Skip to content

002 · TECH

The right way to generate documents at scale

Automating document production sounds straightforward: until you try to scale it. Three places it usually breaks, and the shape of a setup that holds.

The hours nobody counts

Most teams reach a point where the same handful of documents, contracts, certificates, convocations, invoices, audit packets, eat several hours a week of work that nobody enjoys and nobody budgets. The work is invisible because it is spread thin: ten minutes here to fill a template, five there to find the right version, a morning once a quarter to rebuild the packet an auditor wants.

Generating them automatically is the obvious answer. Doing it in a way that survives contact with reality is the harder one, and the difference between the two is where most document projects go to die.

Where it breaks: templates

The first failure is template rigidity. When engineers own the templates, in code, in a PDF library, in a layout nobody else can open, every wording change waits in their backlog. The legal team wants a clause moved; the change ships in three weeks. After the second such wait, the team goes back to Word and the automation quietly stops being used.

The fix is to let the people who own the words own the template: a Word or HTML document with named placeholders, stored where the content team can edit it, versioned so that last year's certificate still renders like last year's. The engine fills it; it does not define it.

Where it breaks: data and triggers

The second failure is data access. Off-the-shelf generators see only their own database, and the fields a real document needs live in three systems: the CRM has the name, the ERP has the amount, the training platform has the dates. Somebody exports, pastes and checks. A thin connector layer, one function per source, returning exactly the fields a template can ask for, removes that person from the loop.

The third failure is the trigger. A human still has to remember to launch generation, and a process with three places to forget a step will forget it. Documents should follow events: a session ends, the certificate is produced; a contract is signed, the confirmation goes out; the month closes, the packet assembles itself. Event-driven generation is what makes the difference between a tool people use and a tool people are told to use.

What good looks like

Editable templates owned by the content team. A connector layer that exposes every system holding relevant data. Event-driven generation that ships through whichever channel the receiver expects: e-mail, a portal, a signed PDF, an archive. Pull the three together and document production stops being a category of administrative work; it just happens while the business runs.

At C-Certif, that shape produces the certificate, the funding file and the audit trail the moment an assessment is validated. Nobody generates anything. The team's time went to the exceptions, a name misspelt at registration, a funder with a bespoke form, which is the only part of the work that ever needed a human.