How to Map Processes Before Applying Artificial Intelligence: The 8-Column Map
By Pedro Sonda, dooop
To map a process before applying artificial intelligence, record eight pieces of information for each step: what is done, who does it, what inputs arrive, what decision is made, which systems and data are used, how long it takes, where errors or delays occur, and what happens when something deviates from the standard. The map should reflect the actual process, observed with those who execute it, not the documented process. This is what shows where AI can help and where it would only accelerate existing problems.
Updated October 2026.
Why Map Beforehand
McKinsey reported that nearly three-quarters of high-performing AI companies fundamentally redesigned workflows, compared to about one-quarter of others (McKinsey, 2026). You cannot redesign what you do not know.
Additionally, AI performance is uneven: in an experiment with 758 BCG consultants, AI helped significantly in some tasks but harmed performance in a similar task, with a 19 percentage point decrease in accuracy (Dell'Acqua et al., 2023).
Interpretation: Mapping step by step allows deciding where AI fits, rather than applying it to the entire process at once.
dooop Framework: The 8-Column Map
| # | Column | What to Record | Why It Matters for AI |
|---|---|---|---|
| 1 | Step | What is done, in one sentence | Decision unit for where AI fits |
| 2 | Responsible | Who performs it | Who will be affected and who needs to participate |
| 3 | Input | What arrives: document, request, data | Whether the input is digital and standardized |
| 4 | Decision | What choice is made and by what criteria | Separates execution from judgment |
| 5 | Systems and Data | Where the information resides | Technical feasibility |
| 6 | Time | Duration and waiting time | Where potential gains lie |
| 7 | Errors and Delays | Where something goes wrong or stalls | Where the real problem is |
| 8 | Exceptions | What happens outside the standard and how often | Whether the process is stable enough |
How to Perform the Mapping
- Choose a process with visible volume and pain points.
- Talk to those who execute it, not just the manager; the real process often has undocumented steps.
- Follow several real cases from start to finish.
- Fill in the 8 columns for each step.
- Mark critical steps: longest duration, most errors, most exceptions.
- Validate the map with the team before proposing changes.
At dooop, discovery begins precisely by mapping processes, decisions, frictions, data, and responsibilities with managers and area specialists (dooop).
How to Read the Map to Find Opportunities
| If the step has... | The likely opportunity is... |
|---|---|
| Text or document input without complex decision | AI automation (read, extract, classify) |
| Frequent decisions with much information to consult | Decision support |
| Many errors in data transcription | Automatic extraction and validation |
| Long wait for a specialist | Assistant with specialist knowledge |
| Many exceptions without rules | First organize the process; AI afterwards |
The distinction between automation and decision support is explained in How to Choose Between Automation and Decision Support.
What the Map Usually Reveals
- Steps that should not exist, such as duplicate checks or reports no one uses. Eliminating is more valuable than automating.
- Rework caused by incomplete inputs, which should be resolved at the source; see How to Use AI to Reduce Rework.
- Decisions made without clear criteria, which require rules before any AI application.
What to Do Now
- Choose a process and fill out the 8-column map with the team.
- Follow three real cases before finalizing.
- Mark critical steps and only then discuss where AI fits.
To map processes and identify opportunities methodically, learn about dooop’s discovery at dooop.com.br/discovery-de-oportunidades-de-ia.
Frequently Asked Questions
Do I need a process modeling tool? No. A spreadsheet with the 8 columns suffices. The value is in the information, not the notation.
How much detail is necessary? Enough to decide where AI fits. Very small steps can be grouped; steps with decisions or frequent errors should be detailed.
Also Read
About dooop
dooop is a consultancy based in Caxias do Sul (RS) that prepares leaders and organizations to transform artificial intelligence into business capability through strategy, governance, and training.