Work-Intelligence Method: map the work before the tool

Most automation projects start with a tool. Someone saw a demo, a vendor promised hours back, and now there is a subscription looking for a problem.

I work the other way round. Not tool first, but work first. Before anyone talks about AI, software or integrations, I want to see what actually happens in the business on a normal Tuesday.

I call this the Work-Intelligence Method. It is a simple diagnostic you can run in about 90 minutes. One map, one structured conversation, one scoring rubric. The output is one page, not a deck.

Why start with the work

Owners of growing businesses rarely lie awake thinking about language models. They think about the order that came in late and has to reach the team by morning. The invoice that sits on someone's desk for three days. The customer who called twice because nobody followed up.

The problem is usually not missing technology. It is that people, processes, data and software do not work together. If you automate a broken process, you get the same mess, only faster.

So the first questions are boring on purpose. Where does information get stuck? What gets typed twice? What waits for approval? Who hands work to whom?

Part 1: the work map

The work map is one page with five zones:

• Inbound: what enters the business every day. Orders, requests, emails, calls, documents.
• Processing: where the work spends its time.
• Decisions: where things wait until someone says yes.
• Handoffs: where work moves between people, teams or systems.
• Outbound: what leaves the business. Quotes, deliveries, invoices, reports.

You fill the map in with the owner or the operations lead, not for them. Keep it visible during the conversation, on a whiteboard or a shared screen. And use their language, not yours. If the team calls it "the order sheet", you call it the order sheet.

Part 2: the interview

The interview takes about 45 minutes and runs in four phases:

• The day. What actually happens? What lands in your inbox that you wish would handle itself?
• The bottlenecks. What always takes longer than it should? Where does the same information get entered more than once?
• The wish. If one part of your operation could be instant and error-free, which one would it be?
• The close. Of everything we discussed, what would you want fixed first?

One rule matters more than the questions: do not suggest solutions during the interview. Your only job is mapping. The fix comes later, from the rubric. And if they go quiet, wait. The useful answers often come after the silence.

Part 3: the decision rubric

The rubric turns observations into a ranked list. Each observation gets two scores from 1 to 5.

Impact: how much it matters. 5 means more than ten hours a week saved, or revenue that is currently blocked. 1 means cosmetic.

Complexity: how hard it is to build and maintain. 1 means configuring a tool you already have. 5 means a custom build that needs an engineering team.

Priority is impact divided by complexity. Sort by it, drop what the team will clearly resist, and pick three: one quick win, one that compounds over time, and one bigger bet.

Four patterns that show up again and again

• Data re-entry. The same data typed into two or more systems. Often the fastest win: connect what is already there.
• Approvals. Work stops until one specific person says yes. Clear decision rules often help more than any software.
• Document generation. Quotes, confirmations and reports that follow a template but are still assembled by hand.
• Finding information. People digging through emails and folders, or asking the one colleague who knows.

Notice that none of these start with AI. Sometimes the right answer is a better form, a shared list or one clear rule. Sometimes it is automation. Sometimes AI really helps. The rubric decides, not the hype.

What you get

One page. Three ranked opportunities. A rough estimate of time saved, a complexity level, and one concrete first step for each. Not a proposal, but a clear picture the owner can act on, with or without outside help.

How to run it yourself this week

1. Block 90 minutes, on your own or with your operations lead. Open a blank page or a whiteboard.
2. Draw the five zones and walk through the four interview phases. Write the answers into the zones, in your team's words.
3. Score every observation: impact 1 to 5, complexity 1 to 5. Then divide impact by complexity.
4. Pick your top three. One quick win, one that compounds, one bigger bet.
5. Take one concrete step on number one this week.

The method works because it is boring. No models, no APIs, no architecture diagrams. Just a structured conversation, a map, and a simple way to turn daily friction into ranked actions.