Process Intelligence Software

Every organisation has two versions of each of its processes. The one in the documentation, and the one that actually happens.

The documented version describes what should occur: the request arrives, it is checked, approved, fulfilled, and closed. The actual version includes the workaround for the system that has been broken since 2019, the approval someone grants informally over chat, the step that exists because a customer complained once in 2016, and the four routes through the process that nobody drew.

Process intelligence software reconstructs the second version from evidence. It reads the traces that systems leave behind and tells you what your organisation is genuinely doing, which is frequently not what anyone believed.

That is a valuable capability and an uncomfortable one, and both properties explain the category's history.

Mining the system and mining the person

Forrester defines process intelligence software as products providing analytics about an organisation's processes, both those operated by humans and those implemented in IT systems, based mainly but not exclusively on task and process mining, enabling decisions about how to improve process performance.

Two techniques sit underneath that, and the distinction matters.

Process mining reads event logs from enterprise systems. Every time a record is created, modified, approved, or closed, a system writes a timestamped entry. Reassemble those entries by case identifier and you have the actual path each instance took, with durations, loops, and deviations visible.

Task mining watches what people do on their desktops. Clicks, application switches, copy and paste, keystrokes. It captures the work that happens between systems, which is invisible to event logs and is frequently where the time goes.

Forrester's landscape research identified these converging, with most process mining vendors adding task mining and most task mining software extending to cover long-running transactional processes that used to be process mining's territory. Blending both produces a holistic view of process operations.

The convergence matters because the two see different failures. A process may look efficient in the event log and involve a person spending eleven minutes reconciling two screens between steps that the log records as consecutive.

The lineage

Forrester published The Process Intelligence Software Landscape, Q2 2023, authored by principal analyst Bernhard Schaffrik, mapping the market and identifying three things worth knowing: that process and task mining were converging, that time to value remained a primary challenge, and that automation software vendors had become dominant players through acquisitions, partnerships, and native development.

Its characterisation of the market's structure is unusual and accurate: it consists of very small but mature vendors alongside large and equally mature ones. Process mining has academic roots going back decades, so several of the specialists are technically deep and commercially small.

The Forrester Wave: Process Intelligence Software, Q3 2023 was the first scored evaluation, covering fourteen vendors against twenty eight criteria across current offering, strategy, and market presence.

ARIS, then part of Software AG, took the highest current offering score of all fourteen with maximum marks in twenty one of the twenty eight criteria, and Forrester's assessment was that it pioneered and still led in process intelligence.

iGrafx also placed as a Leader, credited for classic process improvement capability including simulation, mapping, benchmarking, and standards definition, with maximum scores in process simulation, modelling, and standard definition.

The Process Intelligence Software Landscape, Q2 2025 mapped twenty one vendors, and The Forrester Wave: Process Intelligence Software, Q3 2025 evaluated fifteen of them, sorted into Leaders, Strong Performers, and Contenders, with customer feedback replacing market presence as the third dimension.

Inside the Q3 2025 evaluation

Three Leaders were named.

ARIS retained Leader status, with Forrester positioning it for users in regulated industries focused on process control, compliance, and cost optimisation in pursuit of holistic business process management, and noting a roadmap including AI agents that identify and resolve bottlenecks.

iGrafx placed as a Leader with the third-highest current offering score and maximum marks in eight criteria including process compliance, process simulation, design, and pricing flexibility and transparency.

Among Strong Performers, Forrester's positioning statements are unusually specific about fit.

KYP.ai is placed for users challenging AI and automation investments and increasing process effectiveness.

SAP Signavio for customers with an SAP-centred application landscape looking to transform operations or rearchitect application landscapes.

Appian for clients with a data-driven mindset in industries where process differentiation is core to winning.

Apromore for financial services customers pursuing general process improvement or prioritising continuous process compliance.

IBM as tailored more to midsize organisations while also suiting large ones wanting to connect process insights to subsequent orchestration.

Read those together and the market is not one market. Regulated compliance, SAP landscape transformation, financial services conformance, automation investment justification, and orchestration integration are five different buying problems, and the vendors have specialised accordingly.

The audience is changing

Forrester's assessment of the current state is that process intelligence software is undergoing a significant change in its target audience and the business value it provides.

That is worth unpacking, because it describes a repositioning rather than a product improvement.

The original audience was process improvement specialists: continuous improvement teams, Six Sigma practitioners, business process management functions. The value was efficiency, and the output was a project to fix a bottleneck.

The emerging audience is automation and AI leadership, and the value is deciding where to deploy. KYP.ai's positioning around challenging AI and automation investments states it directly.

The logic is straightforward and it has become urgent. An organisation deciding where to apply agents needs to know which processes exist, how much volume runs through each, where the exceptions concentrate, how much human time each consumes, and what the variation looks like. Without that, automation targeting is guesswork informed by whoever complained loudest.

That makes process intelligence the diagnostic layer beneath the automation decision, and it explains why automation vendors became dominant players in the market. A company selling process automation benefits directly from a tool that identifies which processes to automate.

It also creates a conflict worth noticing. Process intelligence supplied by an automation vendor will find opportunities that the vendor's automation products address. That is not dishonest, and it is a reason to weight independence when the tool's primary use is investment justification.

What conformance actually reveals

The capability that generates the most discomfort is conformance checking: comparing the actual process against the designed one and quantifying the divergence.

Organisations running this for the first time consistently discover the same things.

The standard path is followed in a minority of cases. A process designed with one route typically executes with dozens of variants, most occurring a handful of times.

Rework is larger than anyone estimated. Steps repeat, cases loop back, and approvals get re-sought, and none of it appears in the process documentation.

The exceptions consume most of the effort. The eighty percent of cases that follow the standard path take a small share of the total time. The tail takes the rest.

And the workarounds are load-bearing. The informal step someone invented to get around a system limitation is now the only reason the process completes, and removing it without understanding it breaks things.

That last finding is the one that should shape how the output gets used. A deviation is not automatically a defect. Some deviations are people compensating for a badly designed process, and standardising them away without asking why they exist produces a compliant process that works less well.

This is the same distinction that matters when automating: encoding the process as it exists imports the accumulated dysfunction, and fixing it first requires the political capital to tell several departments their local optimisations are going away.

Time to value

Forrester named time to value as a primary challenge in this market, and identified the causes: process data analytics skills, cross-departmental collaboration along end-to-end processes, and software configuration.

The middle one is the real constraint.

An end-to-end process crosses functions by definition. Order to cash runs through sales, operations, logistics, and finance. Analysing it requires data from each, cooperation from each, and agreement on where the process starts and ends.

That agreement is harder than it sounds, because each function has a view of the process bounded by its own responsibility, and none of them owns the whole thing. A project that begins as a technical exercise becomes an organisational negotiation about scope and ownership, and it stalls there.

The data extraction is genuinely difficult too. Event logs need a case identifier, an activity name, and a timestamp, and enterprise systems do not always provide all three cleanly. Reconstructing a case identifier that spans three systems is engineering work before any analysis begins.

Organisations that get value quickly scope narrowly: one process, one clear owner, one question worth answering. Those that attempt an enterprise-wide process model first spend a year building something nobody uses.

Where this sits now

Process intelligence has moved from a specialist improvement tool to a prerequisite for something larger.

Forrester maintains adjacent categories for digital process automation and adaptive process orchestration, both of which assume you know what your processes are. Its agentic coverage assumes the same. An organisation deploying agents into business processes without a current, evidence-based picture of those processes is automating a description rather than a reality.

The uncomfortable version of that observation is that process intelligence is most valuable to organisations that will act on what it finds, and most organisations discover that acting requires changing something a function currently controls.

The tooling has never been the constraint. Reconstructing what actually happens has been technically feasible for years and is now genuinely good. What remains scarce is the willingness to look at the result and change the thing it points at, which is the same finding that appears in every analytical category and lands harder here because the evidence is unusually difficult to dispute.

Forrester does not endorse any vendor named here, and tier placement should not be read as a recommendation to buy.