By James Proctor, Co-Founder and Managing Director, The Inteq Group
You streamline approval workflows without losing control by replacing uniform approval with risk-calibrated review: inventory every approval point, trace each one to its origin and to the specific exposure it supposedly controls, keep and deepen the approvals that map to real risk, and retire the rest into monitoring. Control is not reduced by this. It is conserved and concentrated, because scrutiny moves from being spread evenly across everything to being applied deliberately where exposure actually lives. What gets reduced is latency, and the two are not the same thing, however much your approval matrix has taught people to conflate them.
Why Do Approval Chains Resist Streamlining?
Because of an accountability asymmetry nobody designed and everybody obeys. The person who removes a signature owns the next incident, whatever its cause. The person who keeps a useless signature owns nothing, ever. Under that asymmetry, approval chains only grow, and here is the test that exposes the result: ask anyone in your organization to defend the full approval matrix on a routine authorization, signature by signature, in terms of the risk each one controls. Nobody can. Nobody in your company can defend your approval matrix, and everyone enforces it, which tells you the matrix is not a control design. It is an accumulation with a compliance culture attached.
“Every signature you keep for comfort is latency you charge to customers.”
What Is the Subtraction Method?
Five steps, in order. Inventory: list every approval point in the process, with its nominal purpose. Trace: for each, establish its origin, an incident, an audit finding, a delegation limit from a prior structure, and the exposure it claims to control. Test: for a sample of real cases, ask whether the approval changed the outcome, ever; an approval with a near-zero intervention rate is not controlling anything, it is witnessing. Redesign: map the process's genuine exposures against the confidence and risk of each case class, and place deliberate review exactly there. Replace: convert the retired approvals into monitoring, post-action sampling, thresholds, and pattern surveillance, so the removed signature leaves behind an instrument instead of a void. The replace step is what separates streamlining from recklessness, and it is the step decree-style simplification programs always skip.
What Does Subtraction Look Like in a Real Process?
The oil and gas authorization for expenditure is the genre classic. At many operators, a routine well workover AFE collects signatures from operations, engineering, finance, land, joint-venture accounting, and several layers of line management, roughly a dozen approvals for work the field team has performed a hundred times. Trace the chain and every signature has an origin story, none of them current. Tested against real AFEs, most signatures show intervention rates near zero: the approval is ceremonial, but each one adds days, and the well produces nothing while the folder circulates. Redesigned, routine AFEs within engineering standards and budget parameters flow with post-approval sampling and variance monitoring, while genuinely consequential authorizations, unusual scope, environmental sensitivity, capital thresholds, receive fewer, deeper, better-informed reviews than the old chain ever provided. Cycle time collapses. Control, measured honestly as intervention where exposure lives, goes up.
A note on sequencing, because the politics are half the work. Do not open with a decree to simplify everything. Open with one process and publish its intervention-rate table, because evidence changes the conversation in a way argument never does. A signature holder can debate philosophy all day. It is much harder to defend a signature that the data shows intervened three times in four thousand authorizations, and once the first table circulates, other process owners start requesting their own.
Running the trace-and-test analysis, and designing the monitoring that replaces retired approvals, is core work inside our agentic AI consulting practice, and the intervention-rate table alone typically ends the debate.
