We help commerce
and supply chain businesses.

A London consultancy building AI agents that work inside the software you already run, even systems built just for you.

Book a conversation See what we do

We build inside the systems you already run, off the shelf or custom:

  • Shopify
  • WooCommerce
  • Algolia
  • Sage
  • Xero
  • SAP
  • Zendesk
  • WhatsApp

What we do

Three ways we put AI to work in a trading or retail business. Most clients start with one.

Pale concrete warehouse aisle: a tipped-over heap of plain cartons on one side, the same cartons squared away on the racking opposite, one of them cobalt Agents

AI agents for your business

Agents that do the work your people retype by hand: reading the email, drafting the quote, updating the system, flagging the exception. They run inside the software you already have, off the shelf or built just for you.

View examples
A line of identical pale concrete monoliths on an open plaza, one of them cobalt and lit by a single shaft of sunlight Visibility

A website ChatGPT recommends

Your customers now ask ChatGPT, Claude and Perplexity who to buy from. We rebuild your site so you are visible and citable when they do, starting with what those answers say about you today.

Get a free website audit
Pale concrete facade assembled from precast blocks of different sizes fitted together, every joint between them a cobalt seam Software

Custom software, built AI-fast

The tools and systems your business actually needs, at a fixed price and owned outright by you. We build with AI, which is why we take on work that used to be uneconomic.

Book a conversation

How AI agents can help you

Four agents we have designed for merchants and retailers. Each one runs inside the software the business already had, and each takes over work that used to sit on a person's desk. Open one to see how it plays out.

Quote Agent

Reads the request, prices it against live stock, drafts it for approval.

Runs in Sage, Xero, SAP and a shared inbox

Wall of pale concrete pigeonholes holding squared stacks of paper, one compartment lined in cobalt

Aftersales Agent

Handles order status, claims, returns and refunds end to end.

Runs in Zendesk, Shopify and WhatsApp

Row of identical pale parcels on a concrete bench with one lifted out onto a cobalt plinth

Catalogue Agent

Turns supplier datasheets into product data customers can actually search.

Runs in Algolia, WooCommerce and your PIM

Deep grid of pale concrete shelving filled with identical white boxes, one cobalt box pulled forward

Counter Agent

The advice your best people give, answering on your site at nine at night.

Runs on your site, WhatsApp and Zendesk

Pale concrete trade counter with a cobalt front panel and a blank tablet lying on the top

Every engagement follows the same pattern: one process, one system, one number that moves. Something else eating your margin, such as ad operations, supplier onboarding or credit control? We build custom AI agents for those too. Start with an audit.

Where it pays for itself

What published industry studies find when AI is put to work in each sector.

Wholesale goods-in yard under a concrete canopy, rows of wrapped pallets with one wrapped in cobalt

Manufacturing & wholesale

  • Order-to-ship lead times cut by up to 77%, with half the inventory 1
  • Defects down by up to 71%, with raw material costs cut by a fifth 2
  • Tariff codes right 95% of the time, and export paperwork 60% faster 3
  • Up to 30% less stock on hand, with fill rates up to 8% higher 4
Calm retail interior in pale concrete, shelving runs of plain cartons with one run stocked in cobalt

Retail

  • AI agents resolving 80% of common customer service issues on their own, at 30% lower cost 5
  • Lost sales from empty shelves cut by up to 65%, forecast errors by up to half
  • Product and marketing copy written 40% faster, and 18% better
  • Half of all service cases handled by AI by 2027 6
Sources
  1. World Economic Forum, Global Lighthouse Network, June 2026
  2. World Economic Forum, Global Lighthouse Network, January 2026
  3. McKinsey, How generative AI is disrupting distribution, 2024
  4. McKinsey, AI in distribution operations, 2024
  5. Gartner, agentic AI in customer service prediction, 2025
  6. Salesforce, State of Service, 2025

How we work

Audit

We come inside your business, map where hours and margin leak, and rank where AI pays back fastest. You get a plan in plain English, at a fixed price agreed up front.

Pilot

One agent, built and running on your real data. You see it work, and see the number move, before committing to anything more.

Run

We ship it into daily use, train your team and stay on to keep it sharp as your business changes.

  • Fixed price, agreed before we start
  • Built on your systems, owned by you
  • A human signs off until a workflow earns autonomy
  • Your data stays yours, under UK GDPR

About us

Abington brings the calibre of AI that Fortune 500 companies already rely on to the mid-sized businesses that keep commerce and supply chains moving. We are led by two London founders, one from technology and one from commerce, each with nearly two decades of building and running companies.

Nik is a CTO and technology entrepreneur who co-founded an AI company whose platform serves Fortune 500 teams, having begun his career integrating ERP systems at Toyota. Mike has co-founded and led a business since 2013, and brings an operator's command of margin, cash flow and what it takes for a team to adopt something new. Together, they design and build every agent themselves.

Questions owners ask

What does it cost?

Every engagement is fixed-price, agreed before work starts. No day rates, no surprises. The audit is the smallest commitment and tells you exactly what the rest would cost and return.

Do we need to replace our systems?

No. The agents we build sit on top of what you already run, whether that is a full commerce stack and a proper ERP or a shared inbox and a price list in Excel. If your stack genuinely blocks the result, we will say so in the audit.

Our software was built just for us. Can you work with it?

Yes, that is rather the point. Custom systems usually have a database, an API, or at the very least screens and exports. We build AI agents around what your system can already do, and we never ask you to throw it away.

What happens to our data?

It stays yours. We work under UK GDPR, put data agreements in writing before touching anything, and never use your data to train public models.

We're not technical. Who runs this day to day?

We do. Your team approves what the agents produce and handles the cases that need judgement. Nobody on your side needs to learn to code, and we train the people who work alongside them.

What is the first thing we actually see?

The audit gives you a plan in plain English. After that a pilot runs on your own data, so the first thing you see is your real work being done, not a demo. We agree the schedule up front, and if it ever slips you hear it from us first.

What if it doesn't work for us?

The pilot ends with an honest go or no-go, backed by the numbers it produced. If the answer is no, you keep every finding.

Bring us the messiest process you've got.

A 30-minute conversation. We'll tell you honestly whether AI can move a number you care about.

Every booked call comes with a free AEO readiness check: what ChatGPT and Perplexity say when customers ask about your business.

Book a conversation

AI agent

Quote Agent

An AI agent that reads inbound quote requests, whether they arrive by email, web form or phone, prices them against your live stock file and drafts them for one-click approval. Anything it cannot price goes to a human with the reason attached.

Book a conversation How we build it

What it takes off the desk

  • Reads requests from email, web forms and phone notes
  • Prices every line against the live stock and cost file
  • Drafts the quote for one-click approval
  • Hands back anything it cannot price, with the reason

A roofing merchant turning over six million. Every sale starts as a quote: a form, an email, a phone call.

By hand

Someone reads the request, checks the stock, prices it and types it up. Same day, if they're lucky. The counter's best people spend their afternoons quoting instead of selling.

With a Quote Agent

The request is read, priced from the live stock file and drafted in minutes. The counter team taps approve. Nobody is hired, replaced or sent on a training course.

AI agent

Aftersales Agent

An agent that handles order status, damage claims, returns and refunds end to end, and escalates cleanly the moment a case needs human judgement. It arrives at a person with the file already built rather than a ticket number.

Book a conversation How we build it

What it takes off the desk

  • Answers "where is my order" from live carrier and ERP data
  • Opens damage claims with the photos and paperwork attached
  • Processes returns and refunds inside your own policy
  • Escalates judgement calls with the case already assembled

A distributor shipping four hundred orders a day, and one inbox for all of it.

By hand

Every "where is my order" lands in the same queue as the claim that is about to cost you a customer. The team works down the list in the order it arrived, and the serious cases wait their turn.

With an Aftersales Agent

Routine status, returns and refunds are closed as they arrive, out of hours included. What reaches a person is the short list that genuinely needs one, with the history already gathered.

AI agent

Catalogue Agent

AI that reads supplier datasheets to enrich your product data, powers search that understands what a customer actually means, and gets thousands of SKUs properly listed online rather than sitting in a backlog.

Book a conversation How we build it

What it takes off the desk

  • Reads supplier PDFs and datasheets into structured data
  • Fills the attributes your filters and search depend on
  • Matches what a customer meant, not only what they typed
  • Flags contradictions between supplier and your own data

Forty thousand SKUs on the price list. Nine thousand of them online, and half of those with an empty spec.

By hand

One person copies specs out of supplier PDFs, a few hundred lines a month. Search returns nothing useful, so the customer rings the counter instead, or buys it somewhere else.

With a Catalogue Agent

Datasheets are read into structured attributes at volume, the gaps are filled and search answers the question that was actually asked. The backlog stops being a decade of work.

AI agent

Counter Agent

An agent trained on the advice your best people give over the counter, so that knowledge answers customers on your website and on WhatsApp, and is there for every new starter from their first shift.

Book a conversation How we build it

What it takes off the desk

  • Learns from the answers your counter team already gives
  • Answers on the website and on WhatsApp, out of hours
  • Cites the product, the spec and the stock position
  • Sits beside a new starter instead of a fortnight of shadowing

Thirty years of knowing which fixing suits which substrate, held by four people.

By hand

The advice that closes the sale lives with a handful of long servers. When they are on holiday the answer is slower, and when they retire it leaves the building with them.

With a Counter Agent

That same advice answers a customer at nine at night on the website, and sits beside a new starter on their first shift instead of a fortnight of shadowing.