Ask what it costs to catalogue an archive and you will mostly get silence. Software vendors say “contact us for a quote”. Project budgets bury the figure inside salaries and overheads. Funding applications estimate it afresh every time, from scratch, with no published benchmark to lean on. For a sector that runs on accountability — to funders, to committees, to the public — the economics of its own core activity are strangely undocumented.
This post is our attempt to put real numbers on the table: what manual cataloguing actually costs, how big the uncatalogued backlog really is, what AI-assisted cataloguing costs with Archivers.ai — every price published — and a full worked example for a 7,000-item backlog. If you are writing a grant bid, a committee paper, or a business case, the figures below are yours to use.
What manual cataloguing costs
Start with the human reality. A qualified archivist costs around £36,000 a year under ARA 2026 salary guidance — before employer’s NI, pension, desk, and management overhead. That salary buys deep professional judgement: appraisal, arrangement, contextual knowledge, the ability to recognise that an unmarked photograph shows the old gasworks before the 1936 fire. It does not buy unlimited throughput. Item-level description is painstaking work, and there are only so many records a person can examine, interpret, and describe in a day.
Grant-funded projects let us put a unit price on that work. Derived from Archives Revealed project outcomes, manual cataloguing in funded projects typically works out at £3–5 per item. That figure is not a criticism of anyone’s productivity — it is simply what careful, standards-compliant description costs when a professional does every item by hand: reading, identifying, dating, describing, assigning subject and name access points, and entering it all into a system.
At £3–5 per item, the arithmetic gets uncomfortable quickly:
| Backlog size | Manual cost at £3/item | Manual cost at £5/item |
|---|---|---|
| 1,000 items | £3,000 | £5,000 |
| 7,000 items | £21,000 | £35,000 |
| 25,000 items | £75,000 | £125,000 |
| 100,000 items | £300,000 | £500,000 |
For most archives, those numbers are not budgets — they are reasons the backlog exists.
The backlog is structural, not accidental
The scale of the problem is well documented. The National Archives’ sector survey found that around 13% of local-authority collections are uncatalogued, and for museums the position is worse: over 20% of museum collections have no catalogue record at all. Uncatalogued material is, for practical purposes, invisible — it cannot be searched, cannot be produced for a researcher, cannot support an exhibition or an FOI response, and cannot demonstrate its own significance when budgets are reviewed. We have written before about the hidden cost of uncatalogued collections; the short version is that an undescribed box costs money to keep and returns nothing until somebody describes it.
Multiply the unit cost by the backlog and the conclusion is unavoidable: at £3–5 per item, most backlogs will never be catalogued manually. Not because archivists are slow, but because no realistic funding settlement covers six figures of description work for a single mid-sized collection. The choice facing most services is not “manual cataloguing versus AI-assisted cataloguing”. It is “AI-assisted cataloguing versus the backlog staying exactly where it is”.
What AI-assisted cataloguing costs — published, per item
Archivers.ai prices cataloguing in credits, and every price is published on our pricing page — plans, top-up packs, and unit costs, including the institutional tier. One credit catalogues one photograph or object, one document page, one minute of video, or five minutes of audio (transcription included on paid plans). One Explore research message costs one credit; deep analyses such as whole-collection analysis or fonds drafting cost five. The app shows you the credit cost before you commit, monthly plan credits reset each month, and top-up credits last 12 months.
The plans:
| Plan | Monthly | Annual equivalent | Credits per month |
|---|---|---|---|
| Community | Free | Free | 30 |
| Starter | £40/mo | £32/mo | 500 |
| Professional | £150/mo | £120/mo | 2,000 |
| Institution | £400/mo | £320/mo | 6,000 |
| Enterprise / Consortium | Custom | Custom | Pooled, custom |
And for backlog projects, top-up credit packs on any paid plan:
| Pack | Price | Per credit |
|---|---|---|
| 500 credits | £70 | 14p |
| 2,500 credits | £300 | 12p |
| 10,000 credits | £1,000 | 10p |
| 50,000 credits | £4,000 | 8p — sized for an Archives Revealed cataloguing grant |
| 250,000 credits | £15,000 | 6p |
All prices exclude VAT. In practice, most customers pay 7–10p per catalogued item or page — typically 10–40× cheaper than the £3–5 manual equivalent. Those are unit prices you can drop straight into a grant budget or a committee paper, which is precisely why we publish them. If you are building a funding bid around these numbers, our heritage funding guide covers how AI-assisted cataloguing fits NLHF and Archives Revealed applications.
The same work, not lesser work
A fair question: is 7–10p per item buying the same thing as £3–5 per item? It is buying a different division of labour. The AI produces the first draft of every field — title, date, description, names, subjects — with the evidence for each cited, and nothing enters your catalogue until a human reviews and accepts it. Outputs are standards-native: ISAD(G), EAD3, Dublin Core, AtoM and ArchivesSpace CSV, Spectrum CSV, BagIt with PREMIS. The professional judgement still happens; it is simply spent on review and the difficult cases rather than on typing every field from a blank screen. We have set out how that review gate and provenance trail work in our post on defensible AI in the archive.
A worked example: the 7,000-item backlog
Take a concrete case — a community archive or small record office with a digitised but undescribed backlog of 7,000 photographs and documents.
Manual route. At the midpoint of £4 per item, describing 7,000 items costs roughly £28,000 of professional cataloguing time. For most services, that is a substantial grant bid or several years of opportunistic effort — and the backlog keeps growing while you wait.
Archivers.ai route. A Professional plan paid annually costs £1,440 for the year (£120/mo) and includes 2,000 credits a month plus the full analysis suite. Add one 10,000-credit pack at £1,000 and the backlog is funded on day one — 7,000 credits clear the 7,000 items as fast as your review capacity allows, and the remaining 3,000 pack credits (valid 12 months) plus the monthly allowance cover new accessions, Explore research, and deep analyses all year.
| Manual | Archivers.ai (year one) | |
|---|---|---|
| 7,000-item backlog | ~£28,000 at £4/item | £1,440 plan (annual) + £1,000 credit pack |
| Total | ~£28,000 | £2,440 |
| Effective cost per backlog item | £4.00 | ~35p all-in (7–10p marginal per credit) |
| Also included | — | Public portal, standards exports, Explore, 24,000 plan credits across the year |
That is an 11×-plus saving in year one — and the comparison is generous to the manual route, because £28,000 buys only the description work, while £2,440 buys the platform, the public access portal, and the export pipeline as well.
The honest caveat: your archivist’s time is still part of the equation, because every AI-suggested record needs human review before it is accepted. But reviewing a drafted record is a fraction of the effort of creating one, and evidence marks let reviewers concentrate on the fields the system could not support from the page. The saving is real; it is just a saving in hours as well as pounds.
What the numbers mean for decision-makers
If you are a head of service or a director, three things follow. First, the backlog now has a price — a defensible, published, per-unit price you can put in front of a committee, rather than an open-ended staffing request. Second, the comparison is no longer “AI versus archivist”: at these prices you are cataloguing what would never otherwise be done, using professional time for the judgement only professionals can supply. Third, procurement gets simpler — published pricing means no discovery calls to find out whether a tool is even in budget, and most plans sit comfortably under standard purchasing thresholds. We have written more for budget-holders on our directors and heads of service page.
See the numbers against your own backlog
Every figure in this post is on our pricing page, in public, including the institutional tier — no “contact us to find out”. And the arithmetic is easy to test: count your backlog, multiply by £3–5 for the manual route, then price the same backlog in credits.
Better still, run the experiment. The Community plan is free, forever, with no card — 30 credits a month to catalogue real items from your own collection and see exactly what the review workflow and the output quality look like before you spend a pound.
The backlog has always had a cost. Now, at least, it has a price.