Construction businesses do not get paid for how busy they feel. They have to turn a promise into a completed building, a recoverable margin and cash that arrives before the next obligation falls due. Studying where that chain broke for someone else is one of the most practical ways to improve your own business.
That is the purpose of TEMRIK’s free ebook, Find the Solution Before Failure Finds You. Its 25 chapters turn patterns of business deterioration into warning signals, review questions and first actions. Read this way, a book about failure becomes a practical guide to construction success: choose better work, protect the margin, fund delivery, resolve dependencies and act while options remain.
The named AI workflows in the ebook are proposed designs to evaluate and configure, not a claim that all 25 are available as ready-made features.
Learn from the failure mechanism, not just the headline
A collapse headline gives you an ending. A useful investigation asks what happened before it: which assumptions changed, what evidence was available, who could act and when the remaining options narrowed. The objective is to identify a decision you can improve in your own business.
The ebook’s documented case discussions illustrate different questions. Its Carillion discussion draws on parliamentary findings about adverse internal evidence and optimistic contract reporting: could a project team’s warning actually change the reported forecast? Its PBS Building discussion draws on an administrator’s preliminary assessment of working-capital constraints: was a potentially recoverable claim being confused with cash available now? Its WBHO Australia discussion uses the parent’s account of losses, unrecovered claims and withdrawal of support: how dependent was the portfolio on funding continuing?
Those are different types of evidence, with different limits. The ebook also uses Crossrail as a delivery and integration example, not as a corporate insolvency. Its source notes let readers follow the underlying records. None of these cases establishes that AI would have saved the company or project. They do identify questions worth asking before the same exposures become expensive in your business.
The central principle is to make a lesson operational. “Watch cash” becomes a dated forecast, verified receipts, a downside scenario and an owner. “Improve coordination” becomes a release gate with resolved interfaces and approved drawings. Knowledge becomes useful when it changes what happens next.
The 25 chapters: from failure pattern to better construction practice
The following summaries follow the ebook’s chapter order. Each pairs the failure mechanism with the habit that can improve outcomes and a concrete starting action. The 30-minute checks begin a review; they do not replace a full forecast, contract assessment or safety review.
01 Cash-flow failure
Profit and available cash answer different questions. A job can earn a margin while wages and suppliers fall due before the client pays. The proposed Cash Sentinel workflow assembles a 13-week forecast and highlights uncertain receipts. The success principle is to manage dated obligations, not rely on today’s bank balance. Start: move the largest uncertain receipt two weeks later and identify the first funding gap.
02 Margin erosion
A tender margin is an opening assumption. Unpaid commitments, remaining work, rework and extra preliminaries can consume it long before the accounts reveal the final loss. Margin Watch compares current expected final cost with recoverable revenue. The discipline is an honest cost-to-complete review. Start: reconcile one package’s actuals, unpaid commitments and uncommitted remaining cost, keeping disputed variations separate.
03 Fixed-price exposure
A fixed selling price does not freeze labour or material costs. The vulnerable part is often the work sold but not yet procured. A Contract Stress Test examines exposed packages and cost scenarios against the actual contract. Success means understanding the downside before accepting it. Start: classify major packages on three jobs as fixed, variable or unknown, then test increases only on the exposed amounts.
04 Contract risk
Obligations can enter through amendments and incorporated schedules that never reach the delivery team. Contract Intelligence compares the tender, qualifications and executed agreement, producing a referenced obligations register for qualified review. The principle is to know, price and allocate the promise you actually signed. Start: locate the complete executed versions of three major contracts and identify obligations without an operational owner.
05 Scope gaps
“By others” is not an allocation until somebody accepts it. Unowned interfaces and duplicated allowances create disputes and expensive late work. A Scope Comparator cross-checks scopes, specifications, quantities and current drawings. Success comes from resolving boundaries before installation restricts access. Start: review one trade interface and record who supplies, installs, tests and accepts each component.
06 Poor estimating
Repeating an old rate also repeats the assumptions hidden inside it. Access, sequence, complexity and supervision can make apparently similar work commercially different. Estimate Challenger uses completed-job evidence to question the next bid. The principle is to learn from actual delivery rather than copy yesterday’s spreadsheet. Start: compare one losing package across three completed jobs and add the recurring omission to the bid checklist.
07 Cost escalation
An expired quotation can quietly remain inside a forecast as though it were a commitment. Freight, accommodation, fuel and availability also affect delivered cost. Escalation Radar connects quote validity and unplaced orders to forecast exposure. Success requires updating time and cost together. Start: obtain written price and delivery confirmation for the ten largest unplaced orders.
08 Finance pressure
A facility limit is not the same as money available to draw. Existing borrowings, conditions, review dates and guarantees affect usable headroom. The Finance Scenario Engine organises those terms for verified calculations and professional review. The principle is to understand funding conditions before dependence becomes urgent. Start: prepare a debt schedule showing balances, available amounts, review dates and key conditions.
09 Working-capital shortage
Working capital finances the gap between spending and collection. Deposits, certification periods and retentions can make profitable work demanding to fund. The Growth Funding Model maps that operating cycle. Success means pricing and funding the delivery journey as well as the completed job. Start: map one project from its first outgoing payment to its final receipt and identify the deepest funding requirement.
10 Late payment
Submitted, assessed, approved and paid are separate states. A generic reminder does little when a claim is missing evidence or genuinely disputed. The Receivables Agent prepares a collection queue based on the actual blockage. The useful habit is specific follow-up with a credible receipt date. Start: give the ten largest unpaid items a reason, next action, owner and evidence-backed timing assumption.
11 Programme delay
A delay can increase overhead and postpone income at the same time. It may also lose a scarce trade or equipment slot. Programme Intelligence connects the programme with approvals, procurement and site progress, with scheduling calculations checked by competent planners. Success means managing prerequisites. Start: examine the next three weeks of critical work and assign each unresolved constraint a last useful decision date.
12 Design coordination failure
Two discipline drawings can each appear correct and still be impossible to build together. Revision confusion compounds that problem. The Design Conflict Engine flags conflicting information and routes specialist checks to suitable tools and reviewers. The principle is coordinated release before fabrication. Start: check one riser or plant room against current drawings, access requirements and unresolved technical questions.
13 Procurement failure
A quoted lead time may start after approved drawings or a deposit, not when the order is discussed. Procurement Radar joins approval, manufacture, inspection, shipping and installation dates. Successful buying protects the delivery sequence as well as the purchase price. Start: work backwards from installation for the five longest-lead items and verify every prerequisite.
14 Labour and productivity
Hours record effort; accepted quantities record useful output. Waiting, poor access and rework can consume skilled time without advancing the job. Productivity Monitor compares hours, quantities and constraint records. The principle is to remove blockers before adding people or assigning blame. Start: examine one repeatable task with its supervisor and identify the main reasons for lost productive time.
15 Poor management information
A reassuring report is dangerous when its dates, commitments and revenue assumptions cannot be reconciled. The Executive Signal Engine presents material movements with sources, owners and required decisions. Success depends on evidence being able to change the forecast, even when the result is uncomfortable. Start: trace the main numbers in the latest management pack to their source records and cut-off dates.
16 Administrative overload
Repeatedly copying status between emails and spreadsheets uses capacity needed for commercial and delivery decisions. The Work Elimination Engine identifies work to stop, simplify, automate or delegate. The principle is to remove unnecessary hand-offs while retaining required controls. Start: review yesterday’s tasks and eliminate one repeated entry of the same information.
17 Approval delay
An approval register is incomplete if it omits the information needed for a decision. The Approval Tracker connects prerequisites, responsible people, promised dates, expiry and programme consequences. Success means exposing the actual blockage, including internal omissions. Start: examine the five oldest approvals and establish what is missing, who can resolve it and which activity depends on it.
18 Safety-control failure
Possessing a signed form does not prove a physical control is in place. Changed activities and overlapping trades require competent attention. Safety Intelligence is a proposed way to flag missing information and conflicting activities, never to authorise work. The principle is verified controls at the workface. Start: review tomorrow’s highest-risk activity with the responsible supervisor, including who verifies controls and what stops work.
19 Defects and rework
Closing defect tickets individually can leave a repeated underlying fault untouched. Quality Intelligence groups recurring patterns and links inspection evidence for technical investigation. Success means changing the method that reproduces the defect and verifying the repair. Start: sample the last 20 closed defects, check acceptance evidence and assign a review of the most persistent category.
20 Unmanaged claims and disputes
Performing additional work and proving entitlement are different tasks. Instructions, notices, service records, costs and time effects need to remain connected. Claims Intelligence prepares an evidence chronology for commercial and legal review. The principle is to preserve the record while events are fresh. Start: trace the largest unresolved variation from instruction to response and mark every missing link.
21 Client concentration
One customer group may dominate debtors, pipeline and management attention as well as sales. Concentration Monitor maps these connected exposures using verified customer relationships. Success means understanding the effect of a delayed payment or lost award across the business. Start: rank customer groups by receivables and gross profit, then test the loss of the largest uncontracted opportunity.
22 Overtrading
Growth can exceed the funding and supervision available to deliver it. The Growth Governor brings project starts, cash requirements and delivery capacity into one portfolio view. The principle is to accept work at a pace the organisation can support. Start: map active and likely projects across 90 days, looking for overlapping funding peaks and overstretched supervisors.
23 Failure to adapt
Recurring friction deserves an experiment, but buying software is not evidence of improvement. The Ideas Engine turns repeated problems into small tests with measurable outcomes. Success means retaining changes that improve the work after checking costs and errors. Start: choose one persistent problem and set a two-week test with a baseline, owner, success measure and stop condition.
24 Knowledge loss
A business is fragile when critical decisions can only be explained by someone who is absent. Company Brain describes permissioned, source-linked knowledge and approved playbooks. The principle is transferable competence, not a bigger document dump. Start: identify ten tasks dependent on one person, capture one walkthrough and ask another competent team member to follow it.
25 Priority failure
A warning has little value when nobody makes the required decision. Failure Intelligence brings connected signals into a review of five priority risks, with evidence, trajectory, consequences, owners and decision dates. Success means closing the gap between knowing and acting. Start: ask which five unresolved issues most need a decision this week and who has the authority to make it.
Why playbooks make the lessons usable
A playbook maps what should happen when a recognisable situation occurs. It specifies the trigger, required evidence, checks, responsible person, permitted action, approval and proof of completion. That makes it a practical answer to a recurring cause of avoidable trouble: an important task depends on somebody remembering what to do under pressure.
A procedure that says “review variations promptly” leaves too much open. A usable playbook says that a new instruction starts a record; the current contract and drawings are attached; cost and time effects are assessed; an authorised person reviews any proposed notice; service evidence is retained; and unresolved items return to an escalation queue. The exact timing and requirements must come from the applicable contract and qualified advice.
The distinction between the three financial chapters illustrates why precision matters. Cash flow asks whether obligations can be met on their dates. Working capital asks how much funding the operating cycle consumes. Overtrading asks whether proposed growth exceeds available funding and delivery capacity. One generic “check finances” instruction cannot do all three jobs.
Playbooks can therefore help solve repeatable process failures. They cannot supply missing capital, make a loss-making contract profitable by themselves or remove an external shock. Their value is earlier recognition, more consistent execution and a clearer route to intervention. A badly designed playbook can repeat a mistake more efficiently, so it needs an owner, testing and revision.
AI prepares the work. A control layer governs the action.
AI can help bring the right knowledge into a live task: find the approved procedure, compare revisions, extract an obligation, assemble supporting records or prepare an exception brief. That reduces the distance between a senior person’s experience and the next team member who needs it. The objective is rapid, consistent application of approved knowledge across the team.
For an operating approach such as TEMRIK, the control layer should make the boundaries explicit. Who can access this project? Which procedure version applies? What evidence is mandatory? Which steps can run automatically? Who can approve an external commitment? What happens when a source is missing or two records disagree? These are implementation requirements to verify, not outcomes guaranteed by a product name.
| Layer | Required behaviour | Construction example |
|---|---|---|
| Approved knowledge | Show the current source, owner and revision. | Use the executed contract and approved drawing issue. |
| Deterministic rules | Check required fields, permissions and calculations outside free-form AI output. | Do not count a purchase order and its invoice twice. |
| AI assistance | Prepare referenced findings and show uncertainty. | Flag an instruction that may affect scope or programme. |
| Human approval | Route consequential decisions to an authorised, competent reviewer. | Review a contractual notice before it is issued. |
| Outcome record | Retain the decision, completion evidence and next review. | Record service, response and whether the issue remains unresolved. |
Strict execution means strict evidence and authority, not blind obedience to a model. Where a required check is incomplete, the configured workflow should hold the consequential action and escalate. A permitted exception needs a reason and an authorised approver. An unsafe, contradictory or obsolete instruction needs a stop and review path. People must be able to challenge the playbook.
Fast deployment is most useful after that discipline is established. An approved improvement can then reach every relevant project through a versioned workflow, clear notifications and training, rather than waiting for informal knowledge to spread. Audit records should show which version each team used and whether the change improved decisions. Speed without those checks can spread the wrong instruction just as quickly.
A practical example: one delayed approval, five connected risks
Consider an illustrative project where equipment approval is late. The manufacturer cannot release production, the installation slot moves, site overhead increases and a payment milestone slips. Approval, procurement, programme, margin and cash are now parts of one problem.
A useful playbook would identify the missing approval evidence, confirm the supplier’s actual release deadline, have the planner assess the revised sequence and ask finance to recalculate receipt timing. AI could prepare the evidence and questions. The project manager would coordinate the response; authorised commercial and financial reviewers would decide any commitments.
The action remains open until the team verifies the result. Sending a reminder is evidence of activity. Confirming the approval and a feasible delivery date is evidence that the blockage has changed. That distinction is how the guide moves from reporting failure risk to improving the work.
Use the guide to build one reliable process in 30 days
Begin with one recurring decision, such as the weekly cash review or long-lead procurement meeting. Name the owner, define the records they may use and describe an acceptable output. Establish how long the current process takes and where it misses information.
Run the proposed AI assistance alongside the existing process before relying on it. Compare outputs with an experienced reviewer. Test missing files, stale revisions, conflicting dates and unsupported conclusions. Measure review time as well as preparation time; a quick draft that needs extensive correction has not solved the problem.
Only expand after the checks are reliable and approval boundaries work. Track time to a reviewed decision, overdue material issues, missing evidence and repeat errors. Keep a way to pause the workflow. The ebook’s final section also supplies a five-risk review, a worked decision record and an assurance checklist to support this process.
Two useful external AI playbook resources
The Chartered Institute of Building’s Artificial Intelligence Playbook 2024 is directly relevant to construction teams. It covers practical implementation, data quality and the legal and contractual dimensions of AI, alongside potential uses. It provides useful wider reading when deciding where AI belongs in a construction workflow.
The NIST AI Risk Management Framework Playbook offers voluntary suggested actions organised around Govern, Map, Measure and Manage. It is useful for structuring how an organisation evaluates and manages the risks of its AI systems. NIST explicitly allows organisations to tailor the suggestions to their circumstances.
These sources inform the implementation and governance discussion. Neither endorses TEMRIK nor demonstrates that an AI playbook prevents insolvency. The 25-chapter summaries above come from the TEMRIK ebook; the operating examples and recommendations are editorial analysis.
Published 24 September 2026. General educational commentary. Contractual, financial and safety decisions require appropriately qualified people and current, case-specific information.