AI Best Practices: Lessons from DSC's AI Conference

After speaking at the AI conference that we hosted last week, one thing was clear: a lot of people are keen to explore AI, but unsure how to do it well. So, here is a simple guide to help you use AI tools in a way that is smart, ethical and in line with your organisation’s values.

1. Is this content public or private?

Before you share anything with ChatGPT, Gemini or any other tool, ask yourself: is this public information? If it’s internal or involves people’s data, it is best not to share it.

Always double-check your organisation’s Data Protection Policy. If the information is confidential (and if the information complies with GDPR). And if you have doubts about it, do not share it with AI tools. (And if you think your organisation needs to update your policies, here you can buy our Policies and Procedures Templates).

2. House style

AI can write copy quickly, but it won’t know your tone of voice and style unless you give it proper instructions. If your organisation has brand guidelines and/or a house style, make sure you stick to them. If you don’t have one, this is your sign to create one!

If you don’t get these things right and end up copy/pasting whatever your AI tool gives you, this could potentially impact how your beneficiaries perceive your work, your organisation and your credibility (personal and organisational).

3. Keep an eye out for bias

Generative AI is trained on vast amounts of online content and unfortunately, the internet is not always the kindest place. That means that when you ask AI to create something (text, video, audio or image), there is a potential risk of obtaining biased results. A significant risk exists that stereotypes may be reinforced, or certain demographics excluded, through the utilisation of biased language or imagery.

4. Fact-checking

AI can be helpful for research, but it’s not a 100% reliable source. For instance, Gemini, has a disclaimer stating that it can make mistakes, including inaccuracies about people. So, always double-check stats, quotes and names; and cross-reference anything important before using it in reports, blogs or emails.

5. Only use it when you need to

It is easy to get carried away and try the latest trends and generate a picture of yourself in Studio Ghibli style or have a long discussion with your AI tool about if 100 men could defeat 1 silverback gorilla, but remember that AI uses a lot of computing power, which means a bigger carbon footprint.

Before generating anything, ask yourself:

• Do I really need this?
• Is there a quicker, more sustainable way to do it?

6. Pick one tool and master it

Before you try out all the new tools that are emerging, it would be important to focus on one, test it thoroughly, and learn as much as possible about it before considering using another (also it is recommended to do some analysis before getting one tool, see some comparisons between tools and video tutorials if you are not sure which one to pick first).

Also, it is recommended that you talk with your colleagues and that everyone uses the same tools. This helps create familiarity with the tool, and if someone is on leave (or leaves permanently), the rest of the staff can continue with the work without the need for extensive handovers.

7. Be curious, but be cautious

AI, like any other tool, is only as good as the person using it. So, take your time, stay curious, lead with values and always ask yourself these questions before starting to use AI:

• Is this information private or public?
• Does it comply with our house style and guidelines?
• Is the information that the AI is giving me reliable?
• And something that we also ask ourselves at DSC, based on our values (which you can see here), is: Is this (document, image, video) excellent and empathetic?

If you want to get to grips with AI, prompt engineering and Agents in more detail, check out our upcoming bitesize session on Tuesday 5 August: Introduction to AI, Large Language Models (LLMs), Prompt Engineering and Agents.