Valtrun
Our work

Proof you can click on.

Platforms our team built and still runs, the AI we are adding to them, and two working demos you can try right now.

01Selected work

Built to last. Now with AI.

Our engineering team designs, builds and runs production platforms for retailers, consumer brands and member networks. Each case shows what they built, and the AI we are adding to it now.

The hard part is never the AI model. It is live systems, messy data, measured accuracy, human sign-off and security, which is why each upgrade takes an engineering team rather than a chatbot.

Case 01 · Luxury retail

Six stores. One view of every customer.

The brief

A family-owned perfume and cosmetics retailer with six stores

A long-established family business selling designer fragrance and cosmetics from six stores, known for the advice of its in-store makeup artists and skincare specialists. Moving online, it needed a store that carried that service to the web instead of competing with its own shops.

The AI upgrade

Every store knows every customer

Each of the six stores knows its own regulars, stock and specialists, but nothing is shared. Every day staff answer the same shade and availability questions by phone and email, ring other stores to check stock and put items aside, and take consultation bookings by hand. A customer who asks online about a shade, or buys a fragrance in one store, is a stranger in the next. Every question that waits for an answer is a sale that can go elsewhere.

A stock question becomes an item held in a named store within minutes, and about 20 hours a week back across six store teams.

Case 02 · Skincare brand

Trade orders, without the typing.

The brief

A doctor-founded skincare brand selling worldwide

An award-winning skincare brand founded by a practising doctor, selling direct to customers in several countries and wholesale to salons, clinics and stockists. It needed one site that served both: the right currency for every shopper, and trade prices only for approved businesses.

The AI upgrade

Hands-free trade orders

Trade customers still order the old way: emails, PDFs, spreadsheets, even photos of handwritten lists. For every order, someone on the order desk reads it, finds the products behind old names and shorthand, applies the account's terms and keys it in, then chases the stockist about anything unclear. Orders wait a day or two before they are confirmed, so goods ship and invoices go out late, and a mis-keyed line becomes a wrong shipment and a credit note.

Trade orders confirmed the same day they arrive, and about 18 hours a week back for the order desk.

Case 03 · Impact investing

Better introductions, every week.

The brief

An invitation-only network of impact investors

An invitation-only network of impact investors, partnered with peer platforms, wanted one place where investors and institutions could find entrepreneurs in the cities, regions and sectors they care about.

The AI upgrade

Deal-flow matching

Every pitch deck that reaches the network is read by a small team, its facts typed into the platform by hand, and the right investors worked out from memory and spreadsheets. Each introduction is then written one at a time. Decks sit in the queue for weeks, and investors only see the opportunities someone happened to remember them for. The network's value is the quality of its introductions, and that is capped by how many decks a few people can read.

Every deck in front of the right investors the week it arrives, and about 15 hours a week back for the team.

Case 04 · Membership network

Every event, followed up.

The brief

A multi-city membership network for social investors

A network that began around a boardroom table and grew into a community of family offices, foundations, funds and intermediaries meeting in several cities. It needed a strong identity, a website that attracted members, and a way for staff to run events everywhere.

The AI upgrade

Follow-up that actually happens

Every event means invitations, introductions, notes and follow-ups, written by hand for each city by a small staff. The value of an event is in the follow-up, and that is what gets dropped when the next event is already being planned. First-time guests are the network's best source of new members, and they are the ones whose follow-up is most often missed.

Every attendee followed up and every first-time guest invited to join, with about 10 hours a week back for the events team.

Clients of our engineering team, described, not named.

02Live demos

Don't take our word for it. Try it.

Live demo · built by Valtrun

The Quote Desk

A freight quote request answered in seconds: read, priced from a rate card, replied to, checked, and followed up.

What it shows

Pick a shipper's enquiry or write your own. The AI reads it and drafts the reply; ordinary rules set the price, ask for anything missing, and check the reply before a person approves it.

Try it →
Live demo · built by Valtrun

The Paperwork Desk

Freight paperwork read by AI, checked against a TMS, and either posted automatically or queued for a person.

What it shows

Pick a rate confirmation, bill of lading or carrier invoice — or upload your own. The AI reads it live; ordinary rules match it to a load, catch unagreed charges and shortages, and log every step.

Try it →
Live demo · built by Valtrun

Ask the Handbook

A support assistant that answers only from one business's own policy document, and says so when the answer is not there.

What it shows

Edit the document, then ask it anything. It answers from the source and declines to guess — the same approach we use for customer and staff assistants.

Try it →
03Example scenarios

How we would approach it.

These are illustrations of common problems, not past client work. Each is clearly labelled.

01Document processing

The broker drowning in rate confirmations

Example scenario · illustrative

The situation

A small freight brokerage receives 60+ rate confirmations a week as PDFs. Each one is retyped into the TMS and matched to its load by hand, and billing errors slip through at month-end.

What we'd build

Incoming PDFs are read automatically, every field is checked against the load, and the result is entered in the TMS. Only mismatches go to a person, with the discrepancy highlighted.

Document AI · TMS integration · Exceptions queue · Audit trail

Likely impact

Loads billed without retyping, billing errors caught before month-end, and 6–8 hours a week back.

Illustrative estimate.

02Business automation

The contractor who quotes at 11pm

Example scenario · illustrative

The situation

A home-services business sends 15–20 quotes a week, each built by hand in the evening: line items copied in, prices looked up, a PDF formatted, the signature chased, then everything typed again into the accounting system.

What we'd build

One short form produces a branded quote and sends it for e-signature. When the customer signs, the invoice and the job record are created on their own, and a reminder goes out if a quote goes quiet.

Quote generation · E-signature · Accounting integration · Follow-up reminders

Likely impact

Quotes out the same day and fewer jobs lost to a faster competitor, with about 8 hours a week back.

Illustrative estimate.

03AI assistants

The same twenty questions, all day

Example scenario · illustrative

The situation

A clinic or small retailer answers 40–60 routine questions a day: opening hours, returns, stock, booking changes. A generic chatbot made it worse, because it invented answers it did not know.

What we'd build

An assistant that answers only from the business's own policies, catalogue and hours. When it is unsure, it hands over to a person instead of guessing, and every question is logged so the team can see what customers really ask.

Grounded AI answers · Website chat · Email · Hand-off to staff

Likely impact

Answers at any hour with nothing made up, so fewer customers go elsewhere, and about 8 hours a week back for the front desk.

Illustrative estimate.

04Custom tools

The business run on twelve spreadsheets

Example scenario · illustrative

The situation

A growing service business tracks jobs, staff and clients across a dozen shared spreadsheets. Nobody is sure which version is current, two people edit the same row, and the Friday report takes someone three hours to copy together.

What we'd build

One internal app built around how the team already works: every job and its status, who is on it, the client's history and this week at a glance, with each person seeing only what they need.

Internal app · Shared database · Role-based access · Live reporting

Likely impact

One live view of every job, no double bookings, no Friday report to build, and about 6 hours a week back.

Illustrative estimate.

05Connected systems

When every sale means admin in three systems

Example scenario · illustrative

The situation

A small brand takes 100–150 online orders a week, but stock lives in a spreadsheet, fulfilment is manual and US orders need currency and tax fixed by hand. Every order means updating three systems, and items still get oversold.

What we'd build

The store, stock and fulfilment connected, so counts stay right and orders go straight to the warehouse. Cross-border orders arrive with currency and tax already handled.

Store integration · Stock sync · Fulfilment · Multi-currency

Likely impact

No oversold stock, US orders handled like local ones, and about 8 hours a week back.

Illustrative estimate.

06AI assistants

The AI pilot that never left the demo

Example scenario · illustrative

The situation

A corporate team built an impressive AI pilot for supplier questions. Six months later it still runs on sample data, because nobody has connected it to the ERP, passed the security review or agreed who checks its answers. Meanwhile procurement still answers 150–200 supplier questions a week by hand.

What we'd build

One process taken to production: the assistant connected to live ERP data with read-only access, answers tested against real past queries, and a named reviewer for anything it is unsure about. We work with your IT and security teams, not around them.

ERP integration · Security review · Accuracy testing · Human review

Likely impact

The return the pilot promised, about 20 hours a week back for procurement, with accuracy measured and signed off.

Illustrative estimate.

07AI for your team

Everyone uses AI, nobody the same way

Example scenario · illustrative

The situation

Staff at a 30-person firm use free AI chat tools for emails, proposals and social posts. Nobody knows what customer information is being pasted in, the tone changes with every writer, and none of it connects to the CRM.

What we'd build

One company AI workspace connected to the firm's documents, CRM and brand guide, with shared templates for proposals, follow-ups and posts, simple rules on what can be shared, and a short training for every team.

Company AI workspace · CRM connection · Brand voice · Team training

Likely impact

Proposals and follow-ups out faster and on brand, customer data kept safe, and about 10 hours a week back across the team.

Illustrative estimate.

Begin a conversation

Where is work slipping away?

Slow replies, quotes nobody chased, customers gone quiet, work your team dreads. Tell us where it happens. In 30 minutes we will tell you what we would fix first, roughly what it would cost, and whether it is worth doing at all.

Book a free call