Fish Food: Episode 646 - Using GenAI in Strategy

An AI strategy workflow, Meeker's mega-deck, slower-than-expected AI impact, the death of daydreaming, and interviewing English Tudors

This week’s provocation: Integrating AI into the strategy workflow

I’m a big fan of the idea of AI amplifying human capability and have long been working to fully embed AI into my strategy work. For me it’s been a real game-changer. It’s not just about efficiency, although it is like having a mini-team of researchers always on hand which is amazing. It’s also that it gets me to places that I don’t think I would have got to on my own.

So I thought it would be useful to talk about how I’ve integrated AI into my strategy process. I’m going to refer to my use of ChatGPT Project Space as an example for this. I’ve only really been using it to its full potential for the past few months but I honestly think that it’s one of the most undervalued and underused features in ChatGPT (it’s over there in the left sidebar). There’s a particular reason why I find it so useful - it provides an evolving strategic thinking environment with a persistent context and AI fully embedded. Establishing a defined context for every project space which remains persistent across multiple conversations and files is particularly helpful. It’s like having a thinking partner that you can pick up a conversation with at any time, but who instantly remembers your project objectives, angle, research and the work done so far.

Project Spaces is only available to the paid tier (ChatGPT+) and the Teams and Enterprise versions, and to my knowledge Gemini and Claude don’t have equivalents (although Notebook LM is a very useful research tool), but I think many of the principles of what I’m talking about here have broader application. I’ve summarised a few key techniques for how I use it to create long-running, named workspaces (kind of like a combined digital whiteboard and research analyst), and to make it as useful as possible I’ve broken down some key steps, and given you a downloadable template at the end.

Start with a ‘project charter’

Start by naming the project, and then under ‘Add instructions’ you can set up specific contexts to inform the GPT and tailor how you want it to respond throughout the project:

• Objective: explain what you’re trying to achieve, goals and deliverables, key questions to answer

• Context: brief but relevant background to the project itself, such as why it’s important now, relevant challenges or opportunities. You can also set a strategic context for the GPT to ‘hold in mind’ throughout e.g. key priorities that you want to focus on, or known constraints and things to avoid.

• Timeline: A rough timeline can help. For example research (week 1), strategy design (weeks 2 & 3), final recommendations (week 4)

• Role and tone: I’m a big fan of giving AI context by asking it to play a specific role in prompts so here you can use the ‘add instructions’ box to set a default strategic mindset and tone for the project. For example, you could ask it to assume the role of a world-class strategy consultant specialising in a particular field, or to adopt a challenger mindset throughout the project. You can also set out your preferences for communication style and how you like to see outputs. And you can inform it of your preferred working style (e.g. ‘I like to collaborate iteratively and may ask you to refine or build on your outputs’ or ‘actively help manage the project by reminding me of past threads if relevant’ or ‘propose next steps if you see a logical progression’).

This set up is key as it establishes the defined context for how the GPT can add value to your strategic process, and it means that it retains a consistency of perspective and that you don’t need to rebrief it every time.

Curate your inputs

The project space can be used to pull together all kinds of inputs from research reports, to transcripts, to persona docs, strategy decks, customer journey information, PDFs, stakeholder interviews, briefing docs, brand guidelines, email summaries, meeting notes and so on. Curating these inputs is (for me at least) a key part of the strategic process as it enables you to synthesise insights from across a fragmented set of source material. I think of this as a repository for anything that may be relevant to the project.

As well as summarising key insights you can use this to give context, identify patterns or potential areas of tension. In a way you’re creating your own RAG (Retrieval-Augmented Generation: finding relevant information from specific knowledge sources, and using that to give context for the LLM to generate more accurate responses). Remember that your curated inputs will provide important context throughout so it’s important to keep these updated with new information as you get it. I’ve actually gone as far as creating cheat sheets for different strategic concepts or models that I can use as inputs to inform how the GPT thinks about specific contexts.

Toggle between more specific personas as you go

Once you’ve set the project space up in this way you can begin your analysis. The classic ‘4Cs’ (Customer, Company, Category, Culture) is a great way to delineate different research threads that you can then bring together. It can be helpful to use personas and to assign specific roles to the GPT (and even defined mental models) for different parts of the project. For example, asking it to act as a competitive intelligence analyst or a McKinsey strategy-consultant or a behavioural economist.

Toggling between different perspectives like this is great for testing thinking from different angles and getting diverse viewpoints. I’m an independent consultant and so tactical personas can amplify my ability to bring in different perspectives or approaches. It’s like having a multidisciplinary team at your disposal. You can think of the role that you put into ‘Add Instructions’ as a strategic co-pilot that is always there and context aware, and the tactical roles as temporary invited guests with individual skills to enrich or reframe thinking.

Track strategic thinking over time

You can build a kind of narrative arc of strategy as you go. Separate threads can be labelled and used for different stages of the project. For example, phase one might be research and landscape scanning, phase two could be emerging insights and tensions, phase three might be strategy synthesis, and phase four could be a communication or execution plan and an SLT-ready narrative.

You can ask it to cross-reference findings across inputs or threads and to keep a running log of key insights as you go. At various points you can ask it to summarise what has been discovered so far, or to reframe project goals based on the insights that have been uncovered, or to define the key threads that have emerged from the analysis. This tracks momentum but also helps with clarity and structure, and makes iterative learning possible and visible.

Co-developing frameworks and roadmaps

Being honest, this takes a bit more work to get it good, but it can be useful to see how the GPT assimilates information as a matrix, or a framework, or another output. You can use these at different stages of the process. I may use them as inspiration but I usually end up doing the final work myself as this is an important part for me to think through rather than the AI. But it can provide a useful starting point.

Running scenarios

Bearing in mind you will have loaded the GPT with potentially significant amounts of context and information to work from it can be really useful to run scenarios or foresight exercises as part of the process. For example, you could ask it for three future scenarios based on weak signals or insights which you’ve defined, or to set out what black swan events are most likely to disrupt the plan, or the second and third order effects of a competitor entering the space.

Distilling and packaging outputs

Again, this one takes a bit of work to get good but you can easily generate summary slide decks, roadmaps or other outputs. I prefer to get it to suggest structure and essential flow before compiling the deck myself and, as I think I’ve said before, I always like to write myself rather than allowing the AI to do it as writing is how I think and learn. But it can provide a great starting point from which to work.

The key thing that I’ve found is that using Project Spaces makes it far easier to truly embed AI throughout the process, and to build a context-rich chain of thought. You don’t have to deliberately remember ‘Oh, this might be good for AI to input into’ because it’s right there. For me, it’s kind of like having a team of research assistants that are always up-to-date with the context for the project and understand the full journey, or a living notebook, or a strategy room that you go into to think.

The biggest lesson for me? Rather than thinking about AI just as a productivity tool, you need to view it as a thought partner and a thinking environment.

I’ve created a full template for how to use ChatGPT Project Space in the strategic process, which you are free to use and which you can access here.

Rewind and catch up:

How streaming changed music

AI, and inflection points in the creative industries

Thinking in systems

Decoding the value of AI

Photo by Octavian-Dan Craciun on Unsplash

---

If you do one thing this week…

For almost 20 years Mary Meeker used to create an annual mega-deck on the state of digital that everyone would go nuts about for two weeks a year. She moved into becoming an investor and VC and hasn’t done one since 2019 but has just released a brand new mega-deck on the state of AI, and mega is the word - the full report is 340 pages long (!). Lots of graphs going up and to the right, and a diverse range of useful data points and charts covering AI adoption, spend, model capability and cost trends, AI’s impact on work, and much more besides. Cue lots of AI LinkedIn gurus using AI to summarise the AI report (very meta). The best of these that I saw came from Ross Dawson.

---

Links of the week

• Every year IBM release a big old Global C-suite survey which is a pretty good barometer of what the most senior people running the biggest organisations in the world are focused on. One of the most interesting findings is that AI initiatives are progressing slower than anticipated: ‘Last year, more than two-thirds of the CEOs we surveyed said they expected to move beyond AI pilots by 2025. But this year, 60% say they are still in the piloting phase.’ A reflection perhaps of the initial over-hype and the reality of how hard it is to drive real change in big businesses

• OpenAI have opened up an academy and there are 11 free courses on there including ones on prompt engineering, using Deep Research, reasoning and projects (see above)

• In a recent poll 77% of Americans said that they preferred slowing down AI development to ensure it is done correctly, even at the cost of delayed breakthroughs. Do we really need to go so fast with progress in AI?

• A good counterpoint to consider in the AI-will-take-white-collar-jobs debate

• Sangeet Paul Chowday has a wonderful way of bringing concepts to life through stories (something I try and do in my talks), and I loved this post on how NOT to sell AI into the enterprise featuring the East India company, barcodes and treasure maps. His point about mapping systems is also brought to life in this story about London’s Cholera outbreak in 1854.

• Thought provoking read of the week - Christine Rosen on the death of daydreaming (with a foreword by Jon Haidt (who wrote The Anxious Generation)

---

And finally…

I loved these AI-generated interviews of people from the Tudor period in England (1485-1603). They’re hilarious. But also, what a potentially amazing way to learn about history.

Weeknotes

This week I ran the latest version of the IPA Application of AI in Advertising course which was lots of fun, and also did a session with my African bank client and had a few new business meetings. I’m coming up to another period of work travel (a mini-European road trip this time) so I’m enjoying being in one place, working away, and happy that it’s finally rained.

Thanks for subscribing to and reading Only Dead Fish. It means a lot. This newsletter is 100% free to read so if you liked this episode please do like, share and pass it on.

If you’d like more from me my blog is over here and my personal site is here, and do get in touch if you’d like me to give a talk to your team or talk about working together.

My favourite quote captures what I try to do every day, and it’s from renowned Creative Director Paul Arden: ‘Do not covet your ideas. Give away all you know, and more will come back to you’.

And remember - only dead fish go with the flow.