Document back-office and written output
The easiest gain to obtain, the hardest to bank
120 staff spending a quarter of their time on professional writing: minutes, notes, standard replies, summaries. A deliberately conservative model, with an explicit redeployment rate.
- Modelled yearly gain
- €186k – €421k
- Payback
- 3.4 to 7.7 months
- Over 3 years
- €328k to €1.03M net over 3 years
modelled yearly gain
One hundred and twenty support-function staff spend roughly a quarter of their time producing professional writing: minutes, briefing notes, standard letters, replies to internal requests, documentation. A controlled experiment published in Science measured the effect of a generative assistant on exactly that kind of task, across several hundred college-educated professionals.
- Drafting assistance for structured, recurring documents, based on in-house templates
- Summarisation of long documents and meeting minutes
- Professional rewriting and translation without going through a vendor
- Information extraction across internal document corpora
Everything on the table
- Headcount in scope
- 120 staff × 1,500 h = 180,000 h/year
- Share of time spent on professional writing
- 25% → 45,000 h
- Share of that writing genuinely comparable to the tasks tested (haircut)
- 50% → 22,500 eligible hours
- Loaded hourly cost
- €55/h
- Year 1 investment (licences, integration, templates, training)
- €120,000
- Annual running cost in later years
- €55,000/year
Three scenarios, not one number
| Parameter | Conservative | Central | High |
|---|---|---|---|
| Writing-time reduction applied | −25% | −35% | −40% |
| Hours freed | 5,625 h | 7,875 h | 9,000 h |
| Share of freed hours genuinely redeployed to value | 60% | 75% | 85% |
| Hours valued | 3,375 h | 5,906 h | 7,650 h |
| Yearly gain | €186k | €325k | €421k |
| Payback period | 7.7 months | 4.4 months | 3.4 months |
| Net cumulative gain over 3 years | €328k | €745k | €1,033k |
- This is where the gap between theoretical and banked gain is widest. Time handed back to people who are already under-loaded appears in no P&L.
- The study covers short, self-contained writing tasks. A document that requires cross-referencing internal sources, committing the company or meeting a regulatory formalism falls outside that scope.
- The measured quality gain benefits the least confident writers most. On a team of expert writers, expect a real time gain but a weak quality gain — negative if review discipline slips.
- Any data leaving the information system needs an approved usage framework — confidentiality, trade secrets, personal data. That point drives the schedule more often than the technology does.
- In-house templates: the value comes from grounding in your formats, not from a generic tool.
- An explicit decision by managers, before rollout, on how the freed time is used.
- A clear review rule scaled to how far each document commits the company.
- A written usage policy and a list of documents that never go through an external tool.
What these scenarios are, and what they are not
These are models, built for illustration on public sector benchmarks. They are not results observed at Cardan-AI clients, and they constitute no commitment as to outcome. The improvement rates come from the studies cited; the choice of assumptions, the scope haircuts and the arithmetic are Cardan-AI's, and are shown in full so they can be challenged. Transposed to your organisation, these orders of magnitude can vary widely — in both directions.
Test these assumptions against yours (30 min, free)
We swap this scenario’s assumptions for your real figures, and you leave with a model you can defend internally.