A thinking professional's guide to using AI at work
Here are five principles for using AI without losing your craft.
Over the past year, I’ve had the chance to keenly explore how journalists can use AI tools in everyday tasks. I have talked at length with a whole spectrum of people across organisations—from spoilsport naysayers (and there are many in newsrooms; after all, being sceptical is fundamental to being a journalist!) to mindless proponents (whose excitement to stay ahead of the curve often gets the better of rational behaviour). Neither extreme is the right way to be. But without both of them, we couldn’t ever properly understand the nuances of how to, and how not to, approach AI at all.
Those who are mildly, or let’s say, cautiously excited about AI cannot by themselves become good users of AI tools. The naysayers push them to be responsible and ethical; the excitable ones help them think deeper about the possibilities that AI offers but they are missing out on.
Based on my learnings from both sides, I’m listing down five key principles that you could consider if you’re thinking about the right way to use AI. The crux of the idea will likely hold true not just for journalism, but for most sectors that involve any research or knowledge production.
First, even if your organisation lacks an AI policy, have one for yourself. Your personal AI playbook should protect you from all risks—from getting research or data wrong (which your employer will also care about) to the possibility of losing your original creative instinct over the long run (which your employer has no stakes in). This will help you decide when to use AI and to what extent, and when not to.
But on the flip side, your policy should actively liberate you to consider using AI to give wings to your work so that you can do and explore things that you were just not able to earlier. I don’t mean using AI just for the bare minimum (e.g., summarising long PDFs or doing preliminary research), but the need to constantly attempt more and more, even if it won’t show in your final output. Learning a tool may not always mean immediate adoption and use. In the case of AI (which involves building a relationship over time, like you build one with a human), it may simply entail prep for the future when the tool will (inevitably) get a million times smarter.
While you must use AI responsibly, not learning and using it actively is actually irresponsible. Your personal rulebook should help you step back and practise reasonable but not compulsive scepticism.
Start by thinking carefully about the raison d’être of your line of work. For a writer, originality of their craft is vital, so they won’t use AI to write. But these limits shouldn’t be based on personal egos or pride: “I won’t use AI to kick off my literature review” makes little sense because starting literature review, even if you’re proud of how good you are at it, isn’t the raison d’être of your work, is it? It’s just a means to an end.
(However, this must be done at an organisational level: a company may be okay with staff writing end-to-end using AI since it values its role as a communicator; another may frown since it sees its role linked to creativity. For some newsrooms, AI may work for formulaic storytelling for hard news, but never for longform, which needs a signature style.)
Second, your AI rulebook should ensure you’re not adopting AI just to protect your job or become more employable. Modern-day maxims like “AI will replace you if you don’t adapt” feed on insecurities, and aren’t a constructive reason to change. If you want to be excellent at what you do, new skills can expand your body of work, by helping you be either more efficient, or more impactful. If you don’t want to be excellent, you are forever at risk anyway, AI or no AI.
Also protect yourself from the other refrain: “AI will disrupt professions like x and y, but it’s just not good enough to ever replace <insert your profession>”. These maxims come from professional egos, not from a place of reason. If a tool hallucinates today, there’s no reason it will still do so five years later.
The goal of any good business and professional has always been to serve their consumers to the best of their capabilities using the most reliable tools and technologies of the times. This involves caution, but it also involves openness to test the limits of those tools and not be at the sidelines.
Third, in the case of journalists, the essential caveat to any use of any tool is always that at the end of it, you must be confident to take full ownership for what you publish in your name, irrespective of how much you used the tool. As simple as that.
All your checks and verifications will naturally follow (e.g. if you ask an AI tool for a CAGR calculation, add steps to either do it yourself manually, or ask the chatbot to show the working and the formula it used, or perform the same task on a second chatbot to verify).
The fact that you used AI cannot be a disclaimer; at best it can be a disclosure.
While using any external information or idea, your policy must force you to trace, and adequately credit, the primary source. That rule stands, AI or no AI. Secondary sources, such as Wikipedia or an AI-generated response, are necessary to give ideas or leads, but verifying against the primary source should be non-negotiable.
Like in the pre-AI world, your processes must put in deep care about how any information was produced at its point of origin: whether decoding the motivations of a human source in sharing leads with a journalist, or knowing how a survey was conducted or funded, or knowing how a tech tool interprets our query and the biases it operates with.
Make space for strong editorial judgement, aggressive cross-questioning and smell tests, reverse-checking techniques, and verification at all stages, to ensure accurate and unbiased facts, ideas, direction of thinking, and conclusions.
Fourth, while framing your personal AI policy, think retrospectively of the policy you subconsciously apply to pre-AI technologies, like, say, performing Google searches or converting images to text. Think of the kind of checks and balances you have in place to ensure you’re on the right track. With this in mind, think of a framework that works not just for the current AI challenge, but one that would have helped you navigate past tech breakthroughs and will likely stand the test of time even if there are new innovations at the workplace that we cannot imagine yet.
AI-agnostic thinking can help you find much-needed balance. It helps the compulsive naysayer realise that tech’s not all that bad and they have themselves grown dependent on other tools with time (would they still go to a library to scan hundreds of books now that they have Google search at their disposal?). An obsessive cheerleader realises that the smartest tools do need human oversight after all, as they have trained themselves over time (would they entirely trust an interview transcript generated by an app?)
Fifth, and last, I suggest you tie your level of AI use to your experience level. Junior professionals should avoid over-relying on tools so that they can build muscle memory, instinct and craft in doing tasks independently. Those skills will come in handy to develop instinct for future success and for leadership roles. Senior resources who have built those skills already must value themselves more and put their knowledge to greater use for the world by saving time from work that can easily be delegated to AI tools.
I hope these five principles are able to answer most doubts about good-quality, rigorous AI use. Do you have an AI-related principle that you follow for yourself that you’d like to share? Please leave a comment or hit reply to this email!