#208 Gen AI and the end of BI?
Thinking different rather than better is hard but often valuable.
Imagine that you have a question about a specific event for which you need to go the public library and look at newspaper archives. You might go to an index, narrow your search down by region and date and sift through a bunch of local newspapers column by column and hopefully find your answer - if it’s there, it should take a few hours. It’s a scene out of a detective film from yesteryear. Now imagine a sci-fi version of the same film where you pull up a device in your home and ask it the question - and the system comes back in seconds to give you your answer.
One of the biggest challenges of innovation work is being able to think different instead of better. Better is how we're tuned to thinking. It's what Henry Ford called a 'faster horse'. Doing what we do already, but with improved performance. Cheaper, faster, more efficient, and so on. If you sell ground coffee in supermarkets - better leads you down the path of better supply chains, stocking improvements, store promotions, and packaging improvements. It also typically includes tinkering with the product mix - for taste and flavour, or branding and positioning efforts. It doesn't however take you to the point of thinking 'what if we sold ready to drink coffee directly to customers using a globally scalable model?' Which is why coffee brands at companies such as P&G were taken by surprise with the success of Starbucks, and all the other now-ubiquitous coffee chains. And ironically, a few decades later while Starbucks were trying to figure out how to make their model better - better retail, better service, better experiences, better product mix, they didn't think different either. For example, could we give the same experience of fresh coffee to consumers at home? Nespresso did, and suddenly the coffee chains were losing a section of clients to home brewed coffee on kitchen counter friendly coffee machines. This is about the art of thinking different.
The folks at ARIA also had an interesting phrase - at the recently concluded CogX, they were talking about research that creates 'not just a new product but a new industry'. Creating a new industry requires a fundamental rethink of something that we will all now do differently. This too, can be at varying levels of scale. Google maps changed how we navigate on roads, globally. Social media changed how we stay in touch with friends and family. Mega changes, or what Suleyman Mustafa calls 'waves' are even more seismic in their change impact. The Internet, the automobile, the printing press are examples of things that effectively changed the course of civilisation. It created 'different' at a truly global scale.
Which brings us nicely to AI. It seems like AI could be the next big megachange that alters the course of our collective history. It has potential implications that have captivated the worlds greatest leaders, scientists, historians, and business people. Within the broader sweep of AI systems, Gen AI has become particularly topical - largely because of its easy accessibility. And given what Gen AI is good at, it's natural that people are drawn towards areas like summarising documents, writing essays, and in chatbots and virtual assistants. All of which are excellent examples of getting better. But how can we think different? Where could Generative AI change business significantly? Sometimes it helps to reframe the question, and as I like to say, look at the problem "sideways".
I asked myself this question - what is the super-power of GenAI? And it's this: Generative AI understands language and context. It can decipher your questions and intent, no matter how you ask it. And it can translate your question to a set of specific instructions to any other complex system. For example, rather than pore through the manual of a car or any domestic appliance, you could just ask a question. 'How do I replace the engine oil'? And Gen AI can act as the intermediary and fetch you the answer. So you don't have to go looking through the index and the dozen references in the manual you have to scroll, to find the right answer.
Let's extend this a bit more. Think of GenAI as your interface to any business system. In any enterprise, there are dozens of major systems, and possibly a hundred different smaller systems and tools that are used to run and manage the business. All of them hold data and information that you need. Gen AI should be your interface to all of their data. No more having to run SQL queries, or following specific and often complex processes to extract insights from them. These complex instructions usually require technical knowledge, so you're also reliant on your data team. Historically this has led to an entire industry around reporting. BI (Business Intelligence), MI (Management Intelligence), Analytics, predictive and prescriptive analytics, real time data, all of these terms are supported by a plethora of products and armies of data engineers and analysts. An outcome of this is just-in-case reporting. Every business generates encyclopaedic amounts of reports, just in case somebody needs some piece of data. Not only is it inefficient, as most reports go unread, it also has a poor signal to noise ratio, since you have to wade through a lot of information before you find what you need.
Why not end all of this often-mindless reporting, and focus on what you really want to know. If you're in retail and you want to know which stores had the highest or lowest footfall in the past 24 hours, or where the inventory is currently at its highest, simply ask the question, and you should get an answer. Assuming that the data exists somewhere in the business, GenAI should be designed to create your interface with your data. Companies such as seek.ai and Thoughtspot are looking to do just this. I recently saw a demo at client where the queries were quite ambiguously structured, such as 'which ski resort destinations have the highest profitability' - so you're expecting the system to pick out 'ski resorts' or 'beaches' without specifically naming locations. The solution also uses Gen AI to decide what the best format for presenting data is.
It's not as simple as implementing ChatGPT inside the business. The heavy lifting of stitching together data from multiple systems, and relational tables still needs to be done and won't be done by GenAI. But GenAI can act like your executive assistant for data and insights, and trigger the right systems and processes, so all you need to do is ask a simple question in natural language. But I think this is how tools will be set up, and this potentially is an area where GenAI can help us be different, not just better, in businesses large and small.
If you want a reminder of how much of a change this can make, think about the humble remote control. Earlier, if you wanted to find a program on your TV system, or on Netflix or Amazon, you would need to navigate to a virtual 'keyboard' and type out the search by navigating your way around the keys. Clunky and time consuming by any yardstick, but it's what we were used to. Something like this:
Sky tried to improve this by creating the alphabetical bar. Better, but not different.
Then came the next generation of Sky Boxes, with a voice search built into the remote. Now suddenly you didn't have to type, or navigate. You just pressed the blue button and 'said' the name of the program. And the system found the match(es) and showed you the results. I think that qualifies as different, rather than better.
Imagine a similar kind of power for your business - a simple and intuitive way of asking for complex information. Which customers have the highest amounts outstanding? Which suppliers have negotiated the most favourable terms of payments? What was the age and gender split of people who visited our website last month? All you need to do is ask the question, and the AI system will do the rest. Now, isn't that different?
AI Reading
Science: The Economist writes about how AI can revolutionise science - from literature based discovery, to self driving labs. After all every major scientific tool invention such as microscopes or crystallography have paved the way for a number of consequent and bigger breakthroughs. And this piece lists a number of areas where this is happening - from drug discovery, to battery tech, and from understanding the cell-nucleus to tackling climate change. It also points out why scientists won’t be out of jobs soon! (The Economist)
Biology: Mark Zuckerberg and Priscilla Chan, through their foundation CZI are looking to use AI to better model the cell, and create virtual cells for simulation and research. Here’s their piece on the subject - it’s a useful reminder of ecosystem of research at play, and also talks about specific diseases such as cystic fibrosis and the impact of SARS Covid. (MIT Technology Review)
Coke: Will your next can of sugary beverage be designed by an AI? Coke is exploring this. But they’re not alone - so are Kraft, Mars, and KFC. (Quartz)
Leaderspeak: This tongue-in-cheek piece asks whether LLMs like ChatGPT should do your earnings call! It turns out the LLMs are better at being bland and saying a lot without actually communicating specific info, than human beings! So it depends on whether you have good news to share or bad news to hide! (FT)
Improvisation: A study by the Universities at Bergen and Stavanger, both in Norway, found that AI ( 3 LLM based AI models) were on average better improvisors than human beings, when given tasks like identifying new uses for everyday objects. Although the best humans were better than that AI system. The takeaway is that AI can perform creative tasks, but specialist humans might still be better. As of now. (IEEE Spectrum)
Economy: According to a Goldman Sachs report, Gen AI can provide up to a 7% Global GDP boost, based on the task analysis of over 900 jobs. It also suggests that 60% of today’s jobs did not exist in 1940, so an overwhelmingly large % of new employment is driven by emerging tech and the jobs that it creates. (Goldman Sachs)
Extinction Threat: a provocative piece in the WSJ points to the rift between leading thinkers as to the extent of existential risk to humanity posed by AI. The more near term implication on disparity, inequality may be more real. (Wall Street Journal)
Health & Ageing
I was disappointed and a bit surprised to note that Babylon went bust recently. Babylon was one of the real success stories of the early health tech era. Surprising at it comes at a point when medtech and healthtech are still booming. Not surprising that the piece points to issues inside the business. Still it tells a story for a business that was valued at $2bn in 2021,and growing at 400% annually. (Techcrunch)
Private Equity in Healthcare: What happens when care homes have leveraged buy outs? A paper on the impact of Private Equity on healthcare - including hospitals and care homes suggests that PE ownership creates a rise in healthcare costs to patients and a drop in quality. You can read the summary on FT, or the orignal paper on the BMJ.
US Gerontocracy: One of the more interesting areas connected to ageing is its impact on politics. If the next US elections feature another head to head between Trump and Biden, they will be 78 and 81 years old. This piece talks about the wider age skew within the US political environment, and the impact of having an over-representation of the age group, and hence the phrase gerontocracy. (FT)
Other Reading
Failure: Tim Harford talks here about the ‘art of making good mistakes’. Mistakes and failures aren’t black and white. How that is done at an individual and organisational way is where the magic happens. The short answer is that you need to make errors often and own them. Success comes in learning fast from the errors. Here’s a video of Harford talking about the same thing. And in case you wanted more proof, here’s Ed Sheeran saying the same thing. (FT/ Youtube)
thanks for reading and see you soon!