A rapid AI prototype to auto-file email attachments at a private equity firm

A private equity firm now runs a working AI prototype that automatically files inbound email attachments into its document management system. The engagement delivered an indicative estimated saving of up to £10,000 a year, more consistent filing than manual handling had produced, and, most valuably for a firm still forming its AI strategy, a demonstrated capability that built internal momentum for further automation. It is a small-scope, fast-delivery story: one workflow, built quickly, that earned the right to further automation.

The problem it solved was ordinary, and that was the point. Operating partners could see AI's potential but lacked the time and technical experience to build anything themselves, and were wary of committing to a large transformation programme before they had any proof it would work. What they wanted instead was something small, fast and provable. The task they chose is one every deal team recognises: attachments arriving by email that need filing into the right deal, company and folder within the firm's document management system. Done by hand, this is slow and repetitive, and filing choices vary by whoever happens to pick up the email, so folder structures drift and inconsistency accumulates over time, quietly displacing hours that should go into deal work. Fixing it was low-risk and immediately legible to a non-technical sponsor, which made it the right first bet for a firm testing whether AI automation was worth pursuing further.

QuantSpark's delivery method mattered as much as the workflow itself. The engagement followed a value-first, learn-by-doing approach, with success criteria agreed before any code was written, so the prototype could be judged against a clear bar rather than an open-ended scope. From there, the build proceeded in a sequence of practical steps. Inbound triggers were configured first, using rules on sender, target folder and subject line, so that only relevant emails entered the pipeline. Next, an AI layer read and summarised each email's content and extracted the company or deal it related to. A looping sub-folder crawler then iterated the existing file tree in the document management system to find the most relevant location, creating a specific leaf folder where none already fitted. Before anything was filed, a validation and approval step reviewed the proposed destination path, prompting the user for confirmation when required rather than filing unattended. The whole prototype was shipped quickly as a working demonstration, deliberately built to be learned from and iterated on rather than polished up front.

Laid out as a pipeline, the workflow runs cleanly from inbox to filed document: an email arrives and matches the configured rules; its content is read and its company identified; the folder tree is searched for the best match, or a new folder is created; the proposed path is validated, with a human check where needed; and only then is the attachment filed. Each stage is a discrete, auditable step, which is part of why operating partners with no technical background could follow, and trust, what the system was doing.

The systems involved stay categorical rather than naming specific products: an email system supplying the trigger, an AI or large-language-model layer doing the summarising and extraction, automated folder-crawling logic doing the navigation, and the firm's existing document management system as the destination. Nothing proprietary or client-specific needed to be built from scratch; the value came from stitching familiar categories of technology into one dependable path.

On value, the case study is careful to separate what was delivered from what was modelled. The £10,000-a-year figure is explicitly indicative, not an audited outcome, and should be read as a saving of the kind typically estimated by weighing the time manual filing would otherwise take against the volume of attachments handled and the value of the hours freed up. QuantSpark has not published the underlying assumptions behind this particular estimate, so the number should be treated as a modelled saving rather than a measured one. Sitting alongside it are two harder-to-dispute gains: filing that is now more consistent than manual handling produced, and a live example the firm could point to internally when making the case for further automation. For a firm that started this engagement wanting proof before commitment, that combination, small delivered workflow, credible if indicative saving, and genuine internal momentum, is the more durable result.