A B2B portal on BigCommerce with 120 million live SAP prices
Pedigree Wholesale's trade portal, PetProducts.co.uk, rebuilt on BigCommerce B2B Edition in three months, with real-time SAP pricing, restricted catalogues and multi-user accounts.
AI & Data · Build
Automation that does the work, with agents where they earn it.
Quotes, reorders, answers and exceptions handled across your storefront, portal, ERP and CRM. Most of it is a well-designed workflow; where the path can't be known in advance, an agent, inside limits enforced in code, with a person deciding what matters.
The right level for each job
Four levels of AI, chosen per job.
Not a ladder to climb. The question for each job is who decides, and we choose the simplest level that does it well.
Copilot
AI suggests; a person decides. Drafting a quote reply, summarising an account, writing product copy.
Assistant
AI carries out a defined workflow; a person instructs. Most business AI that works in production is this.
Agent
AI plans and carries out the steps itself, within boundaries we set in code. For work whose path can't be drawn in advance.
Agentic system
Several agents sharing the work under one policy, each with its own job and limits. For complex operations that run across teams and systems.
What we build
Automation for commerce, service and operations.
Trade reorders and quotes
Repeat orders and quotes built from account pricing, stock and credit, proposed to the customer or your team to approve.
Customer service
Answers grounded in your policies, orders and the customer's own account, handing over to a person when unsure.
Back-office automation
Order exceptions, product data, returns and invoice matching. The model reads; your systems post.
Integration-first
Connected to SAP, Sage, Microsoft Dynamics, NetSuite, Merlin and the rest of your estate, with only the access each job needs.
In control
Limits in code, not in the prompt.
A prompt can describe the limits. Code is what enforces them. Every system we build has all six.
Its own identity
Its own account in each system, with only the access the job needs, so every action is traceable to it.
Approval where it matters
Anything that commits money, stock or a promise is proposed first. The assistant proposes; the person commits.
Confidence on every output
Three routes, not two: act, show a person with the confidence visible, or escalate. You set the thresholds.
A budget and a stop
A limit on steps, time and cost, and a stop condition the model doesn't control.
A full record
Every input, decision and action logged, so you can see what it did and why.
An off switch
A named owner who can stop it at once, a tested way to do it, and a rule for when.
Our work
Integrated systems we have built.
Automation is only as good as its connection to your systems: live ERP pricing, stock and accounts.
Trade ecommerce on BigCommerce, integrated with NetSuite
A B2B BigCommerce store for Videcon, the trade-only security distributor, integrated with its new NetSuite ERP, with customer pricing, account payments and product kits, promoted with SEO and PPC.
A secure customer portal that cut helpdesk calls by 78%
A bespoke, fully secure car hire and repair portal for Northgate Insurance Services (formerly Auxillis), part of ZIGUP plc, integrated with its case management system and built to keep working when that system is down.
How we work
Plan. Build. Govern.
Plan
Where AI pays, and in what order. For a whole business, the PureNet Group's Converge Methodology; for a single area, a fixed-scope assessment.
Build
A pilot with a number to move, an owner and a way to stop it, then production in your storefront, portal and back office, integrated with your ERP and CRM.
Govern
Govern the use, not the technology. Your own AI on business terms, with an owner, a one-page policy and an inventory; and every system we build for you tested on every change, logged, overseen by a person and fitted with an off switch, under our ISO 27001 controls.
Questions
What people ask us.
Should we have an agent or a workflow?
It depends on the job, and most systems use both. We start with a workflow wherever the steps can be drawn in advance, because it's faster and more predictable, then add agents for the parts that need judgement, where every case is a little different. What is an AI agent?
What stops it doing something it shouldn't?
Limits enforced in code, not described in a prompt. It has its own access and only what the job needs, approval before anything that commits money, stock or a promise, a budget, a stop condition and an off switch. That's also what protects it from prompt injection.
What happens when it isn't sure?
It says so. Every output carries a confidence, and below the threshold you set it shows a person or escalates. Escalation is a path we design, with someone reading it, not an exception.
Will it replace our team?
It takes routine work off them. People stay in the loop for the decisions that matter, and the time saved is yours to decide what to do with.
Which systems can it work with?
The ones you already run. We have integrated SAP, Sage, Microsoft Dynamics, NetSuite, Merlin and many others, and connect each job to only the data and actions it needs.