By James Proctor, Co-Founder and Managing Director, The Inteq Group
AI-enabled processes adapt to changing conditions at three distinct speeds, set by where the change lives. Parameter changes, adjusting decision criteria, thresholds, and routing conditions inside the existing design, take effect in days and sometimes the same day. Structural changes, adding a new decision point or handling path, take weeks. System changes, new integrations or platforms, take what they have always taken. The design objective that follows is plain: push as much of the change you can anticipate into the parameter layer, and size the governance of that layer to move at parameter speed. Organizations that do this respond to market shifts as an editing exercise. Organizations that do not respond to the same shifts as a project.
“Design decides whether change is an edit or a project.”
The parameter layer holds everything the original design anticipated might move: the thresholds, the criteria weights, the conditions that trigger escalation. Changing it changes behavior without changing structure, which is why it is fast and why its risks are boundable. The structural layer holds the shape of the process itself, and touching it means design work, testing, and retraining. The system layer holds the plumbing. Most enterprises experience all change at system speed for a revealing reason: their processes were never designed with a parameter layer at all, so even trivial behavioral adjustments require modifying structure or code. Slowness, in other words, is not a property of change. It is a property of designs that gave change nowhere cheap to live.
This makes anticipation a first-class design activity. The question what is likely to change here, demand, risk appetite, regulatory limits, supply conditions, belongs in every process design review, because whatever the answer names should be built as a parameter, deliberately, on day one. A parameter layer is not discovered later. It is the residue of having asked the question early.
Fast is not the same as ungoverned, and this is where I will say the impolite thing, because it is true in nearly every enterprise I walk into: your change management process is where change goes to die. The process for changing processes is routinely the most rigid one the organization owns, a single advisory-board queue that prices a threshold adjustment and a platform migration at the same bureaucratic rate.
The organization can approve an exception in an hour but needs a quarter to approve an improvement, which is why everything becomes an exception. The alternative is layered governance: pre-authorized ranges within which named owners adjust parameters on their own signature, with every change versioned, attributed, and monitored for effect; structural changes through a faster, design-level review; and the heavyweight process reserved for the system layer it was built for. Speed comes from matching the governance to the layer, not from removing the governance.
Pharmaceutical distribution runs this test constantly, because drug shortages arrive without appointment. When a shortage is declared, the operative questions are immediate: which therapeutic substitutions are acceptable, how remaining inventory is allocated across customer classes, what ordering limits prevent hoarding. In the traditional operation, the answers travel by emergency meeting and memo, applied manually and inconsistently across the network for weeks. In a parameter-layer design, substitution rules, allocation criteria, and order limits are the parameters: the clinical and commercial owners adjust them under their pre-authorized ranges, the network behaves differently that day, uniformly, and every adjustment is on the record with its author and rationale. The shortage is not less real. The response is simply operating at the speed of the decision instead of the speed of the meeting cycle.
Identifying which anticipated changes belong in the parameter layer, and designing the layered governance that moves safely at that speed, is core work in our agentic AI consulting engagements, and it is frequently the difference between owning a capability and owning a backlog.
Related White Paper
Read the foundational white paper examining how agent-enabled processes adapt in real time, absorb volatility, and remain fully auditable.
Resilience by Design: Why Agent-Enabled Processes Adjust in Real Time and Stay Auditable →