Ji-woo Han Advisory memo Decision Call · USD 260
Which AI use case first, and the honest case for none of them
60 minute call, 21 January. Reviewed and signed off by Ji-woo Han on 23 January.
You brought four candidate use cases and asked which to start with. Three of them are worth doing and one is not, but the more useful answer is that you have picked the four that are most visible rather than the four that are most expensive, and those are different lists.
Editorial drafting is the one everyone wants and the one with the worst risk-to-value ratio at a publisher. Metadata tagging is dull, entirely internal, has a clean measure, cannot embarrass you in public, and on your own numbers releases roughly 40 hours a week of production time. Start there, get the governance working on something safe, and revisit drafting in six months.
The question
What was asked, and what was reviewed
As you put it: "The board wants an AI initiative this year. We have four ideas. Which one?"
| Reviewed | Source |
|---|---|
| Four use case descriptions | Supplied, one page each |
| Production workflow, editorial to publication | Walked through on the call |
| Time tracking by production task, 6 months | Supplied |
| Your existing metadata standard | Supplied |
| Author contracts, two representative | Supplied |
| Board minute requesting an AI initiative | Supplied. Worth reading, see clause 4 |
Comparison
The four, on the same basis
| Metadata tagging | Rights search | Marketing copy | Editorial drafting | |
|---|---|---|---|---|
| Time it releases per week | ~40 hrs | ~12 hrs | ~9 hrs | Unclear |
| Measurable against a baseline | Yes, cleanly | Yes | Partly | No |
| Visible to readers if wrong | No | No | Yes | Yes |
| Touches author contracts | No | Yes | No | Yes |
| Needs a human review step | Sampling | Every result | Every output | Every word |
| Staff reaction, predicted | Relief | Indifference | Mixed | Hostility |
| Reversible if it fails | Entirely | Entirely | Yes | Reputationally, no |
| Time to a first result | 6 weeks | 10 weeks | 4 weeks | 6 months |
Marketing copy is the fastest to a result and releases the least time. Metadata tagging releases four times as much and cannot hurt you. That is the whole argument.
Against
Why editorial drafting is the wrong first move
- Your author contracts, on the two I read, are silent on machine-assisted editing. Silence is not permission and it is not prohibition, and establishing which it is with your authors is a conversation to have deliberately rather than after a pilot.
- There is no baseline. You do not measure editorial quality today, so you cannot show the change, and an initiative with no measure becomes an argument about taste.
- Every output needs full human review, which means the saving is in the first draft only, and your editors say the first draft is the part they enjoy.
- It is the only one of the four that is visible to readers if it goes wrong.
- It is the one the board has heard of, which is why it is on the list.
The board
What the minute actually asks for
You described this as the board wanting an AI initiative. The minute of 8 December says something narrower and more useful: "management to report on opportunities to reduce production cost through automation."
| What you heard | What it says | Consequence |
|---|---|---|
| The board wants AI | Reduce production cost through automation | Metadata tagging answers this precisely |
| Something visible and strategic | A report on opportunities | They asked for analysis, not a launch |
| This year | No date stated in the minute | Less time pressure than you think |
Worth re-reading the minute before the next board meeting. On the words as written, the recommendation in this memo is a better answer to the board's actual question than editorial drafting would be.
Next
What to do
- Confirm the 40 hour figure with your production manager. It is derived from your time tracking and it is the number the whole recommendation rests on.
- Scope metadata tagging as a six week pilot with a sampling review, a named owner and a before-and-after measure.
- Take the comparison in clause 2 to the board as the answer to their minute. Four options, scored, one recommended.
- Ask your rights lawyer the author contract question now, separately and unhurriedly, so that drafting is a live option in six months rather than a blocked one.
- Do not start marketing copy at the same time. Two pilots at once halves the attention each gets and this organisation has not run one yet.