We have watched the operations of investment banking change across several decades, and the distance travelled is easy to underestimate from where the industry now stands. In the 1990s, an institutional order or indication of interest moved largely by hand, captured on a terminal, taken over a phone, and re-keyed into separate vendor networks that shared no common format. Within a few years the same flow had become straight-through, captured once and propagated to many systems without manual intervention. We were fortunate to work inside that transition, and what stays with us is less the systems built than the conditions under which they were built.
The foundations that were not yet there
Almost none of the foundations a modern engineer assumes were in place. Secure communication was assembled by hand rather than drawn from hardened, audited libraries. Databases failed often enough that a skilled administrator was a permanent presence rather than an occasional advisor. Public internet access was absent from the trading desk, so the body of practice available to an engineer was the body of practice inside the building. Open source had not yet arrived as a respectable option, and almost every component in production was proprietary, licensed, and supported by a vendor who would send staff to discuss how their system might fit alongside yours. Compute was scarce, the integration technologies of the day were young, and standard utilities that now ship with every platform had to be written from first principles.
An appetite for emerging technology
What we admire most about the period is the readiness of the banks to grasp technology while it was still forming, and to treat it as a source of competitive edge rather than a cost to be contained. Each new tool arrived without an established practice to inherit, and engineers learned it in the same months they shipped production systems on top of it. Investment banks, often caricatured as conservative, were among the most ambitious early adopters of emerging software, willing to commit serious capital and serious talent to technologies that were months rather than years old. The firms that grasped each new tool early took ground from those that waited, and the advantage compounded. Rapid development was not a methodology imported from elsewhere. It was the ordinary condition of the work, conducted under tighter constraints and with thinner tools than a modern team would accept, and it produced systems that ran a global business.
Hardest to convey is the speed of the resulting transformation. Moving from manual workflows to straight-through processing was not one decision but hundreds of separate ones, taken across competing firms in roughly the same direction within a window of a few years. Viewed from inside any one bank it looked like a sequence of internal projects. Viewed across the industry it was a phase change, an entire business reorganising how it moved information in less time than a single system now takes to mature.
What endured, and what we keep
Much of the specific technology has gone, and what replaced it is better in almost every measurable way. The FIX protocol endured. So did the integration problem, now lodged in gateways, brokers, and schema registries rather than solved away. Security improved out of recognition once communication libraries became hardened and standardised. Databases learned to manage and heal themselves. Open source moved from suspicion to default, and the cost of running the industry fell with it. Compute became abundant. Scaffolding that the 1990s lacked is now in place, and the work has risen to higher problems as a result.
What we draw from the period is a lesson about how technology and the people who adopt it mature together. Capable engineering produced working systems before the tools were powerful, through depth of ownership and a direct relationship with every failure. Those foundations are a genuine gift now, and they make excellent work faster and easier. They also make optional a discipline that was once compulsory. The organisations we most admire, then and now, are the ones that keep that discipline even when the scaffolding would let them skip it, and that commit to a new technology while adopting it still looks optional rather than waiting until it becomes urgent.