Notes from the work itself.

A website has a job to do. The right buyer should be able to understand the offer, judge the evidence and take the next step without needing somebody to explain the page to them. The team receiving the enquiry should know enough to respond properly.

That makes the website part of the operation, not a layer of decoration around it. The proposition, approved facts, booking route, consent, performance and release controls all have to work together. A polished interface with nowhere dependable for the enquiry to go has still failed.

I start by asking which decision the website must make easier. Then we agree which facts are approved, what proof can be published and where each kind of visitor should go next. Design and development become one commercial job instead of a queue of disconnected page requests.

A dashboard cannot resolve two systems that disagree. It can only make the disagreement look more finished.

Before choosing a chart, agree the question and which source should be trusted. Decide what happens to missing, duplicated or conflicting records. Keep the exception count visible. Write down how the number is calculated.

The test is simple: can the person responsible explain the number in the meeting? Good reporting shortens that decision. It does not create another reconciliation exercise before the discussion can begin.

A workflow demonstration is easy when every field is complete and every system responds. Real work starts when one of those assumptions fails.

Start with the blocked job. Use the missing asset reference, the photograph nobody can interpret or the approval nobody has authority to make. Decide who needs to act and keep the evidence with the blockage.

I would rather put a smaller workflow into use that handles those cases than build a larger one that only behaves in a demonstration.

The field team can do its job well. The office can do its job well. The operation can still fail in the gap between them.

That gap is where the photograph loses its context, the status changes in one system but not another, or an exception gets solved in chat and never returns to the job record.

Treat the handoff as part of the workflow. Decide what must cross it, who owns the next decision and what happens when the normal route breaks. Automation becomes useful after those answers exist.

AI earns its place when it helps somebody deal with evidence faster or notice something they might otherwise miss. Triage, summarisation, classification and pattern finding can all be useful. None of them should hide an important decision or make accountability disappear.

The pattern I trust is simple. The model makes a suggestion and records why. A named person accepts or changes it. The system keeps the original evidence alongside the final action.

That is less dramatic than trying to automate everything. It is also much easier to trust when the work matters.