We signed up for the real products, clicked through them, and kept screenshots. Facts below are from those teardowns, dated 2026-07-04; each vendor's claims are quoted as their claims. Prices are their published prices for a book of clients like yours.
prices as of 2026-07-04
| Product | 10 clients | 20 clients | 50 clients |
|---|---|---|---|
| AgencyAnalytics flat per-client pricing; Enterprise (sales-led) from 25 clients | $200/mo | $400/mo | sales call |
| DashThis about 4 simple sources per client; data-rich clients hit the source cap and force the next tier | $139/mo | $279/mo | $429+/mo |
| Whatagraph source credits; self-serve ends at $229 for 20 credits - a 10-client, 5-channel book is already sales-gated | ~$812/mo | custom (sales) | custom (sales) |
| Databox agency ladder, 5 sources per client, plus a white-label add-on until $799/mo | ~$286/mo | ~$406/mo | ~$1,006-1,279/mo |
| Swydo 3 sources per client, monthly billing; verified against the in-app estimator | $159/mo | $294/mo | $624/mo |
| Narrative BI $30-40 per connected source per month; no self-serve signup exists | sales-led | sales-led | sales-led |
| A Nightdesk build | priced once - $9,000-10,000 comparable range - editions at pennies, and it is yours | ||
what it does wellThe category leader: flat per-client pricing with no feature gating, and the deepest white-label.
where it stopsBooks past 25 clients route to a sales team, and its accuracy story is "AI paired with human oversight" - their words, an assurance rather than a mechanism.
what we build insteadNo per-client meter, and a mechanical check in place of oversight.
what it does wellThe fastest first report in the category: templates by industry and sample data everywhere.
where it stopsReports stop being accessible when payment stops, and the source catalog is marketing-only.
what we build insteadEditions that are files and a database you own - they never go dark.
what it does wellThe most polished output in the category, and prompt-to-report onboarding that finishes in about a minute.
where it stopsScale is sales-gated, and the AI prose is not bound to the computed numbers.
In our trial, its AI summary and its own charts disagreed inside one report (screenshots on file).
what we build insteadWriting that cannot contradict the charts - it is checked against them mechanically.
what it does wellThe best no-login demo in the category, and an AI analyst over 130+ connections.
where it stopsYou assemble dashboards yourself, and the AI's figures carry no provenance.
Its own public demo displayed a broken comparison card (captured 2026-07-04).
what we build insteadEditions that assemble and send themselves, every figure traced to its pull.
what it does wellExcellent whole-book machinery: linked templates, delivery tracking, source health checks.
where it stopsIts AI summaries are kept away from clients behind four layers of paid, human-copy-only friction.
The fencing itself is the fact: the vendor does not let its own AI text reach a client unreviewed.
what we build insteadA narrative that can ship to clients - because it is proven, not fenced.
what it does wellThe closest idea to ours: scheduled, auto-written data stories with a genuinely good template.
where it stopsIt publishes an accuracy benchmark of roughly three errors per hundred for its shipping model - and ships anyway, with no per-report check.
what we build insteadZero untraceable figures, fail-closed - a guarantee, not a benchmark.
They say: “AI paired with human oversight” - “the same number every time” - “about a three percent error rate”.
We say: a report with one untraceable figure does not ship. Not a promise - a mechanism, proven in our test suite and in production.
| Their own claim the category leader, quoted 2026-07-04 | ~2.5 h per client, monthly |
| The arithmetic 20 clients x 2.5-5 h x $50/h - arithmetic, not a measurement | $2,500-5,000 a month |
| With Nightdesk the human step that remains | review and approve |
And the AI that does the writing is your choice - a fast model for everyday editions, the strongest model for the hardest writing, or open models on your own machines. The first two run on our own subscription at no per-call charge. Price follows your needs, not a per-source meter.
A fast model for everyday editions.
runs on our own model subscription - no per-call charge
The strongest model, for the hardest writing - what writes the sample book today.
the same subscription - a measured $0.00 an edition
Open models on your own machines - we set them up.
the spoken voice already runs locally today
This system already runs. We build it for your business - and you try it first, free.