Know What Will Work Before You Commit
We analyze the structure of your processes and tell you which ones can actually be automated, before you spend the money.
The Core Insight
When the logic of a process is written down completely, automation succeeds. When the logic exists only in people's heads, automation fails. This pattern holds across every case we've examined.
About 70% of large-scale automation projects miss their targets. This number has remained consistent across decades, technologies, and industries. The technology keeps changing; the failure rate does not. The question is why.
The answer turns out to be structural. When the logic of a process is written down completely, automation succeeds. When the logic exists only in people's heads, automation fails. This pattern holds across every case we have examined, from Ford's assembly line in 1913 to modern AI deployments.
The implication is that before automating anything, you should check whether the logic is actually documented. Most organizations skip this step, learning the answer during implementation when it becomes expensive. We do this analysis first, telling you what will work before you commit the resources.
Four Cases, One Pattern
These are well-documented failures with billions of dollars at stake. Each one followed the same trajectory: an attempt to automate work that had never been fully specified. In each case, the structural problem was identifiable before implementation began.
Phoenix Payroll
Canada's federal pay system. Nine years in, it still cannot reliably pay employees. The problem was not the software but the 80,000+ pay rules that existed only in specialist knowledge.
Read the analysisKnight Capital
A trading firm destroyed in 45 minutes. The deployment process depended on knowledge that existed in one engineer's head. When a server was missed, the system behaved in ways no one had specified handling rules for.
Read the analysisHertz/Accenture
A website that was never delivered. Requirements changed continuously because they had never been written down, with each review surfacing new requirements because the rules had never been made explicit.
Read the analysisCrowdStrike
A software update that crashed systems worldwide in 78 minutes. The validation process assumed things about production that were not documented, and when those assumptions did not hold, there were no explicit rules for handling the situation.
Read the analysisWhat We Actually Do
We analyze the structure of work to determine what can automate before you invest. Every information-processing task can be decomposed into combinations of nine operations: receiving data, sending data, storing data, transforming data, comparing values, selecting paths, combining components, imposing constraints, and establishing connections. These are the primitives of computation, derived from the mathematical foundations established by Shannon, Turing, and Wiener between 1936 and 1948. There is nothing else.
When a task operates on explicit, documented rules, its primitives can be automated. When a task requires judgment that has not been specified, it cannot. The ratio of explicit to tacit determines the automation ceiling. A process that is 90% explicit can probably be automated with appropriate human oversight. A process that is 50% judgment will fail regardless of what technology you use.
We map your processes, classify each decision point, and tell you what will work. We show you where the undocumented judgment concentrates, what would need to change for automation to succeed, and whether the investment makes sense given what the structure can support.
The People Who Hire Us
Three situations tend to bring people to structural assessment.
Before a major investment
You have a proposal on your desk and a vendor who says it will work. You want an independent evaluation of whether the underlying process can actually support automation, before you sign the contract.
After a failure
Something did not deliver what was promised. Before trying again, you want to understand what went wrong structurally, not just operationally. The operational post-mortem tells you what happened; the structural analysis tells you why it was predictable.
During due diligence
You are evaluating a company that claims certain automation capabilities. You want to know if those claims are structurally sound or if they depend on tacit knowledge that will not transfer.
In each case, the question is whether the structure of the work supports automation or not. The answer is knowable before you commit resources.
The Assessment
We analyze your process, classify its decision points, and give you a clear answer. Typical engagements take 4-6 weeks.