Condition reports and CapEx plans that update together

01The challenge

Earl Kendrick's condition reports were written in Word and its CapEx plans built in Excel, separately. When one changed the other didn't, so PDFs fell out of step with spreadsheets. Clients received a static PDF and filed it, and EK never found out what work was done, what it cost or whether the forecast had been right.

The cost history was there but unusable: roughly 950 cost plan spreadsheets going back to 2017, in templates that changed year by year. Some used calendar years and some 'Year 7' of a cycle; most had one sheet, one had twenty; M&E was split out or mixed in; cells read 'inc above' instead of a number. Each new plan started from scratch instead of learning from the last hundred.

02What we did

1. 01December 2025 — scoping with EK on how condition reports and cost plans are produced, and what happens to them after issue
2. 02January 2026 — a three-phase plan: Capture for EK's surveyors, Engage for managing agents, Derive for benchmarking and prediction
3. 03March 2026 — EK's asset headings mapped to RICS NRM3 and BCIS, then consolidated into three levels so a surveyor reaches any component in three taps
4. 04Phase I — one data model for condition and cost, linked at the asset level, with combined report generation
5. 052026 — a pipeline specified to bring the historical cost plans into the same taxonomy, with a surveyor reviewing anything the machine is unsure of
6. 06July–August 2026 — Phase II: property manager user stories, then a forecasting tool with graph, elemental analysis and CapEx breakdown views

03What we built

Condition observations and cost forecasts are held once, against the asset. Surveyors classify each element with a consolidated, three-level version of RICS NRM3, so a roof covering, a lift or a run of rainwater goods means the same thing on every job. The same data produces the condition report narrative and the CapEx spreadsheet view, so the two cannot drift apart.

The forecasting tool shows a plan three ways: a graph of spend over time, an elemental analysis of cost by element, and a year-by-year CapEx breakdown. EK locks its 10, 15 and 25-year scenarios; property managers build their own alongside, moving works between years and testing them against a budget without touching EK's version.

For the archive we specified a four-layer pipeline: every cell kept as it appears in the source file, tidy line items with their headings carried down, categorised items with a confidence score, and a human-approved truth set. Rules handle the common phrases and an AI model classifies the rest in batches. Anything uncertain, or anything where rule and AI disagree, goes to a reviewer, and each correction becomes a new rule or example.

• Condition and cost linked per asset
• Report and CapEx views from one dataset
• Three-level RICS NRM3 classification
• Combined condition report generation
• Forecast scenarios over 10, 15 and 25 years
• Graph, elemental and CapEx breakdown views
• Property manager scenarios beside EK's locked plans

04The result

Condition and cost now sit together in EKi, classified to the industry standard rather than to whichever template a surveyor last used. Update an element once and both the report and the plan change, and because every plan shares one taxonomy, plans can be compared by building type, region or element.

Phase II, now in build, puts the plan in the property manager's hands instead of a filing cabinet, and once works are done they record what was actually spent, which feeds EK's benchmarks. That dataset is what Phase III needs: BCIS benchmarks, data exports for TPI and predicted costs for buildings EK has never surveyed.