit Audit. All figures are in Ukrainian hryvnia (UAH) unless marked EUR (one EUR taken at about 48 UAH, an assumption flagged where it matters). English working master; client-facing version issued in Ukrainian. Client-provided facts and our own estimates are clearly separated in Section 1.

0 Verdict · 1 Scope & data reliability · 2 Financial baseline · 3 Operating model & labour · 4 Process value stream · 5 Cost-to-serve teardown · 6 Throughput & queue · 7 Technology & data · 8 Risk register · 9 Opportunity map · 10 Spearhead · 11 90-day plan · 12 What acting looks like · Appendices

Scorecard: how your back office measures up

1   Scope, methodology and data reliability

Scope. A back-office operational and financial audit covering the firm's services, volumes, finances, organisation, processes, systems, and the automation opportunity. We reasoned only from the firm's own documents: financials read from the spreadsheets, turnaround recomputed from the raw timestamps, headcount taken from the firm's own role list.

Documents examined (immutable copies retained)

Data reliability: fact vs estimate

2   Financial baseline (the P&L teardown)

Source: the firm's 12-month management P&L. Absolute values are withheld and the structure shown as shares, so the client cannot be matched to public filings.

Cost structure (share of total operating cost)

3   Operating model and labour analysis

Headcount: about 50 staff plus a few contractors. The role mix is roughly 45% production (engineers, technicians and land surveyors), a large intake and registrar function, and a small management, legal and accounting layer. The administrative layer is large relative to production. (discovery-answers.docx, job-descriptions)

Utilization. About 1,750 worked hours per head a year, at roughly 410,000 UAH of fully-loaded cost per head. We value every recovered hour at the firm's own blended rate of about 240 UAH.

4   Process value stream

The whole chain except field measurement (step 3) is document and data movement; the constraint concentrates at step 5.

5   Cost-to-serve teardown (from real orders)

A private-sector case, a house with five outbuildings. Line costs are shown as a share of the order's billed price rather than in hryvnia, so the firm cannot be matched to its published tariff list:

6   Throughput, queue and capacity

Source: a two-month order export from the firm's own system.

• How long an order takes: half of all orders are finished within 31.6 days, and the average is 34.2 days; 9 in 10 are done within 52.6 days, and the slowest ran to about 280 days (recomputed by us from the raw timestamps). By elapsed days: roughly 700 orders took 15 to 30 days, about 1,100 took 31 to 60 days, and about 160 took over 60 days.
• Where the delay sits: active registry entry is 3 to 6 hours per object, then the order waits about a month. That month is the internal backlog (orders queued for an engineer), not the registry being down: the registry processes well over a thousand orders a month, so it works most of the time and is only intermittently down. Almost the entire cycle is waiting in your own queue, not work, so the constraint is throughput (engineer capacity), and registry instability matters because it wastes engineer hours and so feeds the backlog.
• Orders open at any one time: at roughly 1,600 orders a month and a roughly one-month cycle, about 1,600 orders are open in the system at any moment (in the pipeline, not actively worked), by the standard queue calculation (Little's Law). Each is a client waiting and a balance payment not yet collected.
• Not in steady state: intake rose by about 15% between the two months, and a sharp demand surge began April 2026. Arrivals outpace completions, so the queue and open-order count are rising.
• Volume reconciliation: the firm states roughly 1,600 orders/mo; the earlier two-month export shows about a quarter fewer. These reconcile as demand growth (the export predates the April 2026 surge). The pre-surge passport run-rate was about a quarter below the rate the firm now reports; we size on the current rate and carry the lower one as a downside sensitivity (Section 10).

7   Technology and data audit

• Inventory-case & order management: bespoke in-house Ukrainian modules; they hold the registry object identifier and export to PDF and Excel, the integration surface for the build.
• Land survey: Digitals XE (Ukrainian) + the Public Cadastral Map. Accounting: BAS + M.E.Doc.
• Registries: the state construction registry (daily); the Property Rights Register, direct access restricted by Law No. 4576-IX of 21.08.2025 (caps some ownership lookups); the State Land Cadastre.
• Data: a secure local server plus a paper archive of roughly 190,000 cases, and a modest annual spend on registry access fees.

8   Risk register

9   The opportunity map (ranked, multi-year)

Conservative-to-base ranges, valued at the firm's own blended hourly cost. The total is a multi-year build-out ceiling, addressed in phases; estimates overlap at the margin and need demand to absorb freed capacity. Only the spearhead is committed and guaranteed.

Quick Wins: three moves you can make this month, free, without us

Before any build, three moves you can start this week, free, with no software and without us. Their honest values are below. Two are low-risk and immediate (a registry-timing tweak and an intake checklist); the third, capturing your rush premium, is the largest but is conditional, so confirm it before you count on it.

• 1. Check whether you are capturing your own rush premium. Your tariff doubles the price for delivery within 3 working days, counted from the on-site inspection (rush coefficient of 2). We could not verify how much of this you already capture: the order export carries no prices, and the 3-day clock runs from inspection, not from intake, so the rush-qualifying count cannot be computed from the data. So the first step is a question to yourselves: what share of orders is already billed at the double rate? If the answer is "little", offer 3-working-day delivery selectively to a few willing clients whose orders you are confident you can finish in time (submitting to the registry in its stable hours, move 2), and measure how often you hit the deadline. Note this reallocates capacity rather than adding it (an expedited order pushes the others back), so it is a margin move for a handful of orders, not a throughput fix; reliable, scalable express arrives with the build. For scale, 10 apartment orders a month at the double rate would be about 35,000 UAH/month (~425,000 UAH/year) of extra revenue (ten orders at the firm's average ticket).
• 2. Schedule submissions for the registry's stable hours. You report the biggest delay is when the registry is down. Batching submissions into its low-load windows (early morning, late evening) cuts failed attempts and engineer wait. Even 0.2 h saved on half your passports is roughly 18,000 UAH/month of engineer time. A manual slice of Section 10 you can capture today: it frees engineer time (it is part of the build's recovery, not separate cash).
• 3. Hand intake clerks a one-page document checklist. Incomplete packets cause rework and repeat visits. A checklist that gets the packet right first time saves admin and engineer time. Even on 5% of orders at half an hour each, that is roughly 9,000 UAH/month (about 110,000 UAH/year) of saved cost, plus less client friction.

10   The spearhead: state construction registry entry

The one process: automate the entry of object data into the state construction registry and the submission and retry against the registry. It is the largest billed labour line (Section 5), high volume (~760 passports/mo), and the cleanest to measure.

Scope, the honest boundary. We cannot make a government server faster. We remove the part you control: pre-filling the entry from object data already in the in-house system (removing re-keying), and a queue that prepares submissions and retries failed ones and reports status (removing the engineer's babysitting). Registry processing and downtime are excluded from the saving.

We headline the conservative-to-base band (~0.65 to 1.13M UAH/yr, 1.6 to 2.7 FTE). The firm's own tariff assigns 3 to 6 normo-hours to this step per object, so recovering 0.5 to 1.0 hour is deliberately a fraction. The number to commit to is set from a two-week measured baseline at the start of the build.

Volume sensitivity. If the pre-surge run-rate (roughly a quarter lower) is closer to the forward baseline than the rate the firm now reports, the conservative case falls to about 0.5 million UAH/yr. The volume is confirmed against the live run-rate in week one before the guarantee target is set.

11   The 90-day plan to production

12   What acting looks like

The first step is the spearhead: registry-entry automation, in production in 90 days, with a visible result by week two (Section 11). The freed capacity absorbs the April demand surge without new hires and begins shrinking the 31.6-day queue; from there the rest of the opportunity map (Section 9) is worked down in sequence.

Appendix   Calculations, assumptions and sources

Calculations. The blended hourly cost is total payroll over total worked hours, then divided by headcount for a fully-loaded cost per head. Operating margin is turnover less total cost. Work in progress is monthly order intake multiplied by the cycle time in months. The registry line is expressed as a share of the order's billed price, of its excl-VAT labour, and in normo-hours at the firm's own tariff rate. The spearhead range is annual passport volume multiplied by recoverable hours per passport and by the automation rate, valued at the blended hourly cost; certificates follow the same form. Hire-avoidance is recovered full-time equivalents multiplied by fully-loaded cost per head. The client's absolute figures are withheld from this published version; every ratio and percentage in the document is computed from them unchanged.

Assumptions: recoverable 0.5 to 1.0h/passport (fraction of the firm's 3 to 6h / tariff normo-hours); automation 60 to 75% (under the firm's own electronic-version discounts); passport volume ~760/mo (excludes garden/garage/non-res = upside); EUR/UAH ~48 (affects only EUR tiers and the audit fee); drafting automation 25% (lower confidence). Each is confirmed against the week-1 baseline.

Sources: discovery-answers.docx, income-and-expenses.xlsx, order-intake-delivery-dates-sep-oct-2025.xlsx, tariffs-price-list.pdf, inventory-case-private-sector.pdf, inventory-case-non-residential.pdf, sample-information-products.pdf, blank-forms-orders-contracts-invoices.pdf, job-descriptions-1/2/3.pdf, technical-passport-extract-private-sector.pdf. External: Cabinet of Ministers of Ukraine (kmu.gov.ua).