Re-thinking Discobot and AI-assisted member support in Crossref’s strategy

[boutaleb]

The role and benefits of Discobot raise several questions within the Crossref community. While it plays a useful role in basic orientation, its potential as a tool for learning, guidance and ongoing member support is not always clear, and may be limited in relation to the diversity and scale of current needs.

At the same time, a large number of questions raised on the forum remain relatively basic — for example how to join Crossref, how to register a DOI, or how to correct a simple metadata field. These questions are entirely legitimate, and it is excellent that they receive thoughtful and supportive responses, often directly from Crossref leaders.

This situation highlights an opportunity to reflect collectively on how member support is structured. Part of this foundational assistance could potentially be handled through automated and AI-assisted solutions, allowing Crossref staff and community leaders to focus their expertise on new, complex and genuinely strategic issues, where human judgement and experience add the greatest value.

In this context, automation and artificial intelligence appear to be important elements to consider more explicitly within Crossref’s strategic framework. They could enable more personalised and context-aware support, adapted to the diversity of the global community, while also strengthening metadata quality and operational efficiency.

I would be very interested to hear perspectives from the Crossref team and the wider community on this.
How does Crossref envision the future role of Discobot , and more broadly the integration of automation and AI , in order to enhance member support, improve metadata quality at scale, and better align global strategic priorities with local and community-driven participation?

[ifarley]

Hi @boutaleb ,

Thanks for your message. We’re considering Discobot and other potential solutions within the Community Forum, so I do appreciate you flagging it to us here.

I take your point that many repetitive and straightforward questions could be addressed with such technology and generally agree with that, so it is worth investigating, which we’re doing.

But, at least a couple of the examples you have surfaced - how to register a DOI and how to correct a simple metadata field - are not terribly straightforward, since they differ greatly depending on the content registration method/preferred tool (e.g., OJS (which version is being used?), web deposit form, direct full XML upload, direct resource-only XML upload, bulk upload of XML via HTTPs POST, platform provider, the new metadata manager tool, DOAJ, etc.).

Sure, an AI could potentially help, but we have to be sure that whatever we adopt is simplifying the process for our community.

What do others reading this post think?

-Isaac

[boutaleb]

Thank you, @Isaac, for this thoughtful and particularly nuanced response.

I would like to share a brief personal experience, as I believe it reflects the situation faced by many young or emerging journals. We are a relatively new journal and, within our national context, the number of Crossref members remains quite limited. As a result, local familiarity with Crossref tools, workflows and requirements is still low.

In this context, I had to learn how to work with Crossref largely on my own: understanding membership procedures, the different deposit methods, article registration and DOI generation. This learning process almost always begins with errors and failed attempts, which is entirely normal when engaging with such a structured and demanding system. To make progress, I initially relied heavily on support from the Crossref team and, increasingly over time, on AI tools to help me understand error messages, XML schemas, differences between deposit methods, and the overall logic of the system. By building on the accumulated history of questions and solutions, these AI tools become progressively more robust and effective.

I would like to emphasise the quality of the human support provided by Crossref . The responses I received were consistently generous, pedagogical and respectful, even for questions that may appear basic but are far from simple in practice. As you rightly note, tasks such as registering a DOI or correcting a metadata field can vary greatly depending on the tool used, the software version, or the type of deposit.

It is precisely this complexity that makes structured guidance essential. Over time, the accumulation of questions, errors and resolutions — even at the scale of a small journal — becomes a valuable knowledge base.

If we extend this reasoning to the Crossref ecosystem as a whole, the potential is considerable. The aim would not be to introduce a general-purpose AI, but rather a specialised assistant grounded in Crossref documentation, XML schemas, existing workflows and the history of support requests. A Discobot-type tool , enhanced and more finely targeted and enriched, could offer step-by-step guided pathways adapted to each member’s real context (deposit method, platform used, level of experience).

Such an approach would in no way replace human expertise. On the contrary, it could help reduce the support burden on Crossref teams by handling recurring entry-level questions, while freeing up time for complex, cross-cutting or strategic cases where human intervention remains essential. In this sense, automation and artificial intelligence should be seen not as a threat, but as an efficiency lever, a complement, and potentially a significant asset for the future of the Crossref ecosystem.

On a lighter note, I should admit that my first real encounter with took place after I had been appointed as a Crossref Ambassador. Full of enthusiasm, I initially thought I might be able to answer almost any question — and even tried to make it talk, or perhaps reveal a few secrets! This small but revealing experience is, in fact, at the origin of the reflection I am sharing here.

To conclude on a more personal note, being appointed Crossref Ambassador has been a real honour for me and represents an important commitment. I would like to warmly thank the entire Crossref team, and especially @Obanda, who is always attentive, generous with his time, and consistently responds with kindness to my requests, particularly when it comes to organising meetings