The Target BPMN: Seeing the AI-Transformed Process Side-by-Side With the Current One

The target BPMN is a second version of your process diagram, generated automatically once the AI transformation plan is approved. The shapes are the same as your current-state BPMN: same lanes, same gateways, same flow structure: except that every task the plan recommends automating is rendered as an AI-task node rather than a regular task node. The AI-task nodes are visually distinct (different fill, often a small AI badge in the corner) so the difference between current and target reads at a glance.

Every AI-task node carries metadata that ordinary task nodes do not: the pattern name from the knowledge base it matched, the maturity level chosen (Companion / Automation / Agent), the classifier's confidence score, and the projected monthly savings. Click an AI-task and you see exactly what it does, why this pattern matched, and what financial change it represents in the optimised process. The current-state BPMN tells you how the process runs today; the target BPMN tells you how it would run after transformation, with the receipts attached.

The standard alternative to a target BPMN is a numbered list: "We propose to (1) automate the credit-check task, (2) eliminate the manual approval, (3) integrate the customer-data lookup with the CRM". The list is fine for a memo. It fails in a stakeholder meeting because no one in the room can hold three changes in their head while also evaluating whether the changes preserve the process correctly. The list also fails to make the transformation feel real, it is a description of work that does not yet exist.

Side-by-side BPMNs solve both problems. The original on the left, the target on the right; every change is a node that visibly differs. A stakeholder can scan from left to right, point at the AI-task node that replaces a familiar manual task, and ask "is this really safe to automate?": a question they could not ask if all they had was a numbered list. The diagram makes the proposal interrogatable in a way prose cannot.

Once the transformation plan exists, the generator walks every step in the plan and finds the corresponding task node in the original BPMN by ID. For each match, it clones the node, changes the type from 'task' to 'ai-task', and writes the pattern name, maturity level, confidence score, and projected monthly savings into the node's data. Tasks not referenced by any step in the plan stay as 'task'. Non-task nodes: events, gateways, swimlanes: are copied through unchanged.

Edges follow a similar pattern. Every edge in the original is preserved; edges that connect to an ai-task node get an animated style so the transformed flow segments visually pulse on the canvas. The animation is a deliberate UX choice: it draws attention to the parts of the process that have changed without requiring the reader to mentally diff two static diagrams. After 10 seconds the animation fades to a static highlight; this is enough to orient the eye but not so much it becomes annoying during a long review.

The target BPMN earns its keep in three specific moments. First, the kickoff: open the original on the left, the target on the right, and walk the room through the diagram from start to end. Stop at every AI-task node and read out the pattern, the maturity level, and the projected saving. The narrative writes itself. Second, the objection round: when a stakeholder points at an automation and says "I do not believe this will work", click the AI-task node and read the classifier's reasoning aloud. Most objections evaporate when the basis of the recommendation is on the table.

1. Show both diagrams side by side from the start. Resist the temptation to walk through the changes verbally first; let the diagrams do the framing.
2. Read out the pattern name and maturity level on every AI-task. The pattern names are designed to be self-explanatory; reading them is enough to validate the proposal at a high level.
3. Click into nodes the stakeholders question. The classifier's reasoning is the on-the-spot answer to "why did the AI suggest this".
4. End with the cost dashboard's annual savings figure overlaid on the target diagram. The picture and the number reinforce each other.
5. Export the side-by-side as PDF after the meeting. Stakeholders who could not attend get the same artefact and need no extra explanation.

The target BPMN is a model of what the process will look like after the recommended changes ship. It is not a promise that the changes will ship on time, that the tools will be procured at the projected price, or that the actual savings will match the projection within the percentage the report claims. Every transformation project is also a change-management project; the diagram captures the steady-state outcome but says nothing about the friction of getting there. Treat it as the destination, not the route.

The target also does not capture every kind of optimisation. Eliminate (a removed task) shows up as the absence of a node, but the target does not draw a comparison ghost: the reader has to remember the original was there. Simplify (a smaller form, a shorter procedure) shows as the same task with revised KPIs but the same shape. Standardize (collapsing variants) similarly does not draw the variants that were merged. The target excels at showing intelligize and integrate; it under-shows eliminate, simplify, and standardize. For those, lean on the ROI report waterfall that does decompose every change.