AI may assist
Classify documents; extract candidate quantities and references; detect apparent gaps; draft communications; retrieve similar jobs; summarise evidence.
From automatic quotation generation to the right first AI intervention
Westmoor Precision Engineering Ltd is fictional. This representative example demonstrates the Codance method; it is not presented as commissioned client work.
The initial ambition
Westmoor receives approximately 25–40 requests for quotation each week. Enquiries arrive through a shared inbox, direct email, a website and occasional telephone calls. A typical request may include drawings, CAD files, quantity spreadsheets, material specifications and incomplete or contradictory instructions.
A local automation supplier proposed a discovery phase and prototype costing approximately £12,000. The proposed system would monitor email, extract information and create quotation drafts automatically.
The proposal sounded efficient, but important questions remained unanswered:
- How would drawing information and revisions be validated?
- How would conflicting quantities or specifications be handled?
- How would historic jobs be selected and compared?
- How would JobFlow MRP interact with the enquiry record?
- Which engineering and commercial decisions required approval?
- What would happen when the AI was uncertain, wrong or unavailable?
What the assessment found
The main operational constraint was not simply the time required to calculate a quotation. It was the absence of a consistent intake and control process before estimating began.
| Present condition | Operational consequence |
|---|---|
| Enquiries enter through several routes | Some may remain unread, unassigned or outside the shared process. |
| Outlook, Excel, a shared drive and JobFlow hold different parts of the truth | Nobody can reliably see ownership, status, missing information or the latest clarification. |
| Drawing completeness is checked largely from experience | Important notes and conflicts may be found late. |
| Historic estimates and actual costs are inconsistent | Automated pricing could reproduce weak assumptions with false confidence. |
| Two experienced people hold essential knowledge | Administration consumes scarce technical attention and creates key-person dependence. |
Readiness conclusion
Ready with prerequisites for a limited, human-supervised intake integration. Not ready for autonomous quotation generation.
That conclusion is commercially useful in its own right. It gives Westmoor a reasoned basis to reshape or pause the proposed prototype before spending substantially more.
The correct first intervention
Codance would recommend AI-assisted enquiry intake and estimator preparation. The bounded first stage would:
- capture every authorised enquiry and preserve the original evidence;
- classify documents and extract candidate facts, visibly marked as unverified;
- apply deterministic completeness checks and identify apparent conflicts;
- draft an acknowledgement or clarification request for human approval;
- retrieve potentially similar historic jobs with supporting evidence;
- assemble an estimator review pack and assign it to a named human owner;
- record clarification, approval, status and completion events.
What the £95 Assessment delivered
- a decision not to commission automatic quotation generation in its current form;
- a bounded intake and estimator-preparation opportunity;
- clear source-of-truth, process and access prerequisites;
- a shadow-mode evaluation followed by a human-approved pilot;
- preserved engineering and commercial authority;
- clearer questions and acceptance conditions for the supplier.
Westmoor could act on this decision without buying anything further from Codance.
What the £295 Blueprint would add
| Assessment | Blueprint |
|---|---|
| Determines whether, where and why to proceed. | Defines how the selected integration should operate. |
| Identifies readiness, risks, prerequisites and the best first opportunity. | Defines future workflow, information, responsibilities, approval, systems, failure, stages and acceptance. |
| Ends with an independent next-action decision. | Creates a basis for internal approval, supplier quotation and controlled prototype planning. |
Illustrative pilot acceptance conditions would include:
- no loss or alteration of original customer evidence;
- tested classification and extraction on a representative set;
- mandatory escalation of required-field conflicts;
- no unauthorised customer communication or commercial commitment;
- acceptable human correction burden;
- complete audit history for every pilot enquiry;
- safe manual continuation when the AI service is unavailable.
The result
A smaller, safer first integration; clearer supplier expectations; protected human authority; and a path that creates useful operational data before attempting higher-consequence automation.
Continue reading
The complete fictional Westmoor AI Workflow Readiness Assessment shows the scope, evidence, current workflow, information analysis, decisions, readiness ratings, prerequisites, risks and recommended route a client would actually receive.