DoxaMind
← Insights

Which workflow should you automate first? A scoring framework

12 July 20268 min read

The management meeting ends the way it always does. Someone says the firm should be using AI by now. Everyone agrees. Nobody can say on which workflow, so the topic moves to the parking lot, where it has lived for a year.

The blocker is not conviction. It is that choosing a first workflow feels like a technology question, which nobody in the room owns. It is not a technology question. It is a portfolio question, and you can answer it tonight with a whiteboard and 4 axes.

List the candidates first

Write down every recurring workflow that produces a document: proposals, studies, quotes, courseware, client reports, meeting summaries, tender responses. Ten to fifteen lines is normal. Do not filter yet; filtering is what the scores are for.

Axis 1: volume

How often does this workflow run, and how many senior hours does each run consume? A proposal produced weekly at 12 hours beats a strategy document produced twice a year, regardless of how painful the strategy document feels. Score 1 to 5. Multiply frequency by hours in your head; precision does not matter, order of magnitude does.

Axis 2: judgment shape

The trap axis. The question is not how much judgment the workflow needs, but where it sits. Work where judgment comes at the beginning (framing the question) and the end (approving the result), with structured production in between, scores high. Work where judgment is smeared through every step, like final-round negotiation or original strategy, scores low. A veteran should be able to describe the middle of a high-scoring workflow as steps on one page.

Axis 3: data readiness

Where do the inputs live? Email, shared drives, spreadsheets, and mainstream software with proper interfaces score high. Paper, phone calls, and legacy systems that only a human can click through score low. This axis kills more AI projects than any other, and it is the cheapest one to check: trace one real recent run of the workflow and note where every input came from.

Axis 4: strategic value

If this workflow ran 5 times faster, would the firm make more money, or just feel tidier? Proposals and quotes usually score high: more answered means more won. Internal minutes score low. Be honest here; the axis exists to stop you automating what is merely annoying.

FIG 1 / A WORKED SCOREBOARDDOXAMIND
VOLUMEJUDGMENT SHAPEDATA READYVALUE
PROPOSALS
MARKET STUDIES
CLIENT REPORTING
MEETING NOTES
STRATEGY DECKS
A consultancy's real candidate list, scored. Dark cells are 4-5, tinted cells 2-3, empty cells 0-1. Proposals win on the first pass; the strategy work everyone assumed would be first comes last.

Read the scoreboard, then commit to one

Add the rows. The top score is your first project. The second becomes the roadmap, not a parallel effort: running 2 first projects halves the attention each gets and doubles the odds that neither ships. One workflow, taken all the way into daily use, changes how the firm thinks about the next 4.

FIG 2 / WHY THE ORDER MATTERSDOXAMIND
HOURS TODAYAFTER FIRST BUILD
PROPOSALS
STUDIES
REPORTING
Illustrative senior hours per month, before and after automating in score order vs enthusiasm order. Score order recovers the big block first.

Two sanity checks before you start

First, trace one real run of the winning workflow end to end, with the person who actually does it, not their manager. If the steps they describe differ from the steps everyone assumed, you just saved the project. Second, decide the review gate now: who reads the output before it leaves, and against what checklist. A workflow with no named reviewer is not ready to automate, whatever it scored.

DoxaMind runs this scoring as the first hour of every engagement, then builds a working system of AI agents for the workflow that wins, live inside a week and reviewed by your team. If you want the scoring run on your firm's list, join the waitlist.