We had a problem that probably sounds familiar to every SME owner reading this. Despite being a Cyber Essentials certified technology company, with partnerships including Dell Technologies, HPE, Lenovo and Cisco, we were drowning in routine operational tasks. Sales emails sat unanswered for hours, social media posts needed constant attention, and our team spent more time on administration than on work for clients.
So we built something about it. Not a simple chatbot or a workflow automation, but a complete AI-powered operations platform that we now run the business on. We call it Apaxon Ops, and it has changed how we work as a Manchester-based IT services and hardware company.
Meet the Apaxon Ops AI team
Apaxon Ops isn't a single AI tool. It's a multi-agent platform where each AI agent has a name, a defined role and specific areas of expertise. Think of it as an operations team that works around the clock, with a person checking its work before it reaches a customer.
Victor Apaxon is our CEO Agent, the strategic triage point for everything that comes into the business. When emails arrive, Victor reads them, assesses their commercial priority and routes them to the right specialist agent. He knows our vendor portfolio, including our status as a Dell Technologies Authorised Partner, with Dell infrastructure supplied through Arrow Electronics. Every morning Victor drafts a business briefing, and every week a strategic summary that helps us spot patterns we would otherwise miss.
Alex Carter handles inbound sales email at sales@apaxon.co.uk. She has become good at spotting spam and scam emails, scoring each message for legitimacy and checking product availability across our store catalogue. When a genuine enquiry comes in, Alex builds an itemised quote with VAT and applies our free delivery on orders over £100. She flags end-of-life products and recognises returning customers, and there is one rule she never breaks: the store price is the price.
Riley Shaw runs our marketing operations, and drafted the first version of this post. Riley handles our LinkedIn and Facebook content, writes SEO blog posts on a rotating 27-topic schedule, sends a Monday morning marketing digest and keeps an eye on competitor activity. She owns Apaxon's content voice and keeps everything aligned with our positioning as a serious technology partner rather than just another IT reseller.
Rounding out the core team, Sam Riley looks after customer support, including the live chat on Apaxon Limited, and helpdesk escalations, while Jordan Blake handles invoicing, payment chasing and financial reporting. Each agent works in its own area, but they all run within the same framework.
Three more agents work behind the scenes: Aria, a personal assistant; Dev, our lead developer agent; and Scout, which enriches the product data in our store catalogue.
The human approval layer changes everything
Here's what makes Apaxon Ops different from typical business automation: the work the agents prepare waits for a person before it goes anywhere. When Alex drafts a quote, it sits in a queue for review. When Riley creates a LinkedIn post, it waits for sign-off. When Victor identifies a high-priority strategic issue, it goes to a human decision-maker.
The exceptions are deliberate and narrow. Sam answers questions in the website chat in real time, because a visitor asking about delivery or a product's specification shouldn't wait for a queue. Sam can explain products and policies and take a visitor's details, but cannot agree a price, issue a refund or authorise a return; those stay with the team. And Sentinel, below, can restart a stuck background task on its own.
This approval layer isn't a bottleneck. It's the reason the system works. We review AI-prepared work every day, but we're making decisions on finished drafts rather than starting from a blank page. The agents handle research, analysis, drafting and formatting; people handle judgement, strategy and the final say.
The result is a business where routine work is handled consistently and professionally, while control stays with people. An agent cannot send a sales email, publish a post or make a commercial decision on its own.
Sentinel keeps the platform running
We learned quickly that an AI operations platform needs its own security and maintenance layer. Sentinel monitors the platform itself, watching for brute-force login attempts, unusual out-of-hours access, and sudden spikes in failed orders or payments. It also keeps the platform healthy by automatically resetting tasks that get stuck in processing loops.
When Sentinel sees the same problem repeat, it escalates to a person. That keeps the platform running when an individual component has a problem, and gives us early warning of anything systemic before it affects clients.
A virtual board
Apaxon Ops also has a virtual advisory board: five AI personas, each with a different brief. Every week Victor prepares a board pack, and each member responds the way a board would, with concerns, recommendations and a view on how the business is going. They are not directors and they don't make decisions. They are a structured way of stress-testing the business from angles a founder can easily miss.
Margaret Holloway, Non-Executive Chair: governance and strategy, and holding the founder to account.
David Okafor, Finance & Risk: cash flow, margins, VAT and overdue debt.
Priya Sharma, Commercial & Growth: revenue, pipeline, pricing and competitive position.
James Whitfield, Technology & Operations: platform stability, service delivery and security.
Claudette Bernard, People & Culture: a non-voting observer who asks whether the business is sustainable for the people running it.
What we've learned running AI operations
The biggest change is that nothing waits for someone to find time to start it. A sales enquiry has a drafted reply ready for review instead of sitting in an inbox, our social media posting is consistent and on-brand, financial reporting runs automatically, and our team spends its time on client work rather than administration.
We've also learned that AI operations need careful design. The approval layer isn't optional; it's what makes the system trustworthy. Each agent needs clearly defined boundaries and a clear route for escalating what it can't handle. And monitoring the platform isn't an afterthought; it's essential to keeping the service reliable.
The road ahead
We're continuing to develop Apaxon Ops, including deeper store integration so that Alex can handle inventory alerts and product lifecycle notifications. We're also exploring whether other businesses could run their own AI operations teams on the same platform, particularly IT managed service providers and other professional services firms that handle a steady flow of client communications, quotes and administration.
Apaxon Ops is what we believe AI in business should look like: capable, reliable and firmly under human control. It isn't about replacing human judgement but supporting it with consistent, round-the-clock operational help. For a Manchester IT company handling everything from Cyber Essentials certification to enterprise hardware procurement, it has proven to be exactly what we needed.
If you run an SME or an IT service provider and you're curious what AI-powered operations could look like for your business, we'd be interested to hear from you. The problems we solved probably aren't unique to Apaxon. IT Strategy.
Get the latest IT insights
Expert tips, infrastructure news, and product updates — delivered weekly.
