How AI Turns a Process Map and a Compendial Test Method into First-Draft Procedures

Writing a standard operating procedure from technical source material is often more demanding than it initially appears. A compendial test method may describe the analytical requirements precisely, while a process map may describe the sequence of activities performed within a laboratory or manufacturing environment. Turning those two forms of information into a coherent SOP requires more than copying text.

The information needs to be organized, contextualized and expressed as instructions that people can follow.

Artificial intelligence can assist with this transformation.

When provided with a clearly defined process map and authoritative technical source material such as a compendial test method, AI can generate a structured first draft of a procedure. This does not mean that the resulting document is automatically correct or suitable for approval. Rather, it changes the starting point for the author.

Instead of beginning with a blank page, the subject-matter expert begins with a draft that can be reviewed, corrected and developed.

This distinction is important in regulated documentation. AI can accelerate drafting, but responsibility for the accuracy and suitability of the final procedure remains with appropriately qualified people.

Why combine a process map with a compendial method?

A compendial test method and a process map provide different types of information.

A compendial method establishes technical requirements for performing a test. Depending on the method, this can include information concerning reagents, apparatus, preparation, analytical steps, calculations, system suitability, acceptance criteria or other specified conditions.

A process map provides information about how activities are organized within a particular workflow. It can show the sequence of activities, decision points, inputs, outputs and interactions between stages.

Neither source necessarily constitutes a complete SOP for the organisation using it.

A compendial method may define what the analytical method requires without describing every local procedural detail. A process map may show the operational sequence without containing all the technical information needed to perform the test.

Combining the two creates a more useful basis for drafting.

The process map provides structure. The technical source provides controlled content.

AI can then assist in translating those inputs into procedural language.

Start with the process, not the AI

The quality of an AI-generated draft depends heavily on the information supplied to it.

If the underlying process is poorly understood or the process map is incomplete, asking AI to generate an SOP does not solve the underlying problem. It may instead produce a polished document based on an incomplete representation of the process.

The process should therefore be established before drafting begins.

A process map should identify the relevant stages of the workflow and make important decisions visible. It should distinguish activities that are performed sequentially from those that depend on specific conditions.

For example, an analytical workflow might include sample receipt, identification, preparation, instrument setup, analysis, result evaluation and documentation. The actual structure will depend on the process, but mapping it first provides a framework against which the eventual SOP can be assessed.

AI should not be asked to determine the intended process simply because the process has not yet been mapped.

The technology is most useful when it is given a defined process to describe.

The compendial method provides the technical foundation

A compendial test method can contain highly specific technical requirements.

When such a document is used as an input for AI-assisted drafting, the objective should be to preserve those requirements while organizing them into a procedure suitable for the intended operational context.

The AI may be able to identify procedural components within the source material and organize them under headings such as materials, equipment, preparation, procedure, calculations and acceptance criteria.

However, the existence of a technically plausible section in the draft does not establish that the content is complete or correctly interpreted.

A reviewer must compare the draft against the authoritative source.

The question is not simply whether the AI has included the information. It is whether the information has been transferred without changing its meaning.

This distinction is particularly important when a technical requirement depends on a qualification, condition or defined parameter.

AI can transform structure into a first draft

Once the process and technical source material are organized, AI can perform a useful drafting function.

The system can be instructed to use the process map as the structural framework and the compendial method as the technical source. It can then produce a first draft that follows the sequence of the process while incorporating relevant technical requirements.

This is where AI can save substantial drafting effort.

Instead of manually converting a flow diagram and technical method into paragraphs and procedural steps, an author can review an initial structured document.

The draft may contain headings, numbered steps, conditional instructions and other elements that make the workflow easier to navigate.

The author can then concentrate on the substantive review.

This is a different activity from simply asking an AI system to "write an SOP." The quality of the input and the constraints placed around the drafting task are central to the result.

Plain language can be applied during drafting

AI-assisted drafting also creates an opportunity to apply plain-language principles systematically.

A process map can help establish logical sequence. The drafting instructions can then require the procedure to use clear headings, direct instructions and appropriately separated procedural steps.

Complex activities can be divided into manageable sections. Decision points can be made visible rather than embedded in long paragraphs. Related information can be grouped so that users do not need to search through unrelated material.

The objective is not to remove technical terminology.

A compendial method may require exact scientific terminology, and the resulting SOP must preserve that precision. Plain language should instead be used to remove unnecessary linguistic complexity around the technical requirements.

A technically precise procedure can also be readable.

AI can identify patterns that are difficult to see manually

Beyond initial drafting, AI can assist with some editorial analysis.

For example, it can review a document for inconsistent terminology, repeated information, unusually complex sentences or structural inconsistencies. It can compare the use of defined terms across different sections and identify areas that may warrant human attention.

These functions are useful because document review often involves repetitive tasks.

However, identifying a potential inconsistency is not the same as determining that an error exists.

If two terms appear different, the AI cannot necessarily determine whether they are genuinely inconsistent or whether they intentionally describe different concepts.

Human review remains necessary to make that determination.

The first draft is not the controlled procedure

The phrase "first draft" is critical.

A generated document should be treated as draft material until it has undergone the organisation's appropriate review and approval processes.

This is particularly important because AI systems can generate statements that are grammatically correct and scientifically plausible but unsupported by the source material.

The system may also omit information.

It may interpret a condition incorrectly, merge separate requirements or introduce language that appears reasonable but changes the intended instruction.

These limitations mean that AI-generated content should never be accepted simply because it reads well.

The reviewer must establish whether the document is accurate.

Human review needs to examine the source

A strong review compares the draft with its inputs.

The process map should be checked against the procedure to determine whether the sequence has been preserved. Decision points should remain visible. Activities should not have been added, removed or rearranged without justification.

The compendial method should be checked to ensure that relevant technical requirements have been retained and that the language of the draft does not alter their meaning.

The reviewer should also consider whether local procedural requirements have been represented correctly.

This review is fundamentally different from proofreading.

Proofreading asks whether the text is written correctly.

Technical review asks whether the procedure is correct.

Both can be necessary, but they address different questions.

AI cannot replace process knowledge

One of the greatest risks of AI-assisted documentation is assuming that the system understands the operational environment simply because it can process large quantities of text.

It does not have the same practical understanding as the people who perform and oversee the process.

A process map may not capture every operational consideration. A compendial method may not describe every local activity required to execute the test within a particular quality system.

Experienced personnel may know that an instruction requires additional context, that an apparent exception is significant or that a particular sequence has an important practical reason.

Those observations are part of the review process.

AI can reorganize available information. It cannot be relied upon to supply undocumented operational knowledge.

The most effective workflow is iterative

AI-assisted SOP development is likely to be most useful as an iterative process.

The initial draft provides a starting point. Reviewers identify missing information, incorrect interpretations, unclear instructions and structural problems. The document can then be revised and reviewed again.

This approach also allows the organisation to separate different types of review.

A technical expert can assess scientific accuracy. An operational expert can assess usability and alignment with actual practice. Quality personnel can assess the document against applicable quality-system requirements and established controls.

The result is not simply an AI-generated document.

It is a controlled document developed using AI as one tool within a broader authoring and review process.

Better drafting can create more time for review

The potential value of AI is therefore not limited to faster typing.

If AI can reduce the time required to create an initial structure and translate source material into readable procedural language, human effort can be redirected toward higher-value activities.

Those activities include checking the process, verifying technical requirements, identifying ambiguity, considering exceptions and assessing whether the SOP will work for its intended users.

This is particularly important because the review stage is where much of the critical judgement resides.

An efficient first draft is useful only if the time saved is used to improve the quality of the final document.

A practical model for AI-assisted SOP development

A controlled workflow can be summarized in several stages.

First, establish the process and create a reliable process map.

Second, identify the authoritative technical source material, such as the applicable compendial test method.

Third, determine what the SOP needs to communicate to its intended users.

Fourth, use AI to generate a structured first draft from those inputs.

Fifth, compare the draft against the process map and source material.

Sixth, conduct technical, operational and quality review as appropriate.

Finally, complete the organisation's established document approval and control processes.

This model keeps the technology in its appropriate role.

AI assists with transformation and drafting. People remain responsible for determining whether the resulting procedure is accurate, appropriate and controlled.

The opportunity is better drafting, not automated compliance

AI has considerable potential to change how SOPs are produced. Its greatest practical value may be the ability to transform structured information into a usable first draft quickly.

A process map gives the system a representation of the workflow. A compendial test method provides authoritative technical information. AI can combine those inputs into a draft procedure that provides a starting point for expert review.

But the final quality of the SOP still depends on the quality of the process definition, the reliability of the source material and the expertise applied during review.

The objective should therefore not be to remove people from SOP development.

It should be to remove unnecessary drafting effort so that people can spend more time on the activities that require scientific, operational and quality judgement.

That is the practical opportunity presented by AI-assisted SOP development: not a replacement for expertise, but a faster route from structured knowledge to a reviewable first draft.