REFERENCE SCENARIO · PROFESSIONAL SERVICES
Reference scenario - composite, not a named client. A composite build illustrating our method. Figures are modeled and the model is shown.
This ai content pipeline case study is a reference scenario. No agency sits behind it. It shows how briefing, research, drafting and scheduling can be worked by agents while an editor approves every piece, and how each figure here is calculated.
What this page is not
No agency and no client roster is described here. We have nobody’s written consent to publish a content program, and describing one closely enough to be recognized would be the same problem as naming it.
This is a composite. The pipeline stages are the ones we build against. The account count, the volumes and the minutes per stage are inputs chosen so the arithmetic is checkable. None is a measurement.
The situation this build answers
A small team produces regular content for several accounts at once. Each piece is a similar sequence, run slightly differently every time because a different person runs it.
The sequence is the work. Brief, research, draft, edit, format, publish, cut down for social. Only two of those stages need the editor’s judgment, and all seven currently need the editor’s calendar.
Where the time actually goes
- The brief is rewritten from scratch for every piece, though the shape rarely changes.
- Research is collection, not analysis, and collection is what an agent is for.
- The blank page costs more than the edit, most weeks.
- Formatting and scheduling are pure administration and they land on the most expensive person.
- Social cut-downs get skipped when the week runs short, so the piece underperforms quietly.
The hive, agent by agent
SEO Content Agent
Builds the brief from the account’s own positioning and the target query, assembles sourced research, and produces a first draft with every claim linked.
Social Scheduler Agent
Cuts the approved piece into channel variants, drafts the schedule, and holds everything for one editorial approval.
The orchestrator
Runs the pipeline per account, keeps each piece in one state at a time, and stops the whole sequence at the approval gate rather than around it.
The generative layer underneath is not decoration. Draft quality here is a retrieval and context problem, so this build sits on generative AI development, with the agent behaviors from AI marketing agents.
What the agents may not do
Autonomy boundary
Agents brief, research, draft and schedule. They never publish, never send to a client, and never post to a channel without an approval.
Approval gate
An editor approves every piece and every social variant. This is the one rule in the build with no exception path.
Sourcing
Every factual claim in a draft carries its source. An unsourced claim is flagged in the draft rather than quietly kept.
What stays human
Point of view, the client relationship, anything about a named third party, and the decision that a piece is good enough.
Logging
Every brief, source, draft version and approval is recorded with a timestamp. Read our governance approach for how that is held.
How the build would run
| Phase | Typical | What happens |
|---|---|---|
| Audit | 3-10 days | The last quarter of published pieces is timed stage by stage, which produces the real minutes the model below assumes. |
| One account | 2-3 weeks | Agents brief and research only. The editor still writes. Draft quality is scored against the account’s own published work. |
| Drafting | 3-4 weeks | First drafts move to the agent. Edit time is measured every week, because it is the number the whole model turns on. |
| All accounts | 4-8 weeks | Accounts are added one at a time, each with its own voice reference set. |
How to read this ai content pipeline case study
Every figure below is arithmetic on stated inputs. The multiplication is shown, so the disagreement can happen where it belongs.
Volume is assumed at eight accounts producing three pieces each, so 24 pieces a month.
| Stage | Today | With agents |
|---|---|---|
| Brief | 45 | 10, approving a drafted brief |
| Research | 90 | 15, scanning assembled sources |
| First draft | 180 | 75, editing an agent draft |
| Edit | 60 | included in the 75 above |
| Format and publish | 30 | 5 |
| Social cut-downs | 40 | 10, approving variants |
| Total per piece | 445 | 115 |
The modeled outcome, with the arithmetic
| Figure | The arithmetic | Label |
|---|---|---|
24 pieces a month | 8 accounts × 3 | Modeled |
178 h editorial time today | 24 × 445 min = 10,680 min | Modeled |
46 h editorial time after | 24 × 115 min = 2,760 min | Modeled |
132 h returned per month | 178 − 46 | Modeled |
~74% less editorial time | 132 ÷ 178 | Modeled |
~7.4 h per piece today | 445 min ÷ 60 | Modeled |
~1.9 h per piece after | 115 min ÷ 60 | Modeled |
100% of pieces approved by an editor | Design rule, not a rate | Target |
What we would measure, and what we will not claim
We publish no outcome range for this build. No consented client result exists behind it, and a modeled figure is not a result.
What we would change next time
Give the agent the briefing stage for a month before the drafting stage. A good brief improves a human draft as much as a machine one, and it is the cheapest part of the pipeline to get wrong.
Build the voice reference set from work the editor is proud of, not from everything published. A draft trained on the whole archive averages toward the weakest piece in it.
The parts this build is made of
AI marketing agents
The agents that brief, draft, schedule and report.
Generative AI development
Retrieval, prompts and evals, so draft quality holds past the first month.
SEO Content Agent
Briefs, sourced research and first drafts.
Social Scheduler Agent
Channel variants and a held schedule.
AI agents for professional services
Where else agents pay off in this sector.
