Adaptive process orchestration, with the objective as the specification.
Adaptive process orchestration is the category Forrester opened in 2025 and mapped in its first landscape in 2026: automation that uses AI agents and non-deterministic control flows, alongside traditional deterministic flows, to meet business objectives. Its centre of gravity, the business objective, is where our methodology has lived for years: objective-driven process design specifies every process by the outcome it must reach and the constraints it must respect, so the route can vary while the accountability cannot.
The term and the method
We call it objective-driven process design. Forrester describes the same frontier as adaptive process orchestration.
For twenty years the contract of automation was that the sequence is the specification: the diagram is the spec, and conformance means the path was walked as drawn. An AI agent breaks that contract, because it decides its next step within limits — if the sequence is the specification, an agent is unspecifiable. The category's name points at the resolution: processes defined by objectives, not only by routes.
For that to be more than a slogan, the objective must be a formal object the engine can check: outcomes expressed as business states, one criterion per outcome with a target, a tolerance, a direction and a declared source, constraints typed, risks attached. That is what our methodology formalizes — and it is why conformance here is consulted, not opined: deterministic flows, human steps and, at the declared frontier, agents can vary the route while the objective and its evidence hold.
The mapping
What the category describes, and where it lives in Dexon BPM
The three elements of the definition — objectives, deterministic flows, the non-deterministic frontier — with our coverage as it stands.
The objective as specification
The formal objective →
Stakeholders, outcomes as business states, one criterion per outcome with target, tolerance, direction and declared source, typed constraints and risks — a structured object, not a paragraph.
Objective-driven process design →
The discipline of deriving the process from its objective — documented, published and downloadable in the methodology.
The essay: the objective is the specification →
Why the sequence-as-spec contract breaks with agents, and the three questions to ask any vendor in this category.
Deterministic flows
Executable BPMN →
The deterministic backbone: modelled routes, states and transitions, executed on the engine with evidence per step — what everything auditable still runs on.
Clocks and calendars →
Deadlines computed against real work calendars and service windows; alerts act before the breach.
Evidence per case →
An unalterable event log per case: who or what took each step, on what data, under which rule — the same trail for humans and machines.
The non-deterministic frontier
The agent as an executor class →
The methodology defines the agent with bounded scope, escalation criteria and evidence per intervention — and the platform applies it: agents are registered with declared limits (allowed actions, confidence threshold, human approval gate, daily cap), an activity whose executor is an agent is handed to it automatically with the human gate kept at closure, and every decision is logged with its reason. Over MCP, Microsoft 365 Copilot, Claude or your own agents use the platform with each user’s permissions. Computer-use agents invoked from the process: available from December 2026.
AI subordinate to the objective →
Classification, prediction and grounded answers run as activities of the governed process, never outside it.
Cases whose route is not predefined →
Adaptive case management in production: tasks, participants and related data added while the case runs, with the same evidence as a modelled route.
Conformance, consulted →
Declared constraints are checked by the engine on demand — the property that makes variation safe to allow.
Bring a process whose route cannot be fully drawn
We will show you its objective specified formally, its deterministic backbone executing, and the activities where governed agents take part — with the evidence an auditor would ask for.