Dear Strategist,
In the midst of the current buzz about ChatGPT, let's pause and ponder: Is this just another tech wave or is it a game changer? Does it matter strategically?
Focus on obvious benefits
In all the technophoria, we sometimes forget that all technology can be both, help- and harmful. You can use your smartphone to quickly answer an email on the move, get directions and check the latest weather. Or you can let your smartphone turn you into a mindless notification addict wasting your hours scrolling through popcorn content.
Unlike smartphones, we have yet to understand when exactly LLMs are helpful. Are they just a waste of time - you staring at ChatGPT's suspense-creating word-for-word generation of a Shakespearean company poem?
LLMs like ChatGPT are a probabilistic technology - they generate an expected answer with some variation using large pools of data. Therefore, LLMs work best for problems where these four criteria apply:
Standard answers rather than extraordinary ones,
Creative rather than tasks with a single correct solution,
Text and image generation rather than numerical tasks,
Lots of data as base for learning rather than niche applications.
Given these (current) strengths, ChatGPT is helpful for marketing, publishing, programming, and - perhaps unfortunately - for writing solid school essays. But it writes uninspiring news columns and struggles with complex number problems.
Companies should focus using LLMs in areas where the above criteria apply, where the technology has obvious benefits at the moment; instead of wasting money on at-the-edge experiments. It may be that it does nothing meaningful outside a few high-impact areas.
Focus on effectiveness, not efficiency
Using LLMs for efficiency gains is, of course, a race to the bottom with an obvious limit of zero. This may be crucial for some low-cost strategies in certain industries such as legal or media. For most others, making marketing and HR more efficient is nice but not game changing - unless the freed-up resources are reinvested into improving effectiveness.
Effectiveness effects are much more interesting from a strategic perspective. Where do LLMs help us do the right things better?
Unfortunately, this is also where we have reached the thickest fog of uncertainty, the uncharted territory of the unknown. Will this technology create vastly better personalized yet automated customer service? Will it revolutionize the media experience with more and better (or less bad) content? From a strategic perspective, this technology is still in its infancy and developing in leaps. Strategists need to think hard.
Competitive edge by leveraging data
At first glance, LLMs seem to be a easily accessible technology with low barriers to entry. How the heck should companies create competitive moat with ChatGPT?
To create competitive advantage you have to combine this general-purpose technology with something proprietary. I don't believe it will be prompt engineering. Prompt engineering is just the new jargon for programming - and best practices are already emerging and spreading.
I think it will be the proprietary data you feed into the LLM. To create competitive advantage, companies should explore ChatGPT as a technology lever for their proprietary data.
Strategic choice: Early mover vs. fast follower
I have tried ChatGPT multiple times, for example, for writing newsletters like this one. Was it helpful? Somewhat, but I also invested quite some time in exploration.
Strategists have a choice: Invest resources in experimentation for early-mover advantage or wait until the dust settles and go for fast follower.
Purpose-built tools on top of LLMs ('Chat GPT wrappers') are already emerging. They could do the experimentation for you. You could just wait for them to mature - maybe already in 2025.
To create an early-mover advantage, experimenting with ChatGPT has to create a non-imitable experience curve. I have been thinking about this for a while, but I just cannot make up my mind if it actually does. What do you think?
Practical strategy nugget to take away: Strategists should explore ChatGPT from three perspectives: Where does it have obvious benefits? How can we combine it with proprietary data to create competitive advantage? Where can we leverage it for effectiveness gains?
Let me know your reactions and thoughts!
Sebastian
Whether IT can be the basis for competitive advantage is an old topic. On the one hand ERPs weren't because everyone could buy one and hire consultants to support implementation.
On the other hand internet was available to everyone. Any business could have built an e-commerce web site. And yet there's only 1 Amazon.
So is AI closer to ERP (more like CRM actually re: you excellent point about proprietary data) or the internet? So far I've seen several articles claiming AI/GenAI could redefine competitive advantage but all the examples were about operations effectiveness...
I'm also thinking about this as we move through space-time.
From first principles, I find it hard to believe that a widely available tool can generate a competitive advantage.
I tend to think about things in terms of fixed costs and market gross profits.
Higher fixed costs (primarily R&D and distribution) relative to market gross profits and customer captivity dictates the number of players who can profitably compete. So high fixed costs => higher profits.
In industries where R&D are significant today but can be dramatically reduced (as perhaps Software), the industry should become more fragmented because it can now room more players. If variable costs gets reduced, it has the same effect. More competition, less profits. In these scenarios, I would expect LLMs to be the inverse of a competitive advantage.
Depending on distribution and customer captivity.
If you still need to pay for displaying adds, then the things you can automate is only a fraction of the cost of distribution. And arguably, the savings will be reallocated to more ads and sales regardless. This should paradoxically dampen profit pressures, because new players have higher barriers to overcome.
In cases where you get a new paradigm (10x better product), you'll also have 10x more competitors. So speed becomes the crucial part (as it always has been). In this sense, maybe the LLMs makes the impact of relative skill less pronounced. It's an enabler in one sense, and a flattener in another.
And if the costs are reduced, businesses that are not viable today may become viable. Many niches, with capped profit potential. In a sense, these pocket may have inherent competitive advantages simply due to the size of the profit pool. But should be equally available to incumbents and new players.
If distribution can be automated or eliminated, customer captivity remains an isolating mechanism for incumbent. Microsoft has huge network effects and distribution advantages, and may be primed to capture both cost savings and revenue uplifts.
So here's how I parse it:
- In some cases, it softens existing competitive advantages by reducing fixed costs, which increases competitive pressures
- In some cases (new paradigms), it may level the playing field compared to earlier generations
- In some cases (new viable businesses) it may be source of competitive advantage if you capture a niche with a small profit pool
- In some cases (incumbents with customer captivity) it can strengthens their position
Sorry for the ramblings :D