Process Discovery

Process discovery for transformation and consulting teams

Process discovery is the practice of capturing how work currently flows through an organisation - the as-is process - as the input to improvement, automation, migration, or audit. This page covers the discovery engagement as a BPMN process map, the data sources discovery teams pull from, and how AI is collapsing the timeline.

By Jack Finnegan · Updated 21 May 2026

Run process discovery faster Process mining software

What it is

What process discovery actually is

Process discovery is the practice of capturing how work actually flows through an organisation. The output is typically an as-is BPMN process map (or a set of them), with variants and exceptions documented, pain points quantified, and a handover to whatever programme is going to act on the discovery: improvement, automation, modernization, M&A integration, or audit. The discipline sits at the front end of most transformation engagements and inside the early phase of most consulting engagements.

Modern discovery combines four data sources: structured interviews with the people who run the process, observation (often video walkthroughs or shadowing), document review (SOPs, runbooks, training materials), and increasingly event-log mining from the systems the process touches. Each source corrects the others - what people say they do (interviews), what they actually do (observation), what the manual says they do (documents), what the systems record they did (mining).

The problem today

Process discovery still takes weeks because the analyst draws diagrams by hand

A senior analyst doing as-is process mapping in legacy tooling spends three to five days on a single end-to-end process. Multiply by 50 processes in a typical mid-size transformation and discovery becomes a six-month programme before any improvement work begins. The team gets pressure to skip variants, accept the happy path as good enough, and move on - which is exactly the missing detail that makes the subsequent improvement work miss the mark.

AI changes the economics. A generative model that understands BPMN semantics can turn a transcript, a SOP document, a screenshot, or a meeting recording into a valid first-draft BPMN diagram in seconds. The analyst's time shifts from drag-and-drop to interpretation, validation, and variant capture - the parts where the human judgement actually adds value.

Four pillars

Four pillars of process discovery

Multiple data sources

Interviews, observation, document review, and event-log mining. Each corrects the others. Single-source discovery is biased discovery.

Variants and exceptions

The happy path is rarely the interesting one. Discovery quality is measured by how completely it captures the variants and exceptions the formal documentation hides.

Owner validation

Maps must be validated by the people who actually run the process, not just signed off by their manager. Owner validation is where most discoveries surface their final missing variants.

Pain-point quantification

Cycle time, rework rate, cost per execution, SLA breach rate. Quantification is what turns discovery into a business case the improvement programme can build on.

Process Map

A discovery engagement as a process map

The canonical discovery flow - identify, collect, document, surface variants, validate, quantify, hand off - with the validation loop that catches missed variants before the hand-off.

Open in editor

A process discovery engagement as a process map

A canonical process discovery engagement rendered as a BPMN 2.0 process. Identify the candidate, collect data, document the as-is, surface variants and exceptions, validate with the owner, quantify pain points, and hand off to improvement.

Identify the candidate process - via referral, prioritisation, or process mining heat-map.
Collect data through interviews, observation, document review, and system event logs.
Document the as-is process map covering the happy path plus the major variants.
Surface exceptions, manual workarounds, and shadow processes the formal documentation hides.
Validate the map with the process owner and key participants.
Quantify pain points - cycle time, rework rate, cost per execution - to justify the improvement business case.
Hand off to the improvement, automation, or migration programme.

What this diagram shows: The engagement starts once a candidate process has been selected. Data collection pulls from interviews, observation, document review, and event-log mining in parallel. As-is documentation produces the first-draft map; the variants and exceptions task adds the workarounds and edge cases. Validation with the process owner is where misses surface - the gateway routes back to documentation if revisions are needed. Once approved, pain-point quantification produces the numbers that justify the next-stage programme, and the discovery hands off cleanly.

FAQ

Frequently asked questions

What is process discovery?

Process discovery is the practice of capturing how work actually flows through an organisation as the input to improvement, automation, modernization, M&A integration, or audit. The output is typically an as-is BPMN process map with variants, exceptions, and pain points documented.

What is the difference between process discovery and process mining?

Process mining reverse-engineers process models from system event logs - it tells you what the systems recorded happened. Process discovery is the broader discipline that combines mining with interviews, observation, and document review. Mining is an input to discovery, not a substitute - it cannot capture the manual steps that happen outside systems, the workarounds that bypass the system, or the human judgements that shape the process.

How long does process discovery take?

Historically, three to five days per end-to-end process for a senior analyst working in legacy diagramming tools. With AI-assisted first drafts (BA Copilot and similar), the analyst time per process drops to a few hours - the time savings mostly come from removing the drag-and-drop diagramming, freeing the analyst to spend more time on variant capture and validation, which is where the value is.

When should you do process discovery?

At the front end of: improvement programmes (you need to know the as-is before you can improve it), automation programmes (RPA and workflow engines need detailed as-is process maps), modernization programmes (migrating to a new system without discovering the current functionality is a recipe for missed requirements), M&A integration (acquiree processes are usually poorly documented), and major audits (auditors expect to see the as-is documented before testing controls).

How does BA Copilot speed up process discovery?

BA Copilot turns plain-English descriptions, meeting transcripts, SOP documents, and screenshots into valid BPMN 2.0 first-draft diagrams in seconds. The analyst then refines - validating with the owner, capturing variants, quantifying pain points. The bottleneck was always the drag-and-drop diagramming; once that is removed, discovery throughput jumps without losing analyst judgement.

Jack Finnegan, Founder & CEO, BA Copilot

From the founder

14 Years in BPMN

I'm Jack Finnegan. I've spent fourteen years working hands-on with BPMN, as an analyst, an engineer, and a product director, where I felt every sharp edge of legacy business process platforms.

BA Copilot is the platform I wanted on every one of these projects: AI-first process management, which treats BPMN as a first-class output rather than an export afterthought.

Jack Finnegan on LinkedIn

Speed up your next discovery

Open the discovery engagement template, feed your interview transcripts or SOPs into BA Copilot, and produce first-draft BPMN diagrams in seconds - then refine with the analyst time you just freed up.