The authority problem that copilot (and others) will reveal with our records.

I was talking to another consultant recently, and he was describing a situation in which he'd traffic lit the reports for an organisation (ie, red, yellow, green) based on the quality of the lineage available for the data supporting the reports (lineage and provenance as concepts are roughly equivalent).

It can be a controversial move - in one case, he had red lit every report the organisation used because there was no lineage available - the reports really weren't worth anything because they didn't understand where the data was coming from, and how it was defined.

The challenge with copilot and other tools like it, is that they're going to answer questions based on the corpus of documents available to the user.

There are obviously going to be challenges around security - because the swiss cheese of most document repository security means that people will be able to ask questions like "how much money does x make" and "who has been put up for disciplinary action" - and get answers.

The bigger challenge though, is going to be authority.

Just like lineage, deciding what the authoritative document is always requires a process solution.

It's people, agreeing to a technical process where they agree that they're going to name a document xyz when its the authoritative one, or they're going to have a publish function - and actually use it, or they're going to have soem other process that everyone in the organisation knows means "this is the authoritative document, the one that I should base my decisions on."

The problem we have, is that for most of us, most of the time, we'd have to mark most of our repositories red when it comes to how authoritative the documents in them are, becuase they're a dumping ground, and only people with deep process knowledge (about the business process that was taking place), or who know the organisational structure (so they know what an edit from someone means) will be able to make sense of it.

I put a post up a few weeks ago with idea I was toying with about how email servers have the best provenance.

The main surviving idea from that train of thought, is that if someone put a fact or a directive into an email, they meant for the recipient to use that fact, or act on that directive - which naturally carries a level of authority with it. It's incomplete, and may skip a lot of what was going on in the real world - but it was expected to be acted on, becuase it was communicated in writing, and people don't communicate things in writing unless they expect them to be used.

Most of what we put in our document repositories doesn't have this characteristic. Half of it is drafts that didn't make it - or that changed substantially between the draft and the finished product.

The level of authority in the average EDRMS is staggeringly low - at least for documents that the organisation was creating. Which might put us in the slighly ironic position of being able to ask good questions about things from other organisations, but not our own.

All of this is before we even get into badly organised "repositories" like file servers and personal storage - services that generally have no meaningful provenance.

There's so much that is going to be interesting in the coming years about how these services work, this is just one more challenge that's going to bring core records and document expertise back as something that organisations begin to seek out.

For those of us practicing records as a capture focused discipline, it might be worth dusting off our document control hats and starting to broach the subject with the people who have the power to mandate organisation wide process.

Internal forms of generative AI are very likely to be a huge productivity improver, but the truth is that most organisations aren't going to be able to switch them on because they'll create a security problem and others will switch them on and get gibberish (because the corpus will have such a low level of authority it will appear to be "making up" the results).

There will be a select few that can turn it on, and they'll reap the benefits.

They'll be the ones who solve the authority problem - at scale.