Finding the AI uses worth trying and getting the tools vetted
A writer needs to know which uses of AI tools are worth their time, and can look for them in what others publish, find them by using the tools, or rely on someone else, such as a professional association or a company library, to judge the tools on the writer's behalf.
Use cases and why few are published
Maturing the use of AI tools needs clear use cases, in the view one regulatory writer gave in December 2023. On his account, the tools hold possibilities but leave a natural hesitation over how to use them effectively and in what cases, and use cases that set out, for a given activity, which tools to use, how to use them and how to benefit from them without running into their challenges, such as security concerns, would let writers take hold of what AI can do.
An engineer who had spent much time exploring the tools held in December 2023 that doing so is time-consuming and keeps a writer from their real work, and that a regular person with a real life cannot possibly keep up with the tools and try them all out. She suggested that few use cases are published because people in the know would lose their competitive advantage by sharing them.
Where writers find use cases
She pointed writers to the tools' own promotional activities, to YouTube, where many young scientists and researchers make good tutorials of around five minutes explaining real-world cases, to Twitter, where following the right people yields a lot of information, and to Reddit. On YouTube, a writer can pick a tool they like and then search for it. The regulatory writer replied that this route relies on each writer's own interpretation of each tool and of how to use it.
A whole world of tools exists to help researchers and medical writers do their work more effectively, and many scientists and entrepreneurs with a science background are working on specific tools, in the view of a speaker from the author-services industry in May 2024. On his account, many share a frustration that ChatGPT and similar tools feel as though they are not meant for scientists. He advised that a writer who tried ChatGPT once and was not happy with it, or who worries about privacy, should explore this world of tools for science publishing before giving up on AI, and start from the specific use case. The use cases he named include writing, such as a tool for author manuscripts, and clinical trials.
Finding uses by using the tools
One publisher's senior editor, no technology expert, was overwhelmed when generative AI arrived, so she challenged herself to start using it, and came to a better appreciation of what it can do without becoming any more of an expert. In 2024 she held that the best way for her to learn about it was to start using it and talking about it, and to be her editorial team's go-to person on AI. Her challenge in everyday use was finding what she could use it for, because as an editor she cannot feed content into it.
She found a use while writing an email to a team of editors to explain a new marketing platform she did not understand well. Five minutes in, she turned to ChatGPT, which in five minutes produced the whole email, a simple step-by-step explanation that she then changed a little. She saw then what people meant when they said it helps them write communications and emails.
She tells her team to try to use it once a week, because the more they use it, the more they see how it can be applied. She also tells them not to be afraid to use and explore it as long as they follow their company's ethics, since nothing wrong will happen and they will always see the output and can check it. On her account, using it regularly is how a person discovers how to incorporate it into their job. The gap still to be closed, she held, is how someone whose job does not call for it every day comes to understand how they can use it.
Endorsement by a professional association
In December 2023 the regulatory writer asked whether a body such as EMWA could endorse a tool as good for a particular purpose, without certifying tools as such, so that writers without the technical capability to understand the intricacies of each tool could use it in their own work on that body's word. In December 2023 EMWA's AI working group planned some categorisation and list of the AI tools available to medical writers and the uses they would have for them, without full due diligence on each tool and without recommending particular tools. Recommending tools would be quite dangerous, the engineer warned, because none of those involved is affiliated with any of the tools, and a writer whose recommended tool then made a mistake, as such systems do, could say that EMWA had recommended it.
The company library as vetter
A large pharmaceutical company probably has a library or information centre, and a writer working for one should go and look for it, in the advice of one such company's librarian in May 2024. By then her library team at Takeda, seven people among them information specialists, had been following the ethics of AI, AI in science and the guidelines journals were issuing on what they would accept in articles written with AI. The team decided to become the company's subject-matter experts on everything to do with AI. It does not build AI tools, but answers the questions of those who do, such as where to find articles to help build a large language model.
The librarian who leads the team held that librarians are uniquely positioned to address AI gaps in medical writing, as subject-matter experts in information verification, dissemination, curation and scientific search training and teaching, and above all in vetting, testing, licensing and paying for new tools. The library's goal is to promote AI literacy, helping medical writers to critically evaluate AI technologies, collaborate effectively with AI tools, understand the basics of how AI works, use AI effectively and ethically and make informed decisions about how and when to use it. It also promotes appropriate use cases, since not every use case is a good one for an AI tool. She held that it is too hard for a writer to go it alone, as there are too many tools.
How the library vets a tool
On her account, most of the company's medical writers who have attended even one conference receive all kinds of emails from new companies, each touting a new tool as the best there is, and have to decide whether to use it. Writers and writing teams direct the vendors to the library, which vets each tool by experimenting with its capabilities. It puts a set of questions to the vendor and asks the vendor to go through a set of exercises. It includes teams by department in the evaluation, such as regulatory writers and manuscript writers, because their needs for a tool are not always the same, and invites them to put in their own use cases.
The library recommends signing no contract longer than six months with a new AI vendor, because the technology will change quickly and the company wants to be comfortable before spending a lot of money. It may instead ask for a six-month paid trial for interested users rather than the whole company, which also seems to work well. By May 2024 it had recommended two tools that were working out pretty well.
How the library shares what it learns
In May 2024 the library provided a dashboard listing every journal's guidelines on AI, for a writer who wants to know how much AI a journal allows and how it has to be cited. It showcases AI tools and their utility across the company through newsletters, weekly trainings, half-day sprints and lunch-and-learns. It aims at medical writers in particular, because they interact with the literature most often and because the head of the library had been one. It wants them to bring it what they learn from vendors at conferences, so that it can compare and curate that information. It publishes a monthly AI newsletter covering trends in AI and AI in the company's therapeutic areas, and takes part in a company-wide ethics and AI council.