AI can simplify, clean up, and improve your ISO 9001 documentation. But it falls short in real-world processes and communication.

At Yonder, we are in the unsexy, no-nonsense, but relevant business of electronic documentation.

Awwww, electronic documentation. How exciting.

Well, ever thought of digitizing the cockpit of an aircraft? Or about bringing the relevant information to firefighters on the frontline to get their mission done? Or about taking out some of the pains of ISO 9001 / 27001 for small companies?

We are involved in all of these fields, and we think that electronic documentation is actually quite exciting.

Let’s focus on the ISO 9001 / 27001 documentation in this article. And let’s not focus on what ChatGPT could do in our product, but on the pros and cons of using Artificial Intelligence in general in the context of ISO 9001 / 27001 documentation.

The No-Brainers

1. Simplify documentation for the end user, not the regulator

The raison d’être of a norm is that there is a regulatory body that enacts the norm, and all certified organizations have to comply with it.

The advantage is that there is no discussion of what norm compliance means, as it is the same for everybody. The disadvantage is that more often than not, the language used in norms is highly technical and administrative, and users might have the feeling that the documents weren’t written for them, but for the regulator.

Today’s generative AI tools are good at summarizing, reformulating, and changing the style of existing writing. Here is what generative AI could do to simplify your ISO 9001 / 27001 documentation:

  • “Find and resolve all nested sentences in my documents, without changing the meaning of the content.”
  • “Explain all instances of shall/should/must in plain text explanations, and insert them as footnotes at the end of the paragraph.”

2. Find and eliminate duplicates

Most of our customers have large, complex documentation landscapes that have organically grown. Normally, people lose the big picture of the entire documentation landscape, which leads to duplicates in the content. That becomes challenging when one copy of that duplicate content needs updating, but isn’t detected in all the places it is used. Here are some things that generative AI could do for you:

  • “Find all duplicate passages in my documents, and tag them with a specific label.”
  • “Suggest which passages in my documents I could reuse, and suggest which document to use as the master document for my reused content.”

3. Find and eliminate inconsistencies

We all know it. We don’t mean evil, but inconsistencies in the documentation are a common drawback of human work. We get tired, and we use the full term in one location, but an abbreviation without an explanation in another location. Some authors write numbers from one to twelve as words, others as figures.

Generative AI can detect all of that for you:

  • “Create a list of abbreviations, use the abbreviations in all texts, and place a reference to the list of abbreviations.”
  • “Make sure all numbers from one to twelve are written as words, and make all changes autonomously.”

4. Find and notify non-conformities

If you link your ISO 9001 / 27001 documentation to the underlying norm paragraphs and controls, tracking norm changes automatically doesn’t need AI.

However, checking whether each norm paragraph and control is properly described in your ISO 9001 / 27001 documentation, can be simplified using generative AI tools:

  • “How have other companies of our size and in our industry described their risk management process?”
  • “Is one of my process documents missing the link to our risk management process?”

The Challenges

Now for the hard part. AI is not a solution to all the problems, or at least not yet. Today’s generative AI tools are bad at creating new knowledge because they rely on training through vast amounts of text data that are publicly available.

Well, I guess your internal processes are not public knowledge. And I hope there is more than just the process descriptions in your QMS and ISMS.

Let’s look at some of those challenges.

1. Processes vs. team structure

Legend has it that when your company grows, your team needs more resources. Whilst that is certainly true in functions like customer success or customer support, let me share a secret with you:

Sometimes, it helps to have fewer people on your team.

At 30–50 people, you don’t need a full-time QM or CISO. You don’t need six people on your executive team. And you certainly don’t need a finance guy who prepares tons of Excel sheets, the CEO can do finance himself.

Parkinson’s Law at its finest. You need fewer chiefs, but more Indians. When things go badly, don’t hire an additional manager. It’s better to fire a manager, not replace him, and organize the work more efficiently.

Tell me how today’s AI tools can adapt your processes to the true needs of your organization, and how those tools will be able to spot inefficiencies due to the different actions and behaviors of your team members.

2. Processes vs. communication

Writing down processes is easy. But they only ever are effective when an organization’s leaders can communicate them to their teams.

Whilst productivity guys swear by working asynchronously, which often means writing tons of documents and emails, I’m afraid there is no way around talking to each other.

That’s why you need leaders. One of their primary tasks is to make sure people talk to each other rather than write to each other. What sounds simple is a never-ending marathon. And I am sure no AI tool of our times can take over this difficult and tedious task (yet) from human leaders.

3. Process vs. real world

Processes shouldn’t be carved in stone but adapted if and when the organization evolves. Processes also need to make sense for your specific activities, and it’s often when an organization evolves that processes don’t make sense anymore.

AI tools and humans share the problem that they struggle to make sense of the real world. How many silly processes and stupid rules exist in our world?

And now on to the final question: Are those humans who came up with those silly processes and stupid rules able to create AI tools that will eventually eliminate all that red tape and generate not just process documentation, but effective and efficient processes?