Business Systems · Operator Playbook

How to Map a Business Process Before You Automate It

Automating a messy process only makes the mess faster: map how the work really flows with the people who do it, remove and fix what is broken, then automate the steps that are stable.

Worktables 14 min read

Three colleagues map a process on a whiteboard of sticky notes arranged in swimlanes, one holding a coffee cup, next to the headline “Map it first, then automate.” Worktables, monday.com Silver Partner.

Map a process before you automate it by walking through how the work really happens today, with the people who do it: where it starts and ends, every step and handoff, who owns each one, what information moves between them and where work waits. Then fix the broken steps, and automate only what is stable.

This article is for operations leads, monday.com admins and anyone about to automate a workflow. It is the method we use at the start of most client projects: how to scope the process, run the mapping session, capture the right details, read the map for waste, and turn it into a list of what to automate, what to fix and what to leave alone.

Why map before you automate

Automation makes a process faster. It does not make it better. If a request bounces between three people before anyone owns it, an automation will bounce it between three people in seconds instead of hours, and the confusion arrives sooner.

A common pattern when a first automation attempt fails: a team automated what they thought the process was, usually the version in the onboarding doc or in a manager's head. The real process had extra approvals, side spreadsheets and a person who quietly fixed data every Friday. The automations were built for a process that did not exist, so people worked around them.

Mapping first gives you three things:

  • A shared picture. Everyone involved agrees on how the work actually flows today, including the parts nobody wrote down.

  • A list of problems to fix before automating. Duplicate entry, unclear ownership, approvals nobody needs.

  • A clear scope for automation. The steps that are stable, frequent and rule-based, which is exactly what automations are good at.

The American Society for Quality (ASQ) describes the flowchart as one of the seven basic quality tools and notes that, once drawn, it helps identify and eliminate non-value-added steps. That is the job here: see the waste before you speed it up.

Step 1: Pick one process and set its boundaries

Start with one process, not a department. "Sales" is not a process. "From signed quote to first invoice" is.

Good first candidates share three traits: they run often, they cross at least two teams, and people complain about them. A process that runs twice a year is rarely worth automating. A process that stays inside one person's head is rarely worth mapping as a group.

Then write down the boundaries in one line each:

  • Trigger: what starts the process. "A customer signs the quote."

  • End: what finishes it. "The first invoice is sent."

  • Out of scope: what you are deliberately not mapping. "Renewals and change orders."

ASQ's guidance on SIPOC diagrams makes the same point: define the start and end points so that everyone understands the limits of the analysis. Most mapping sessions that run long have skipped this step.

Step 2: Do a quick SIPOC first

Before drawing every step, spend 20 minutes on a SIPOC: Suppliers, Inputs, Process, Outputs, Customers. ASQ recommends it before constructing a flowchart, because it gathers the basics of the process in one view.

ElementQuestionExample (quote to invoice)
SuppliersWho or what provides what the process needs?Sales, the customer, the pricing sheet
InputsWhat information or material goes in?Signed quote, billing details, PO number
ProcessWhat are the five to seven core steps?Hand off, kick off, deliver, approve, invoice
OutputsWhat does the process produce?Delivered project, approved timesheet, invoice
CustomersWho uses the outputs?The client, finance, the account manager

Keep the Process column high level: ASQ suggests five to seven core activities. The detail comes in the next step. The SIPOC's real value is in the Inputs column, because missing or late inputs are behind most of the delays you will find later.

Step 3: Map the current state with the people who do the work

This is the core of the method. Book 90 minutes with one person from each team that touches the process. Not their managers: the people who actually click the buttons, send the emails and chase the approvals.

Run the session like this:

  1. Walk one real, recent case. Pick an actual order, ticket or request from last week and follow it from trigger to end. Real cases surface the exceptions that a generic walkthrough hides.

  2. Write each step as verb plus object. "Create project board," "Send kickoff email," "Approve timesheet." One action per sticky note or box.

  3. Put each step in its owner's lane. Use one horizontal lane per role or team, often called a swimlane map. Every time the work crosses a lane, that is a handoff.

  4. Mark decisions. Where the work splits ("Is the PO number present?"), draw the decision and both paths.

  5. Ask "then what?" until you reach the end. And ask "what happens when that doesn't work?" at least once per lane.

  6. Capture the side systems. Every spreadsheet, inbox, chat thread or notebook that holds part of the process goes on the map.

Map the process as it is, not as it should be. When someone says "we're supposed to," write down what they actually do and note the gap. The Lean Enterprise Institute, describing value stream mapping, notes that teams usually build the current-state map first and the future-state map after. Skipping straight to the future state is the most common shortcut, and the most expensive one.

Step 4: Capture the details that matter for automation

A map that only shows boxes and arrows is not enough to automate from. For each step, add a few facts. You do not need all of them for every step; the first four matter most.

DetailWhat to write downWhy it matters
OwnerOne role, not the teamAn automation needs someone to assign to and notify
TriggerWhat tells the owner to startThis becomes the automation trigger
SystemWhere the step happensShows duplicate entry and integration needs
Data in and outWhich fields are read and writtenDefines the columns you need
VolumeHow many times per week or monthTells you whether automating is worth it
Wait timeHow long work sits before this stepUsually larger than the work itself
Touch timeHow long the step takes when someone works on itShows where people spend effort
ExceptionsHow often the step goes off the normal pathHigh exception rates are poor automation candidates

The Lean Enterprise Institute lists cycle time and lead time among the data recorded in each process box of a value stream map. You do not need a stopwatch. Rough numbers from the people in the room ("about two days," "maybe 30 a week") are enough to see where the time goes.

Step 5: Read the map for waste and handoffs

With the map on the wall, look for these patterns. Each one is a problem to fix before you automate anything around it.

  • Handoffs with no clear trigger. The work moves lanes, but the next owner only finds out when someone chases them. This is one of the most common sources of delay.

  • Long waits. Steps where work sits for days while the actual task takes minutes. The wait, not the work, is the problem.

  • Duplicate entry. The same data typed into two systems, or into a board and a spreadsheet. Every copy is a chance for the two to disagree.

  • Loops. Work that goes back to an earlier step, usually because an input was missing. Fix the input at the source instead of automating the loop.

  • Approvals nobody uses. Approvals that are always granted, or granted by someone without the information to refuse. Ask what would happen if they disappeared.

  • One-person dependencies. A step only one person knows how to do. Document it before you automate around it.

  • Side systems. Spreadsheets and inboxes that exist because the main system does not hold something. These are often what a well-built board should replace.

Circle each problem on the map and give it a short name. You will use the list in the next step.

Step 6: Decide what to fix, automate or leave alone

Now sort every step and every circled problem into one of four buckets.

BucketWhen it appliesWhat to do
RemoveThe step adds no value: an unused approval, a duplicate entryDelete it before building anything
FixThe step is needed but broken: unclear owner, missing input, loopsRedesign it, then run it manually for a few weeks
AutomateThe step is frequent, stable, rule-based and has a clear triggerBuild it as an automation or integration
Keep manualThe step needs judgment, happens rarely or changes oftenLeave it with a person, or consider AI later

The order matters: remove, then fix, then automate. Automating a step you later remove is wasted work, and automating a broken step locks the problem in.

For the steps that need judgment, such as reading a request and deciding its category, a plain automation is not enough. Our article on monday.com AI agents vs. automations explains when an AI block or an agent is the better fit.

Step 7: Draw the future state and build in small steps

With the buckets sorted, draw the future-state map: the same lanes, fewer steps, clear triggers at every handoff, and the automated steps marked. This is the blueprint you build from.

On monday.com, the map translates fairly directly:

  • Lanes become owners. Each lane is a People column value or a team, so every item has one owner at each stage.

  • Steps become stages. The main steps become Status labels or groups, in order.

  • Handoffs become automations. "When status changes to Approved, assign Finance and notify them" replaces the email nobody answered.

  • Data in and out becomes columns. The fields you captured in Step 4 become the board's columns, created once and filled in one place.

  • Side systems become boards or connections. A spreadsheet that tracked something the process needs usually becomes a board, linked with Connect Boards.

Build one handoff at a time. Run each new automation for a week, check that people trust it, then add the next. A process that changes all at once gives nobody a chance to notice what broke.

A worked example (hypothetical)

The company and numbers below are hypothetical, chosen to show the method. They do not describe a Worktables client.

A 60-person agency maps its "signed quote to first invoice" process. The session includes one person each from sales, delivery and finance.

What the current-state map showed. Eleven steps across three lanes. Sales emailed delivery when a deal closed, but delivery only checked that inbox twice a week. Delivery re-typed the client's details into a project board. Finance learned a project had started only when the first timesheet arrived, then asked sales for the PO number, which was in the original quote PDF. From signature to first invoice took about 24 days. The actual work in those steps added up to about four hours.

What they sorted.

  • Remove: re-typing client details (they already lived in the CRM board), and a manager approval on every kickoff that had never been refused.

  • Fix: the PO number became a required field before a deal can move to Won.

  • Automate: when a deal moves to Won, create the project item with the client data, assign the delivery lead and notify finance with the PO number.

  • Keep manual: the kickoff call and the first timesheet review.

The future state. Seven steps instead of eleven, two handoffs automated, and no step waiting on an inbox. The team would then measure the new signature-to-invoice time for a month before automating anything else.

Facilitator tips for the mapping session

The facilitator's job is to keep the map honest and the session on time. A few habits help:

  • Stand at the map, not at the table. Whoever holds the marker controls the pace. Write what people say, in their words, and ask them to correct you.

  • Park the solutions. People will start fixing the process halfway through. Write each idea on a separate "later" list and go back to the current state.

  • Ask for numbers, accept rough ones. "How many a week?" and "How long does it usually wait?" are enough. Precision can come later, for the steps you decide to automate.

  • Watch for silence. When a team goes quiet, ask them directly what happens in their lane. The quiet lanes often hide the workarounds.

  • Remote sessions need more structure. Share one screen, keep cameras on and walk the case in short turns, one lane at a time.

  • End with the problem list. Before people leave, read out every circled problem and confirm nobody disagrees with how it was described.

Self-check: is this process ready to automate?

Before you build the first automation, the process should pass every line below:

  • The trigger and the end are written down and agreed.

  • Every step has one owner, by role.

  • Every handoff has a clear trigger that does not depend on someone remembering.

  • Steps that add no value have been removed.

  • Broken steps have been fixed and run manually for at least a few weeks.

  • The data each step needs is captured once, in one place.

  • The steps you plan to automate happen often and follow the same rule every time.

  • Someone owns the map and will update it when the process changes.

If a line fails, that is the next thing to work on. Automating around it will cost more later.

Common mistakes

  • Mapping the official process. If the map matches the onboarding doc exactly, you have not found the real process yet.

  • Leaving out the people who do the work. Managers know how the process should run. The people doing it know how it does.

  • Mapping a whole department. Scope one process with a clear trigger and end.

  • Too much detail too early. Start with the SIPOC and the main steps. Add detail where the problems are.

  • Skipping the wait times. Without them, you optimize the steps that take minutes and miss the queues that take days.

  • Automating before fixing. A broken handoff that is automated is still broken, only faster.

  • Never updating the map. Once the new process runs, update the map. It becomes the documentation new hires read.

What to do next

Pick one process that crosses two teams and annoys both of them. Book 90 minutes, bring one recent real case, and map it as it is. You will likely find at least one step to remove before you build anything.

You can also see how we run implementations through our monday.com managed services.

Frequently asked questions

How long does it take to map a business process?

For one process with a clear trigger and end, plan on about two hours: 20 minutes for the SIPOC and around 90 minutes for the current-state session. Reading the map and sorting the steps usually takes another hour, often on a different day. If the session runs much longer, the scope is probably too wide. Split it into two processes and map them separately.

Who should be in the mapping session?

One person from each team that touches the process, chosen from the people who do the work every day, plus one facilitator who keeps time and asks "then what?" Managers can review the map afterwards. Keep the group small, around four to six people, so everyone talks. If one team cannot attend, interview that person separately and add their lane before you read the map.

What tool should we use to draw the map?

Whatever lets people move steps around quickly. Sticky notes on a wall or a whiteboard work well in person, and any online whiteboard works remotely. The tool matters much less than mapping a real case with the right people. Once the future state is agreed, document it where the team will see it, such as a monday.com doc linked to the board that runs the process.

What is the difference between a process map and a value stream map?

A process map, such as a flowchart or swimlane map, shows the steps, decisions and owners of a process. A value stream map, a method from the Toyota Production System, follows material and information flow from order to delivery and records time data at each step to expose waste. For most office workflows, a swimlane map with wait times captures the parts of value stream mapping you need.

Which processes should we map first?

Start with a process that runs often, crosses at least two teams and causes visible frustration, such as quote to invoice, request intake, employee onboarding or purchase approvals. Frequent processes give you a quick return on any fix. Cross-team processes have the handoffs where most delays hide. Avoid starting with a rare or highly variable process; it is hard to map and rarely worth automating first.

How detailed should the map be?

Detailed enough that someone new could follow one case from start to end, and no more. Each box is one action by one owner. Add the extra details, such as volume, wait time and exceptions, only for the steps where you suspect problems or plan to automate. If the map needs a second wall, you are probably mapping several processes at once.

Should we map the process we have or the one we want?

The one you have, first. The current-state map shows the real handoffs, workarounds and side systems, which is where the problems are. If you start from the ideal process, you will design around problems you have not seen yet. Draw the future state only after the current state is agreed and the steps are sorted into remove, fix, automate and keep manual.

What should we do with the map after automating?

Update it to match the new process and keep it where the team works. A current map becomes your training material for new hires and the first thing to check when something breaks. Review it when the process changes, or at least every six months, and note which steps are automated so people know which parts run on their own.

Sources

monday.com Silver Solution Partner

Want a second opinion on a process before you automate it?

Take the two-minute readiness check, or book a free 30-minute working call and bring one process you want to map.