AI content operations cost per asset: a worked method

What to take away

The unit cost is the pricing step beneath the full budget structure in the costs and budget guide. It needs only invoices, a timesheet and a count of what was published.

Define the unit before you count anything

A subscription is a fixed cost, but your output is not. Price one unit: a researched, edited, published article or landing page. Every other figure in the model is divided by the count of those units.

A unit needs a boundary. Decide whether an asset is finished when it is edited or only when it is published, and count it the same way every month. Without that rule, a comparison between two months measures the counting method, not the operation.

Divide the tool spend by volume

Fixed tool spend becomes a unit cost only when you divide it by output. A team paying £400 a month for a writing tool, £150 for a research tool and £90 for a grammar checker has committed about £640 a month before anyone writes a word.

Twelve assets a month gives about £53 of tool cost per asset. Three assets a month gives about £213. Same tools, very different economics, and the difference comes entirely from how many units the tools are asked to support.

Add the other three lines

Separate tool spend, people time, paid distribution and compliance work, because each behaves differently when volume changes. Tool licences are the easiest line to see and often the smallest, so the budget conversation should not start and stop there.

People time is usually the largest line: briefing, prompting, fact-checking, editing and sign-off. Count it in hours per asset, taken from a month of real work, and price those hours at the internal rate.

Paid distribution is the third line. If you promote content, benchmark it against published rates rather than guessing, and LinkedIn's own advertising costs and pricing page is a reasonable starting point for B2B assumptions.

Compliance is the fourth. If content or outreach touches personal data, you must determine your lawful basis before you start using it and document that choice. The wider data protection duties take staff time as well.

Adopting a recognised framework such as the NIST AI Risk Management Framework, which is intended for voluntary use, is a choice that takes staff time too, and auditors ask for the evidence.

Worked example in pounds

The figures here are illustrative, not benchmarks. They show how the four lines combine into one monthly total, and then into a cost for each published asset.

At 12 published assets a month, the unit cost is about £287. If output falls to eight assets, the unit cost rises to about £430 while the total barely moves. That gap is where most budget plans break.

Decide which lines to fix and which to flex

Fix the licence and compliance lines, because they change slowly. Flex the freelance editing and distribution lines, because you can turn them down in a quiet quarter. Staff time sits between the two and swings with how much review each asset needs.

A fixed line is a commitment even when output drops, so it belongs in the plan at full weight. A flexible line is a decision, so write down who may change it and when.

Stress-test the model

Run the same arithmetic at lower output before you commit to a budget. If the unit cost becomes unacceptable, your model depends on volume you may not reach.

Rework, governance and review hours often sit outside the four lines until someone measures them. The hidden costs article scores them against a supplier quote, and it works alongside a unit cost rather than replacing it.

Turn the unit cost into a return case

A cost model earns its keep when it connects to revenue or saved hours. The return on investment method attaches value to each unit, which turns a cost figure into a budget case.

Without that link, you are comparing subscriptions. With it, you are comparing the cost of one published asset against the value that asset is expected to produce.

Common questions

Should I include staff time in the cost per asset?

Yes. If you exclude it, you are measuring software, not operations. Staff time is usually the largest line, and it swings with how much review each asset needs.

How often should I rebuild the model?

Re-run it whenever output volume, headcount or tool pricing changes. A change of cadence should trigger an immediate update, and a quarterly refresh covers the rest.

Is compliance a one-off cost?

No. Lawful-basis decisions, data reviews and framework mapping recur as your processes and suppliers change, so keep a standing allowance rather than a project line.

What if I cannot estimate staff hours yet?

Track hours for one month on a small sample, then extrapolate. A rough measured figure beats a confident guess, and you can refine it as the sample grows.