Adaptive Process Orchestration Software

In the space of a few months, two analyst firms named what is recognisably the same market and gave it two different names. Forrester called it adaptive process orchestration. Gartner called it business orchestration and automation technologies, which the industry has already shortened to BOAT.

When that happens, it usually means something real is forming and nobody has agreed yet what the important part of it is. Comparing the two definitions is the fastest way to see what each firm thinks that important part is, and Forrester's version turns on a single phrase that most coverage skips past.

The definition, taken apart

Forrester defines the category as an automation platform that uses AI agents and, in its own words, "nondeterministic control flows, in addition to traditional deterministic control flows", in order to hit business goals, handle complex tasks, and make decisions without a person authorising each one.

Control flow is just the order in which steps happen. Deterministic control flow means that order is fixed in advance: if the claim value exceeds this threshold, route it there, then do this, then finish. You can draw it. You can test every path. Given the same inputs it does the same thing every time, which is exactly why regulated industries built their operations on it.

Nondeterministic control flow means the sequence is decided at runtime by something that reasons about the situation. Nobody drew the path. There may not be a fixed number of steps. The same input can produce different routes on different days, because the agent evaluates context rather than following a diagram.

Those two things are, on their face, incompatible operating philosophies. One is auditable by construction, the other is not.

Which makes the phrase "in addition to" the load-bearing part of the whole definition. Forrester is not describing a market where agents replace workflow engines. It is describing platforms that run both in the same process, with the deterministic parts holding the compliance-critical spine and the agentic parts handling the branches that were previously too messy to automate at all.

That is a much less exciting claim than most vendor marketing in this space, and a considerably more plausible one.

Why the previous generation of tools fell short

Forrester's argument for creating the category is blunt. Deterministic workflow engines, RPA bots, and digital process automation tools are not powerful enough on their own to deliver autonomous operations, so getting there requires several automation technologies from different domains combined with a process mindset.

Read that as an admission about the last decade of automation spend. RPA automated the tasks that could be described precisely enough to script, and enterprises discovered that this was a smaller share of their work than the business cases assumed. Everything left over involved judgement, exceptions, unstructured inputs, or all three.

Agents can handle that residue. What they cannot do on their own is prove afterwards that they handled it correctly, which is why the category is defined as a combination rather than a replacement.

Bernhard Schaffrik, the Forrester analyst who established this coverage, has framed autonomous operations as the north star for most enterprises rather than as a shipping product. The distinction matters when a vendor tells you the destination has arrived.

What The Adaptive Process Orchestration Software Landscape, Q2 2026 covers

The first evaluative document in the category published in Q2 2026 as The Adaptive Process Orchestration Software Landscape, Q2 2026, authored by Dr. Bernhard Schaffrik and profiling 35 vendors.

A Landscape is not a Wave. It maps who is in a market and how they differ by size, type, and focus. It does not score them, and it does not rank them, so anyone citing this document as evidence that a vendor is a Leader has misread what they are holding. The Wave evaluation is scheduled for later in 2026.

Confirmed participants include Camunda, Flowable, Newgen Software, and Decisions, the last of which recently merged with ProcessMaker. Most of the 35 will be familiar from adjacent categories, particularly digital process automation, where several of these vendors were evaluated in The Forrester Wave: Digital Process Automation Software, Q3 2025.

That overlap is the point rather than a coincidence. This is not a market of new entrants. It is the existing process automation vendor set repositioning around agents, and the Landscape reflects an industry consolidating overlapping tools into a single orchestration backbone.

The themes Forrester flags in the report are worth noting precisely because they are unglamorous: governance, auditability, hybrid execution models, and human-in-the-loop decisioning. Not agent autonomy. Not reasoning quality. The constraints, not the capability.

Two names, two emphases

Set the definitions side by side and the difference in emphasis is clear.

Gartner's BOAT describes a consolidated platform delivering enterprise process automation through orchestration, enterprise connectivity, low-code development, and agentic automation. It is a definition about consolidation. The problem it identifies is tool sprawl, and the answer is one platform doing four things.

Forrester's APO is a definition about control. It says almost nothing about connectivity or low-code, and stakes the category on the coexistence of two fundamentally different execution models in one process.

Neither is wrong. If your pain is fifteen automation tools with overlapping licences, the Gartner framing describes your problem. If your pain is that you cannot let an agent near a regulated process without losing your audit trail, the Forrester framing does. Most enterprises have both problems, which is presumably why both categories found an audience.

Where this leaves you until the Wave publishes

There are no tiers yet, so there is nothing to shortlist from, and that is genuinely useful information. Any vendor currently claiming analyst validation in this category is pointing at a report that does not score anyone.

What the Landscape does support is a longlist, and the definition supports a test you can apply yourself. Ask a vendor to show you a single process where deterministic and agentic steps run together, then ask what the audit record looks like for the agentic portion. Ask what happens when the agent picks a path nobody anticipated, and who finds out.

The vendors that have genuinely built for this will have an answer involving constraints, logging, and escalation. The ones that have relabelled an existing workflow product will talk about the model.

Forrester does not endorse any vendor named here, and inclusion in a Landscape report is not a rating or a recommendation to buy.