01
Compliance
Technical standards, safety requirements and final suitability need accountable review.
A practical point of view
AI is most useful when it gives judgement more room to work.
I am not presenting AI as a finished project I have delivered. I see it as a practical capability for reducing repeatable work in procurement, estimation and commercial analysis — while keeping technical responsibility with the engineer.
The operating idea
Use AI for attention-heavy work. Keep the final call human.
The best starting point is not “where can AI do everything?” It is “which part of this workflow is repeated, structured and easy to verify?”
Good candidateExtracting line items, quantities and supplier descriptions from long documents.
Good candidateBuilding a first-pass comparison table and highlighting unmatched scope.
Good candidateTurning meeting notes into decisions, owners and follow-up actions.
Good candidatePreparing negotiation scenarios from a verified commercial benchmark.
Four practical loops
Productivity is a workflow, not a prompt.
AI becomes useful when the input, output and verification step are clear. These are the loops I would build around real MEP commercial work.
Extract before you analyse.
Start with the quotation, BOQ or specification. Ask for a structured table with source references, not a confident paragraph.
InputPDF quotation, schedule or email thread.
OutputItem, size, quantity, unit, rate, amount and source page.
Compare on a common basis.
Different names and units can make equivalent offers appear different. AI can help prepare the first mapping, but dimensions, product geometry and technical standards still need review.
CheckPipe length, UOM, quantity factors and fitting dimensions.
FlagAngle variance, missing rates, supplier-only items and exclusions.
Turn meetings into working memory.
Meeting notes should not disappear into a folder. A practical AI pass can produce a decision log that shows what changed, who owns the next action and what remains open.
CaptureDecision, evidence, owner and due date.
ReuseBring the last decision back before the next meeting.
Prepare the conversation, not a script.
Use AI to test scenarios: what changes if quantity, lead time, payment terms or delivery certainty changes? The point is to enter the discussion with options, not to outsource the relationship.
ModelOpening position, target, minimum and tradeable terms.
VerifyConfirm every assumption against the actual offer.
The human line
What I would never delegate.
AI can support the work. It should not hide the source, make a technical commitment or take responsibility for the outcome.
02
Commitment
Do not let generated wording become an unintended commercial or contractual promise.
03
Context
A supplier relationship and project constraint cannot be reduced to a single score.
04
Ownership
The person approving the recommendation remains responsible for the decision.
Interested in practical AI?