Use evidence and AI to understand the disagreement, test your own assumptions and take a better-informed next step.
Based on Adam Street’s experience at Hafer Road. Educational guidance, not a decision on your insurance cover.
sharing the repair problem
without warranty cover
net settlement accepted
Our neighbours rebuilt their Battersea block together. Years later, water was coming into homes, the main structural warranty claim was declined and two owners had lost cover because another insurer failed. As a resident and director, I had to help untangle reports, policy wording, costs and difficult collective decisions.
We accepted a settlement and completed the works. AI helped me organise evidence, question explanations and prepare correspondence. Advisers, contractors, directors and neighbours all played a part. AI did not replace them.
Earlier claims had been paid before the main structural repair claim was declined. The accepted settlement reconciles as £656,058.47 agreed cost basis, less £62,325.55 uninsured shares and £17,710.17 excesses, giving £576,022.75 net. This is not the total project budget or a cash-received date.
The outcome is a single structural warranty case, not a typical buildings-insurance result. We cannot isolate AI’s contribution or promise the same result for another claim.
Case basis: Adam’s account and acceptance confirmation of 22 October 2026, retained settlement breakdown and CPA addendum of 13 February 2024. Private underlying documents are not reproduced here.
Read the wider Hafer Road storyMake a brief and find the gaps.
Build my brief →02Check the report, policy and your own argument.
Test my position →03Make a focused request and plan delivery.
Prepare my next move →Keep urgent safety action, damage limitation, notification and relevant deadlines moving while investigating. AI should not delay them.
Write one page: what happened, who is affected and the decision you need. Separate urgent work from the disputed claim.
A brief a new reader can understand.Index the policy, endorsements, reports, photos and correspondence. Give each fact a dated source and page. Keep contradictions.
A timeline with visible gaps.Check basic facts and whether the conclusion follows from the observations. Ask what test would distinguish competing explanations.
A specific error or unanswered question.Read the cited clause with its definitions, exclusions and amendments. Identify which facts the insurer relies on.
A clause-to-evidence comparison.Request a correction, inspection or review. State the strongest evidence, practical consequence and response date.
One focused request to a named person.Reconcile the settlement with scope, uninsured amounts and excesses. Confirm authority, funding, technical checks and resident updates.
An agreed route from decision to repairs.Choose the explanation you need. These are existing educational videos, not endorsements of Building Trust. Each has a written takeaway so watching is optional.
IBM’s Martin Keen uses a refund example to explain the distinction. Apply it to your settlement arithmetic and policy interpretation.
Loading connects to YouTube. Watch on YouTube instead.
A calculation needs checked inputs and rules. AI interpretation needs evidence and scrutiny. Both can be wrong. List one task for a spreadsheet, one for AI-assisted comparison and one needing expert judgement.
“Einstein in your basement”, as used in TIP PERSONAL preparation. Watch 01:02–06:00 and 12:54–15:54.
Loading connects to YouTube. Watch on YouTube instead.
Try asking for comparison, critique and questions rather than a final answer. Give AI your intended outcome and the documents it needs, then ask what it lacks.
Generative AI constructs responses from learned patterns and available context. It can produce convincing errors, omit exceptions or invent references. Tools can add searching and calculation, but results still need checking.
Your “room” of context matters. AI may suggest connected issues, but cannot establish facts you did not supply. A missing endorsement or contradictory report can change the answer. Asking what is missing helps; it is not a completeness guarantee.
Watch or read 3Blue1Brown’s visual explanationUse non-sensitive notes. Nothing typed here is sent to Building Trust or an AI service, and it is not saved when you leave. Download your notes before closing.
Remove unnecessary personal data before using an external AI service. Check its terms and your authority to upload documents.
The aim is a sound decision. Sometimes that means changing your position or stopping a claim.
First check simple facts: dates, floors, location, materials and which inspection actually occurred. Then test whether the conclusion follows.
At Hafer Road, damaged decking above polythene helped challenge the causal account. It did not establish cover by itself. A competent expert still had to assess technical conclusions.
Ask for exact wording and the facts said to bring the claim within it. Read definitions and endorsements together. Do not treat a search snippet or AI paraphrase as the policy.
Keep physical cause, policy coverage, lease responsibility and funding as separate questions.
Use a simple table: statement → source/page → conflict → question → person qualified to resolve it. Distinguish factual errors from technical disagreement.
This is a critique exercise, not inside knowledge. Keep the eventual correspondence factual and proportionate.
Please decide: [the correction, inspection or review].
The evidence is: [source and page].
This matters because: [practical consequence].
Please respond by: [date] to [named lead].
Give the specialist your brief, exact clauses, indexed evidence, alternative explanation and the decision needed. Ask for scope and fee before commissioning work.
Show what is covered, what remains unfunded, who can authorise spending and the choices ahead. Record decisions and explain changes.
Adam used AI for contractor research, tender documents and evaluation, Section 20 documentation, method-statement review and resident notifications.
0 of 5 personal checks marked. This is not a legal or safety assessment.
The original Hafer Road report was authored by an MCIOB surveyor. The later addendum was reviewed by an MRICS/MCIOB surveyor. Do not assume every report is governed by the same professional standard.
TIP: bias in the conversation · PERSONAL preparation · ADVANCED preparation.
Frame the problem, outcome and context. Ask AI to improve the request and identify gaps. Verify sources and calculations, then assess whether the result actually helps its recipient decide.
The longer guide contains source-pack, report, policy, escalation, funding and delivery tools. Use the section you need.
Download the detailed toolkitResources selected 22 October 2026. External pages may change. The organisations linked here do not endorse this guide.
The separate lease-reading error in Adam’s case helped inspire LEASE-iQ. Individual leaseholders can request free access for one residential lease. Managing agents can use it for their own lease questions and explore access for residents. It is not an insurance coverage adjudicator.
Ask Adam about accessBLOCK-iQ is being co-created around the wider shared-building record and decision problem.
What did this guide leave unanswered? A brief description is enough; please do not email private case documents with initial feedback.