Some of this work is about cost. The re-keyed order, the freight quote chased by phone. More of it is about revenue you aren’t capturing: the spec sheet a specifier couldn’t get at 9pm, the installation video that never got made because the quote came back at five figures, the configured product nobody could price without a rep. We build the automation into the systems you already run, not alongside them.
Live systems, in production, that you can ask us about in detail.
A distributor’s reps quote LTL freight in the ERP as they always did, then set one status. Everything after that is automatic: the customer is emailed a secure payment link, the payment settles, the hold clears through the API, and the order releases to the warehouse. Nine steps, seven of them unattended.
Read the case →Work board, client intake, repo access and reporting in one system, with an assistant that answers questions against live project data and respects each person’s permissions. We use it every day.
Requests arriving by email and Slack are read, summarised, matched to the right client and existing work, and queued with a recommendation, before anyone opens them.
Hazmat obligations detected from cart contents, forcing registration and capturing acknowledgement only on the orders that actually require it. The rule is enforced by the system rather than remembered by a person.
Read the case →An instructional reel is a configuration file. Narration is synthesised, captions are aligned to the audio word by word, brand treatment and music are applied, and the finished cut assembles itself. The second video costs a fraction of the first, which is what makes covering a whole product line realistic.
Read the case →An automated assistant answering from a manufacturer’s own catalogue and documentation rather than a generic model, so the specification questions that used to interrupt the sales team get answered correctly at 9pm.
Read the case →A reporting database the client owns, fed nightly from the storefront, the ERP and the marketing data, so a question that needs two of them at once has somewhere to be asked. Cost comes across from the ERP, which is what makes margin reportable rather than just revenue.
Read the case →A WebRTC audio service carrying a live market feed to listeners with the latency budget of a phone call rather than a stream. Built from a prototype proving the transport, then run in production continuously since 2020.
Read the case →Market news read, filtered and assembled into a briefing that goes out before the opening bell every trading day. It is the editorial job of a person, running unattended.
Read the case →These aren’t case studies. They’re the work we’re actively taking on. We’d rather tell you plainly what’s built and what’s being built than blur the two.
Manufacturer product data is usually spread across spreadsheets, PDFs and someone’s head. Agents can normalise attributes, reconcile variants against the ERP, and flag what contradicts itself, across tens of thousands of SKUs rather than the few hundred a person can review.
Specification documents assembled from live product data at the moment they’re requested, so what the engineer downloads matches what the system actually holds.
Configuration rules that live in institutional memory, captured and applied consistently, so a new rep quotes like your best one, and the answer traces back to the rule that produced it.
Grounded in your product data and documentation rather than a generic model, answering the specification questions that currently interrupt your sales team.
When the ERP, the storefront and the accounting system hold different truths, the difference is usually found late and by hand. It can be found nightly and reported instead.
Start there. We’ll tell you honestly whether it’s worth automating yet.
Start a project