Why operations expertise never made it into software
Every operations team has a handful of people who simply know how the work gets done — which exceptions matter, which carriers to call, when to break the rule in the manual. That knowledge is the most valuable asset the team has, and it has never lived anywhere but in their heads.
Operations expertise never made it into software because the systems companies bought store the result of a decision, not the reasoning behind it — and that reasoning is tacit, the judgment experts can perform but can't fully write down. So it stayed in people's heads, and software stayed a place to record work, not do it.
Systems of record store results, not reasoning
A system of record captures what happened, not why. It stores the approved invoice, the booked load, the closed ticket — the output of a decision, with none of the judgment that produced it. The TMS knows which carrier got the load; it doesn't know the dispatcher picked them because they're the only one who answers on a Friday. The result is legible to software; the reasoning never was.
The best operators can't fully explain how they do it
The deeper reason the knowledge stayed put is that it's tacit — the expert can do the work and react to the case in front of them, but the rules live below the level of articulation. Ask a 20-year dispatcher to write the manual and you get a thin sketch that misses everything that actually matters: the exceptions, the escalation paths, the definition of a good outcome. You can't type what you don't know you know. That's why the most capable system in most companies is still a person, not a platform.
Why process docs and RPA never captured it
Two decades of process documentation and robotic process automation tried to close the gap and couldn't. Documentation captures the happy path and goes stale the moment the work changes. RPA automates the steps you can fully specify in advance — and operations is mostly the steps you can't: the carrier who replies in an unexpected format, the invoice that doesn't match, the exception no script anticipated. Both encode the parts of the job that were already easy, and leave the judgment — the actual expertise — untouched.
And the expertise is walking out the door
The knowledge was always fragile, and now it's leaving. Ten thousand experienced professionals retire every day, and 1.9 million operational roles sit unfilled in the US alone — you can't hire the expertise back fast enough to replace what's going. It's also why 95% of GenAI pilots deliver zero P&L impact: the models are capable, but the domain expertise they need lives in people, undocumented and absent from every dataset. The cost of the gap runs to roughly $260B a year in operational inefficiency.
Capturing it is an elicitation problem, not a scraping problem
You can't scrape knowledge that was never written down. So capturing it is an elicitation problem — one drawn from cognitive and behavioural science, not data engineering. Evos captures expertise the way an apprentice would: by working through real cases with the people who run the desk, surfacing the decisions they make without noticing, and structuring them into capabilities an operator can run. How that codification works is a discipline in itself, which we cover in codifying expertise at scale. The point here is simpler — it never made it into software before because everyone treated it as a data problem, when it was always a human one.
FAQ
Why did operations expertise never make it into software? Because systems of record were built to store the result of a decision, not the reasoning behind it, and that reasoning is tacit — judgment experts can perform but can't fully write down. It stayed in people's heads, and software became a place to record work rather than do it.
Why didn't process documentation or RPA capture it? Documentation captures the happy path and goes stale when the work changes; RPA only automates steps you can fully specify in advance. Operations is mostly the steps you can't — the exceptions no script anticipated. Both encoded the easy parts and left the judgment untouched.
What is tacit knowledge in operations? It's the expertise an experienced operator uses without being able to fully explain it — which exceptions matter, which carrier to call, when to break the rule in the manual. It's real and valuable, but it lives below the level of articulation, so it never got written down.
How can that expertise be captured now? As an elicitation problem, not a scraping one — working through real cases with the people who do the job and structuring their decisions into capabilities an AI operator can run. Evos captures it live and deploys operators built from it, live in 24 hours.
Book a demo
See what capturing your team's expertise looks like on your own operation, and get projections built from your own data. Book a demo.