Whether you are a manager, investor, policy maker, or analyst some hue or the other, scenario analyses is part of your job portfolio. Analyses not just analysis, because the world is open to multiple interpretations, some credible, some beguiling. Since ‘analysing data’ has become easier and faster with some AI agent, as even Excel too will be embedded with AI, the job of all these professionals has more to do with building scenarios. How AI can assist you is really a function of the breadth of your thinking.
There is no one ‘text’, but a vast array of texts different in multiple ways. Embedded deep within such a world of texts is a range of insights, pointers, indicators, masks to be unmasked, prejudices deliberately populated, innuendos and the downright bigotry. All crying to be used in some interpretation or the other. Perhaps, more.
In his Tenth Theses of Feuerbach, Karl Marx remarked that ‘philosophers have so far only interpreted the world; the point, however, is to change it’. Just like rebels and revolutionaries intent on ‘changing the world’ by attempting to mould it towards their chosen goals, so does every management of every company and even governments.
Goals and interpretation; interpretation and goals
There are two ways of going about this.
Choose your goal and then search for an interpretation to ‘justify’ the goal you have chosen and drive the world towards accepting it. Typically, such a strategy is built on specious arguments – superficially plausible but actually false or wrong, in the underlying belief that repetition of the same will usher in belief and faith, leading to the desired action. Both governments and companies have been known to have resorted to this strategy, with success a function of how well articulated is the specious argument. Much as it is distasteful, this strategy too does need skill in ‘creating a narrative’ and convincing a large number of people that this is the truth.
On the other hand, you can look at more than one plausible interpretation and their directions and then choose what might be a more realistic goal among a range of possible goals – what will enable, what will disrupt. There is no doubt that this too needs moulding and convincing, but it is of an entirely different character altogether.
Scenario Analyses
Armed with this preface, let us turn to our topic.
No student of management and investment can escape building ‘Scenarios’ likely to prevail in some defined future where a strategy – business or investment – will succeed or fail.
Invariably, a scenario analysis is indexed to some factor. For example, a bond trader or investor would like to visualize what happens if the interest rates moved up or down and by some basis points (100 basis points is one percent). Much will depend on the current or prevailing interest rate regime; we say regime because there are many interest rates.
Or a fast moving consumer goods company will like to see consumption patterns indexed to income growth, family size and cities, to mention one possible index; indexed factor can be a combination because factors tend to influence in baskets rather than as an individual factor. Finding the combination that is your indexing factor is a matter of analysis.
For example, in FMCG, a crucial factor is the average repurchase period, which defines the frequency of buying and therefore the size of an individual unit of purchase. Thanks to Q-commerce and rapid delivery, families need to buy only as much as they need for a day or at best two. More so as families live in small houses. Thus, an FMCG company will likely project how Q-commerce will grow and the cities with small houses.
Where does AI fit in?
The greatest strength of AI, any AI, is abstraction, as I have mentioned in many of my writings. It is this ability to abstract from substance that enables an AI to find the right ‘substance’, which may be interest rates for a bond trader, fuel consumption for an auto analyst, consistency in public/government policy for a policy analyst and so on. The challenge for you is to find the scenarios indexed to the factor you have chosen. And such analyses itself can be executed at multiple levels of granularity.
The world of texts is invariably populated with deliberately misleading ‘indexing factors’. For example, an investor who intends to ‘short’ an asset (short – to sell something you don’t have; you are betting that prices will move down) will portray a bleak picture of the prospects for that asset.
This is not groundless thinking but rooted in reality.
Petroyuan and petrodollar
Ever since the media began to cover news articles on the possible ‘emergence of a sustained petroyuan’ as a rival to petrodollar, a carefully orchestrated media campaign has been carried out in many of the world’s leading media outlets. You will find ‘the myth of the petrodollar’, ‘conspiracy theory’ and so on all with a view to protecting the dollar by arguing that the strength of the dollar is not reducible to one factor – invoicing of oil and gas in dollar.
Such ‘texts’ become part of the universe of texts that an AI will have to ‘go through’ and present a scenario that is credible. How good is the outcome is a function how well you use the AI agent: it is now commonplace knowledge that every AI agent is extremely vulnerable to ‘prompting’. You will need to be aware of all other ‘relevant data’ to discern which is a faithful portrayal and which is ‘beguiling’. Not an easy task, even with AI!
AI and Excel
Again, to repeat what I have noted several times, what Excel can do an AI agent can do on a larger scale: analysis of data. Anyone who has used Excel would know that you can analyse the ‘whole’ and break it down to parts as you choose to define. Every aggregation permits disaggregation. Clusters of data is a matter of definition. In Excel, you can use Pivot tables.
It is entirely possible that different combinations of data point to different scenarios. Your choice is then a function of probability. Whether the AI agent can estimate the probability is something you will discover when you employ an AI agent.
You can have analyses of data with greater accuracy, subject to your choice of the indexed factor. For a portfolio manager, what is of immense interest is the correlation among classes of assets. Using the known correlations, you can visualize scenarios that will sustain the correlation or disrupt it. You can use historical averages or incremental averages, especially indexed to recent changes, or whatever.
The ability to use clusters of data appropriate to a chosen goal is vital and that is a matter of human skill. As yet. Amen!
