AI is everywhere at the moment, but the conversations we have with clients about it are surprisingly down to earth. Nobody is asking whether AI will change the world. They want to know if it can save time, make things more consistent, cut admin, or free their people up for more valuable work.
That’s exactly how we’ve approached it ourselves. Over the last few months, Kinetics Group (OutSource is part of the group) has built and rolled out several AI tools internally. None of them are flashy. Every one of them fixes a real bit of friction for our team or our clients.
Just as importantly, these systems run on AI hardware in our own office. Client information doesn’t leave our environment, so we keep control of security and privacy while still getting the benefits.
Here are three we’re using today.
1. Understanding what matters to each client
In IT, the hard part usually isn’t the technology. It’s understanding what each business actually cares about. For some it’s security. For others it’s reliability, compliance, growth, productivity, cost or reducing risk.
We’ve built an AI-assisted process that organises what we learn in client meetings. Instead of relying on notes, memory and interpretation, it pulls out the themes, priorities and business drivers that come up in conversation.
The result is a much clearer picture of what’s important to each client, so we can match solutions to real needs and spend less time debating technology that doesn’t help. The technology becomes the outcome of the conversation, not where it starts.
2. Turning engineers’ notes into proper records
Most engineers are brilliant problem solvers. Very few got into IT because they love writing documentation! Timesheets and job notes are often bashed out at the end of a long day. They’re usually accurate, but not always clear, consistent or easy for a client to follow.
AI now tidies these records up before they’re finalised. It restructures the entry, fixes the grammar and describes the work more clearly and consistently. It doesn’t change what happened, it just makes it easier for clients, managers and the engineers themselves to understand later, whether that’s for reporting or for figuring out a recurring problem three months down the track.
3. Spotting tickets that need attention
Service desk work never arrives neatly. It comes in waves: a phone call, then an urgent email, someone grabbing an engineer in the hallway, a quick follow-up. The work gets done, but the ticket doesn’t always get updated as well as it should.
Most tickets are fine. The trouble is that patchy updates make it harder to find the few that really do need attention. So we’ve put a new AI agent into testing that keeps reviewing tickets, looking for signs that an action is missing, incomplete or overdue, and flags them for a person to look at. It’s an extra layer of visibility so real exceptions don’t get lost in the noise.
What’s coming next
24/7 AI-based support. People are different. Plenty of clients like to pick up the phone or send an email, but a growing number prefer something more digital, and we need to look after them too. By the end of this year we’ll have chat support in Teams, on the website and in the PC taskbar, and we’ll open up our user portal to every user at every client (secured by their Microsoft 365 login). Users will be able to raise a ticket, and in some cases it will suggest a fix or run a tool to solve the problem automatically.
We’re even experimenting with a voice bot for when all our humans are busy on other calls. (To be very clear, this is NOT about replacing people. We LOVE our people! It’s about seeing what’s possible.)
The next ticket health agent. The next version is almost ready. When it finds follow-up work, it will automatically book time in an engineer’s calendar to do it. AI supporting good discipline, not replacing people.
How we’re building it
AI has also changed how our developers work. In May, our development team moved to a workflow built around Anthropic’s Claude Code and a process called “Brainstorming”. Before any code is written, every change goes through three stages:
- building the specification
- planning the implementation
- development and execution
In the first stage the developer and the AI work through the problem together. The AI looks at the relevant code, suggests alternatives, flags risks and keeps asking questions until the requirements are clear. Then it writes a more detailed plan covering the architecture and technical decisions. Only after that does the work go to implementation agents, which write the code under the direction of the planning model.
It sounds like a lot of extra effort (and tokens). In practice it’s the opposite. Giving the AI several chances to challenge assumptions means problems get found much earlier. The specs and plans are saved with the code too, so there’s a clear record of what the humans and the AI decided. And because planning and building use different models with different strengths, mistakes and blind spots tend to get caught before anything reaches production.
For a two-person development team, the difference has been huge. We recently merged two legacy applications into one platform, functionality and data. By June, only our second month working this way, we’d gone from about 10 releases a month to more than 30, including:
- a refreshed quote builder
- feature flags for controlled roll-outs
- data migration and export tools
- discovery interviews and report generation
- compliance assessments and bulk management tools
- hardening of the application infrastructure
For us, AI hasn’t been about replacing anyone. It’s about removing friction, being more consistent, catching risks earlier and letting skilled people spend more time on the work that really matters.
Wondering where AI could take the friction out of your business? Book a free health check, or call us on 0800 688 768.