How we’re using AI at Which?

Q: How are you using AI in your business now?

A: At Which?, we’re using AI in a range of ways to make our work more efficient and effective. As a publisher, our ultimate goal is that this will allow us to produce more useful and relevant content and create a more personalised and engaging experience for our audience.

Across the business, we use Gemini for everyday tasks such as transcribing meeting notes and assisting with brainstorming and idea generation. We’ve also found it useful for summarising large amounts of information in the initial research stages of business development.

Different teams are adopting a range of AI tools that are more specific to their work. Our developers have been experimenting with Copilot for coding. Our editorial team uses a variety of task-specific tools to support workflows, for example:

• Our art editors work with image generators to develop initial concepts for artwork, such as magazine covers. This helps to streamline the briefing process before a designer creates the final designs.
• Our picture desk uses AI plug-ins to remove the background from product shots and draft alt-text descriptions.
• Our researchers use machine translation to quickly access foreign technical reports, enabling us to publish reviews faster.

We’ve also developed bespoke AI applications to support more specialised work across a range of teams. A company-wide hack day was a particularly effective way to identify and rapidly prototype applications, some of which we’ve now taken into production. For instance, we’ve built a tool to assist our researchers in analysing large sets of product testing data, giving them a headstart when preparing their reviews. Saving time by automating this part of the process gives our reviewers more time to spend on honing their expert insights and creating a more polished and helpful review.

AI is proving to be particularly powerful in enabling us to realise more value from the data we have access to. Using AWS Bedrock, we’ve built a sentiment analysis tool that extracts and summarises insights from the comments users leave on our website. This feedback is now much more accessible to the range of teams that can use it to make improvements — from editorial to product development.

Our consumer policy and advocacy team has also used AI to analyse large volumes of unstructured data, such as ombudsman rulings, to uncover hidden harm in markets where we are campaigning to improve consumer outcomes.

As a business, we’re embracing the potential for AI to have a positive impact on the way we work and for this to translate into improved experiences for our customers. However, we aren’t reliant on it and are careful to ensure that all AI outputs are thoroughly checked and that the content we publish is created by a human. For Which?, AI is a tool to enhance efficiency, creativity and insight, not a replacement for the vast knowledge and expertise of our teams.

Q: How do you expect this to change over the next 12 months?

A: Over the next year, we are keen to expand our knowledge and use of AI as a business. We will be encouraging wider adoption of Gemini across the organisation and providing targeted training to ensure more teams feel confident using AI as part of their day-to-day work.

We’ll continue to explore more bespoke uses of AI, including how it could improve our front-end customer experience, for instance via a natural language search interface. As we all get more used to the sort of conversational interaction that gen AI chatbots offer, we need to experiment with evolving our experience in line with shifting consumer expectations.

Ongoing experimentation will be key — both in the range of models and tools that we use and their application. We will encourage employees to safely experiment with AI, to test and learn how it can enhance and support their roles and to share findings with colleagues across the business. We hope this exploratory approach will help us to continue to find the most valuable uses of AI that help us deliver better outcomes for consumers.

Q: What are your three top tips?

1. Set clear guardrails. There are risks associated with using AI platforms, such as sharing proprietary data or generating inaccurate outputs. Establish robust guidelines that are regularly reviewed and refreshed and always ensure there is human oversight of all AI outputs.
2. Encourage experimentation. Not every AI trial will succeed, and that’s okay. Some of our early experiments such as using it to inform SEO didn’t work out, but as the technology evolves, we revisit them to see if improvements have been made. Staying open to testing and learning helps identify the most valuable applications.
3. Share insights. Encourage teams to explore AI uses and share useful applications and learnings with colleagues. Look both internally and externally for news on the latest developments and uses of AI to stay up to date with best practice and emerging opportunities.