We build AI systems for the business side of books: contracts and rights, royalty statements, manuscripts, metadata and marketing. They run inside your pipeline.
What we see in the market
Each hand-off runs on contracts, statements and metadata, and each one asks for careful, detailed work from your team. Point at any dot to see what the research says.
of North American publishers find it hard to explain to authors how their work is protected from AI.
Swipe the map sideways →
Our research
We read the national surveys behind our AI Maturity Index. The picture repeats from market to market: AI is in use, and strategy, tools and rules lag behind.
of North American publishers have an official AI policy. Among those with 100+ staff, 48.2%.
BISG × BookNet Canada · 2025AI Maturity Index for book publishers
Questions about how AI works in your publishing house, from rights and production to metadata and readers. You get a score out of 100, a breakdown by area with cases from the industry, and the growth areas to work on first.
What we build
Each system helps your team with work they already know well: it does the reading, matching and checking, so people can focus on decisions, authors and partners.
Every build starts from one number that matters, measured before and after on your own data.
It runs inside your perimeter, never trains outside models, and people approve every rights, money and editorial decision.
Everything sits on one knowledge base of your contracts, statements and titles, and the code and models are yours to keep.
Option deadlines, reversion dates, reporting periods and licence terms by language and format, pulled from every contract into one register. Renewal notices and royalty requests go out on schedule. What can you sell right now, and in which language? Seconds, not days.
A combined statement of a thousand pages is split by contract. Each line is matched to its contract and right, including sublicences with no ISBN, and recalculated against the terms: rate tiers, price basis, reserves, advance, tax and commission. Missing contracts and underpayments are flagged with the page and line they came from.
After the check, the system gathers every contract a rights holder has across all publishers for the period and builds their report in the BISG standard. A person approves, the system sends it from your mailbox and keeps the archive. Nothing leaves without approval.
The system knows what each buying publisher and imprint actually publishes, from your past deals and open data. It reads new rights catalogues as they arrive, keeps only titles whose rights are free in that language, and writes each buyer a short list in their language with reasons why each book fits. The agent approves, and replies teach the system what works.
A detailed card on each incoming manuscript or partner title: subject, strengths, comparable books, audience, the author's reputation, potential in other languages, and an estimate of cost and payback on your own print-run figures. The system ranks the flow. The editor decides.
The system finds bloggers, academics and cultural figures in the niches and countries you name, people whose audience is loyal enough to buy their book. A card on each: subject, audience, engagement, past publications, and a first message in their own language. A person leads every conversation.
The full set of materials for every distributor, built to its standard. Keywords, category codes, blurbs, comparable titles and alt texts rewritten across the whole backlist, which is what moves sales in online shops. Alt texts are required for ebooks in the EU since June 2025.
Search the catalogue by topic, question or reader need, using descriptions, contents and permitted text, on your website, in your shop and inside the house. Readers find titles beyond exact keywords, and the backlist gets found again.
The system reads the whole manuscript inside your own perimeter, so the text never goes to an outside service, and returns a working pack: blurbs of several lengths, comparable books, audience profiles, social posts and video scripts, newsletter and ad copy, and a draft press release, in your house style. Your marketing team chooses and edits instead of starting from a blank page.
It answers strictly from your books, courses and materials, citing the edition and page. A reader describes their situation and the assistant walks them through the relevant method, in every language your platform supports. Sold inside a paid course or as its own subscription.
Translation against your house glossary, synthesis in the author's voice, human quality control and clear labelling, in the languages where your audio rights are free.
Harness · Vault
Our whole team works in one environment: work lands in a shared brain, and repetitive tasks become AI agents. Several client pilots run the same way.
How we work
Senior engineers sit with your team and map how the work really runs.
One number that matters, such as time to a checked rights answer.
The smallest system that moves that number, inside your perimeter.
Baseline before, result after, on your own data.
You pay for the milestones you accept.
A dead end shows up in week two, not in month six.
Two to five senior engineers and a CTO-level owner. No junior layer.
Code, models and the runbook transfer to you. No licence fees.
Company
Team
Next step
Tell us where the work piles up. We come back with a first sketch: where AI fits in your systems, how the data stays protected, and what it would take on your side.
About 10 minutes. Your score out of 100, your place among other publishers, and the growth areas to work on first.
Igor Nikolaev, CEO