Business Process Expert·7 min read
The broken golf cart problem
A plant manager at a small manufacturer put it perfectly: implementing AI on broken processes is like strapping a rocket engine to a broken golf cart. The engine is powerful. The cart still does not work. You just crash faster.
This pattern repeats across industries. Leadership gets excited about AI. Teams rush to implement. Nobody stops to ask: do we actually understand how our processes work today?
"I see companies buying AI tools before they can describe their own processes. The AI vendor asks for process data. The company realizes they have none. Six months and a lot of money later, they start mapping. Map first."
What AI needs from you
AI tools - whether they generate documents, route decisions, or predict outcomes - need three things your organization probably has not documented:
- 1.Structured processes - AI cannot optimize a process that exists only in people's heads. It needs a defined flow: steps, decisions, roles, exceptions.
- 2.Clear decision logic - when a gateway says "approved or rejected," AI needs to know the criteria. If the criteria are "ask Janet, she knows," AI cannot help.
- 3.Data about current performance - cycle time, error rate, throughput. Without a baseline, you cannot measure whether AI improved anything.
The three-step sequence
Step 1: Map
Document your current processes in BPMN. Include the workarounds, the manual steps, the exceptions. This is your as-is state.
Step 2: Fix
Before adding AI, remove the waste. Eliminate unnecessary approvals, reduce handoffs, fix broken exception paths. Design the to-be process.
Step 3: Augment with AI
Now add AI to the clean process. AI classifies incoming requests. AI suggests approval decisions. AI generates documents. The process is sound; AI makes it faster.
Where AI actually helps in process work
- -Document classification - AI sorts incoming documents to the right process path (invoice, complaint, application).
- -Decision support - AI recommends approval/rejection based on rules, human makes the final call.
- -Process mining - AI discovers how processes actually run from system event logs.
- -Diagram generation - describe a process in text, AI generates the BPMN diagram.
All of these require documented, structured processes as input. AI is the engine. Your process maps are the road.
Related guides
Keep learning
Frequently asked questions
Can AI map processes for me?▼
AI can generate first drafts from text descriptions. But it cannot interview your stakeholders, understand organizational context, or validate that the map matches reality. AI assists; you lead.
What if our processes are too broken to map?▼
If a process is chaotic, mapping it reveals the chaos - which is the first step to fixing it. You do not need a perfect process to create a useful map. Map reality, then improve it.
How do I convince leadership to map before AI?▼
Ask one question: can you describe your top 5 processes in detail? If they cannot, AI has nothing to work with. The mapping exercise typically takes weeks. An AI project on undocumented processes typically takes months of rework.