AI in practice: Faversham House

Q: What is the current state of play with regards to AI usage?

A: We encourage staff to use AI securely and responsibly, to maximise their potential by concentrating on tasks only they can do. We provide a corporate Claude subscription, with managed connections into multiple company systems, and provide support, training and workshops to encourage experimentation and knowledge sharing. We also invest in other AI products when they genuinely do something better than we could otherwise achieve.

We maintain a living document where team leaders are asked to record their team’s approach to using AI, their current use cases and future plans.

While there are obvious efficiency gains, we see the opportunity not just to do things better but also to do better things. Our primary focus is therefore on creating additional value by improving the quality, relevance and reach of our products and communications, and creating engaging and memorable experiences for specific people, both in person and online.

We recognise the potential for massive disruption of our industry and are adapting our approach to ensure that our main outputs cannot easily be replicated by AI or other competitors. This means creating unique content and experiences, based on a deep understanding of the people we are targeting, with real people at the centre of everything we do.

Some of our usage is well organised and centrally defined, but a lot is still exploratory and piecemeal. There is still a huge amount of untapped potential, and a lot of work to do in making our usage more efficient and systematic.

Q: What are the best opportunities for future development?

A: In my view, the best opportunities lie in these areas: developing deeper understanding of audience members and customers, building products that deliver greater value to them, and enabling our people to do better quality work. These three things are obviously related – and certainly should be!

• Product: On the product side, the opportunity that is imperative for us to address is the changing user interface for content consumers. For many years, that interface has been the browser and through it, we’ve understood how to communicate, control the user’s experience and extract information to learn more about them. We’d love to keep it that way, but it’s essential that we learn how to interact with users who now want to consume our content via agents and LLMs. Otherwise we risk losing our relationship with those people as they either go elsewhere or consume content without our control or knowledge of who they are and what they want. Things we’re looking at here include more nuanced crawler controls, serving agent-specific content and managed MCP access for corporate subscribers.

• Understanding: Looking more internally, we are finding that connecting Claude into our systems and data sources is enabling people to perform analyses that previously would have been a project for our data team. This means people can find insights more easily and make better decisions as a result. For example, we now ask new registrants to tell us what they are interested in, in their own words. We can analyse all the new comments in Claude and update the editorial team with a briefing on what new sign-ups want. We can also combine that with our content to generate a welcome briefing on recent content they would like. There is definitely an opportunity to make this more systematic and put more evidence-based insights into our decision making.

• Process enablement: We all know the value of having good processes, but how often have you banged your head on the desk in despair at them not being followed? We know we should follow the process but when you are super busy and a bit stressed, the option to conveniently forget some steps and take a few short cuts is all too tempting.

Enforcing processes on others is perilous as it tends to produce resistance or reluctant minimal compliance. Process enablement, on the other hand, makes it easier for people to follow the process than not follow it, greatly improving the chances of success. We now have much easier ways of doing this, as an AI tool can be trained on a process and provide assistance and guidance at the same time.

An example of this: we have a new sales process we want people to adopt. It involves some additional steps and thought processes, and it’s probably fair to say our existing sales admin is less than perfect! However, our new Claude skill, trained on the process, will take a call transcript and produce call notes that are compatible with the process, upload them to the Opportunity record and update contact roles, provide coaching tips, and create tasks on our CRM for the recommended and agreed next steps.

There are many opportunities for this kind of tool to make life easier for the user at the same time as ensuring agreed processes happen.

Q: What are your three key learnings?

1. Increasing quality is better than increasing quantity or speed. AI is great at producing things that “look good” but don’t bear close inspection.
2. AI is a lot more powerful and useful when you connect it up to your systems and data – but it’s also a lot more risky.
3. AI can save a lot of work, but it creates more work. Sorry!