)
)
TL;DR
AI is everywhere right now. You can’t open LinkedIn without someone on their soapbox declaring we’ve either entered a golden age of productivity… or the end of civilisation.
At Cube, we’re taking a calmer, more measured approach to AI adoption. From our builds, to our clients, to internal workflows, the last three years we’ve been honing in on what an AI-first software agency looks like when it operates responsibly in GenAI.
To make our internal processes slick, we’re building Custom GPTs that help with the parts of work that are genuinely sticky, repetitive, or hard to learn fast. Not to replace people. Not to “AI-wash” our process. Just to make the day-to-day smoother and the output more consistent.
So, what does that look like in practice?
It looks like our Product crew building a bot that understands how we actually deliver digital products. It looks like Marketing building a persona GPT that helps us speak to real audiences like real humans. And it looks like QA building a test case generator that turns a feature brief into structured coverage in seconds.
Let’s start with the one that’s made the biggest dent so far.
Product: Meet the Product Bot
Our Head of Product, Guy, created a custom GPT trained on the full digital product lifecycle, from scoping and stakeholder management through to delivery and post-launch support.
It’s built on extensive product methodology, technical business analysis practices, and proven external best practice. But the important bit is what it feels like to use.
If you’ve ever worked in Product, you’ll know the job is basically a long series of judgement calls. And for junior Product Owners, the hard part isn’t effort. It’s experience.
The Product Bot helps close that gap.
Instead of hunting through docs, trying to remember where that one “how we do X” thing lives, or interrupting a senior teammate mid-flow, a junior PO can ask the bot and get structured guidance immediately. It’s like having a calm Product brain on tap, ready to sanity-check your thinking and point you in the right direction.
Where it’s already helping
Product work is rarely neat. The bot is most useful in the real-life moments where things get… chaotic.
It can:
help you handle difficult stakeholders by grounding conversations in evidence, feasibility, and the product north star
break down technical or strategic ideas like VMOSA and business model context when you need to explain them clearly
clarify ownership across the Product Owner, Project Manager, and Client Success roles (which can be tricky at the best of times)
What makes it genuinely valuable is that it’s not replying like a generic “product consultant”. It’s rooted in how Cube actually delivers products.
What we learned: the best “AI productivity win” is fewer Slack pings mid-deep-work.
Onboarding that doesn’t rely on osmosis
One of the most exciting uses is onboarding.
Before this, new Product Owners leaned on job descriptions and scattered guidance. Now they’ve got something closer to a living playbook, available in the moment. They can ask questions as they work, not just when they remember to ask them.
It can also validate documents like Product Requirements Documents or technical investigations, checking whether anything important is missing before work goes anywhere near a client.
That’s the kind of AI support we care about: practical, specific, and helpful at exactly the point you need it.
Why clients feel the impact too
At Cube, Product Owners often act as proxy product owners. They safeguard the build until handover and reduce stakeholder chaos for clients. It’s a role that’s valued because it keeps decisions grounded in strategy and evidence, not just opinions and urgency.
Tools like the Product Bot strengthen that protection. Better internal clarity tends to show up as better client outcomes. Quietly, but consistently.
Marketing: Customer Persona GPT
On the Marketing side, we’ve built a Customer Persona GPT that helps us get out of our own heads and into our customers’ shoes.
It supports Sales in tailoring conversations to real client pain points. It helps Client Success build stronger long-term relationships. And it helps Marketing shape segmentation, messaging and campaign direction.
What it helps us do
tailor conversations to what people actually care about (not what we think they care about)
build stronger continuity across the full relationship, not just “campaign moments”
keep messaging specific and relevant, so it feels human and not… generated
The goal isn’t to churn out more content. It’s to make the work more specific and relevant, so the people reading it feel like we actually understand what they’re dealing with.
What we learned: Persona work is empathy, not templates.
QA: Test Case Generator GPT
QA has built a custom GPT that takes structured feature briefs and transforms them into fully formatted, production-ready test cases in CSV format.
It expands scenarios, adds missing negative and edge cases, includes regression, smoke, and accessibility coverage, and makes sure every step has a clear expected result. The output is flat, machine-ready, and compatible with test management and automation workflows.
What it produces (fast)
standardised QA documentation across projects
stronger coverage, including negative, edge, regression, smoke and accessibility testing
less manual formatting and repetitive expansion work
test case creation reduced from hours to seconds
It’s a perfect example of AI doing the boring bits so specialists can focus on quality.
What we learned: Speed is useless if quality drops, so the bar stays high.
The bigger picture
We’re not chasing AI trends. We’re shaping AI around the way good teams already work.
Less friction. More clarity. Better consistency. More time for the work that actually matters.
Because the point isn’t “smarter tools”. It’s smarter humans, supported at the moment they need it.
Want to try this in your own organisation?
If you’re exploring how Custom GPTs could work inside your organisation, the best place to start is safe, structured experimentation.
That’s exactly why we created our free AI Readiness Toolkit. It walks you through a simple six-step process to test real use cases, manage risk responsibly, and move from curiosity to confident adoption.
Need help with your digital product idea? Holla!
Published on 23 February 2026, last updated on 23 February 2026