Underappreciated: instrumentation of business processes

Leaders at large long lived organizations (so, most big companies) generally fail to appreciate one pervasive advantage that Internet- and especially cloud-era companies have on them: almost all of their business processes are instrumented by default.

What I mean by "instrumented": when you build a process using cloud-based services (like the hundreds of compute and storage and processing services offered by AWS, Azure, etc.), whenever that process is used, a rich, timestamped, and very granular log gets recorded. For example, when a chatbot or an AI voice chat service helps a customer, every thing that automated tool gets as input, looks up, uses as logic to determine a response, says in reply, etc., is all recorded. (By contrast, if Jesse, a 23yo employee of a traditional business performs the same kind of customer service, Jesse's logic is not recorded, and many other details will be lost unless the systems he uses are designed to log them.) Because they were designed after storage space became effectively unlimited and ultra-cheap, these cloud-based logs are high volume and richly detailed to a degree that would shock owners of legacy on-prem computing systems. This rich logging thus supports much more useful, real-time, straightforwardly causal analysis of how a business process is performing.

Having deeply instrumented business processes fundamentally changes the kind of management needed. Most importantly, it reduces the need for and value of expert judgment made in the face of uncertainty. Essentially, management becomes easier, and less of it is actually needed per unit of "business" (whether measured by volume, process complexity, or whatever).

I first noticed a similar phenomenon about 15 years ago: that banks and other businesses that made most of their money through financial risk-taking had a management advantage missing in most other industries. Banks could measure (more or less) both profitability and risk in pretty granular ways, often down to individual transactions. You could usefully predict the return on risk for a mortgage loan, or a bond trade, or even approving a single swipe of Joe's credit card, to buy, say, coffee at 6:30AM, or alternately for Joe to grab 30 minutes in the champagne room at his local strip club at 2AM. Much of the success in such businesses comes from picking which risks to take on. Because the banks could measure this critical input into success with absurd granularity, many of the usual judgements that middle and senior management needed to make were obviated. Management in banks with good granular risk and profit measurement was quite simply less necessary than in other industries, and they could scale more with less management.

The same is true for firms that build with cloud-era tooling.