AI Agent document controller — project cost model
Labour + cost of non-quality, by project size · figures in AUD · every assumption below is adjustable
Project size — documents1,000
Saving over project
Saving
of human-run cost
Payback on setup fee
Quality issues prevented
Cost to run document control — this project
Total cost vs project size
100 → 10,000 documents (log scale). The gap widens because quality failures grow faster than headcount — multi-centre execution, people rotation, procedure drift. Steps in the blue curve are real headcount increments (you hire in half-FTEs, not minutes).
Human-run document control
DocController AI + human oversight
Assumptions
Change any number — sceptics welcome. Defaults are deliberately conservative.
Workload
Issue cycles per document (e.g. IFR → IFA → AFC)
Time to process one revision — the client's own benchmark
Transmittals, registers, reporting, chasing — on top of processing
Set duration manually
Otherwise estimated from project size
Labour
Client benchmark: ~$80k
Super, leave, insurance, overheads
~220 days × 8 h
People come in halves, not tenths — staffing is quantised
Quality & non-quality cost
Small co-located team, everyone knows the procedure
Multi-centre, rotation, subcontractors, procedure drift
Chasing, investigating, correcting, re-issuing
Engineering re-work and re-issue effort
Share of issues that reach fabrication, site or the client
Blended: rework, standby, delay. One offshore standby day can exceed $100k
AI solution
Rule-based classes: numbering, issuance purpose, revision & distribution control, metadata
Human approvals and judgment calls stay — by design
One-off: project rules, knowledge base, connector setup
Software service over the project duration