We recently completed our initial internal AI manifesto for SCG. Simply put, it’s a set of guidelines that helps our teams navigate through the noise and produce better outputs for our clients and audiences.
It generated a lot of great discussion, so I thought we should share the key points as others might find it useful.
Manifesto starting point
Members of the senior team prepared all the key points, and we then rolled it out as a 90-minute interactive session with the wider team.
Firstly, we needed to provide overall context: AI isn’t going away, and we will embrace it, but on our terms and with the teams firmly in control.
The key points we discussed were:
- AI is here to stay, and we must clarify our understanding of how to use it effectively across the business.
- AI can be used in every required thinking-based task, evolving at a pace that will only accelerate over time. As a reference, ChatGPT.com is currently being visited 5.2 billion times per month.
- Looking ahead, it is essential to provide guidance on using AI at SCG and what we consider best practices. We already utilise AI, but frameworks are needed, especially due to the speed of how the technologies will advance.
- This is especially important for the content and creative aspects of the business. Our clients and readers pay us for quality work – that can’t be diminished by offloading critical thinking work to a bot.
- Free-range humans have to be front, centre and in control!
Two groups
Next, we mapped out the two user groups that are evolving globally and confirmed which group we need to be in:
- One leads to shortcuts and potentially lower-quality outputs, while the second collaborates with AI but knows “what good looks like.”
- As a company, we need to be in the second group.
To bring those groups to life and be more specific, we fleshed them out:
Group One: AI cruisers
- AI Cruisers get ChatGPT to draft the copy, post, article, or email, give it a glance over and then copy-paste into whatever channel they are working in.
- They start by doing this a couple of times when they’re busy, then get used to doing it daily for everything. Things happen faster, which is excellent, but people notice something is “off.” What is “off” is that you have outsourced all of your work to a bot, and what has been generated lacks your distinct skills, craft and voice.
- Additionally, someone could run what has been created through an AI checker….
- This is not the way. It reduces the quality of what we produce and sidelines your talent.
Group Two: AI collaborators
- AI collaborators work with AI to produce high-quality outputs, while remaining in complete control of the task.
- Examples:
- Just done some banging copy for one of our websites? Get AI to give you some tips on structuring the article for SEO optimisation.
- Developing a draft content calendar? Give AI your framework, including as much detail as possible and ask for initial pointers, which you can then refine with various prompts to shape with additional information etc.
- Got to write a job specification? Put down the main tasks of the role, get AI to do a draft job spec and then refine and craft to deliver something compelling.
- Writing an RFP? Once you have completed your initial draft, upload it and then upload the client’s briefing document. Ask AI if it believes you have missed any areas.
- AI collaborators use AI to delve deeper and push beyond obvious insights. Don’t just use AI to make stuff– use it as a partner that challenges your thinking.
- Finally, never let anything out the door without questioning AI outputs, asking yourself: “Is the correct?” and “Is this me?”
- AI cannot “invent” new ideas, so make sure you are front, centre and in charge.
This part generated some really interesting conversations, but everyone agreed that we all need to be in the second group, otherwise our quality of work and therefore our value diminishes very rapidly indeed!
The final part of the session focused on specifics, providing the team with tangible dos and don'ts. (Obviously, this will evolve.)
This is where we ended up:
Approved uses for AI collaboration
- Administrative
- Pitches – especially tenders
- Awards entries
- Job specs
- Initial collation of client briefs
- Data analysis
- Creative
- All tools within the Adobe suite – Firefly (including image generation)
- Writing
- Assistance with headlines? (This point prompted good discussion about how AI suggestions can be very dull, but a useful starting point)
- Copy tidying up/grammatical polish
- SEO suggestions
- Content repurposing
- Drafting social posts from longer-form content that you have created
- Resizing/batchwork of imagery
- Frameworks
- Content calendar draft assistance
- Social calendar draft assistance
Unapproved uses for AI collaboration
- Article and script writing
- Final editing/proofing/fact checking
- Using off-platform (non-Adobe) tools for image generation due to potential copyright issues
We also talked about the need for our contractors to agree to the above and update any contracts and commissioning docs.
That was the bulk of the main points from the session. We then went into specifics around the various tools we have available and the resources to support them. (We are running a complementary stream of work centralising a core suite of AI tools and developments for everyone to use.)
AI is a transformative technology with some incredibly powerful tools. However, you still need skilled craftspeople and decent frameworks to utilise those tools effectively, producing outputs that are unique and worthwhile.
You can’t just give a bunch of kids some high-end power tools and ask them to build a house. 😉
If you have any questions or comments, feel free to get in touch!