Can a small development team use AI to build more complete features, faster, without compromising quality or giving up human control? That is one of the questions we are exploring at uQualio.
Like many SaaS companies, we have something that never seems to get smaller: our development roadmap.
There are always improvements we want to make, new functionality our customers can benefit from, and ideas we would love to turn into reality. But we are also a lean team. And when you are building a SaaS company with a small development team, every senior engineering hour matters.
So in 2026, we decided to experiment with a different approach to software development.
Together with Clever Spark, we implemented ADLC (an AI-assisted development workflow) designed to support the journey from an initial ticket to a pull request.
And we are already learning some interesting things.
AI isn’t only useful for writing code
When people talk about AI and software development, the conversation often focuses on one thing:
Can AI write code faster?
Our experience suggests there is another part of development where AI can be extremely valuable: before the coding starts.
A feature request can look relatively straightforward when it enters the development pipeline. Then development begins.
Questions appear. Requirements need clarification. Assumptions that nobody realized they were making suddenly need decisions. And work stops while those questions are resolved.
With ADLC, AI is brought into the refinement stage.
It analyzes the ticket together with relevant documents from our repositories, asks questions and helps make implicit assumptions explicit.
As our co-founder Christian Bjerre Nielsen puts it:
“The refinement stage is really what is lifting it.”
Finding those questions before development begins can make the actual build much more coherent.
From several tickets to one complete feature
One of the most interesting changes we’ve seen is the ability to approach larger pieces of functionality as a whole. Work that might normally have been divided into two to five separate tickets can, in some cases, be developed as one coherent feature.
A good example is our new magic-link feature developed at uQualio. Rather than breaking the functionality into numerous smaller development tasks, the feature was built in one pass. Even better, its core can be reused as we develop the next part of the product.
Christian describes the difference this way:
“It has made it possible to create a full feature without having to necessarily understand all the things before you start refining and building.”
For a small SaaS team, that is exciting. Not because AI suddenly replaces the development process, but because it can help us make better use of the development capacity we already have.
Humans still make the decisions
This part is important.
Our goal isn’t autonomous software development where AI decides what to build and what to put into production.
Humans remain in control.
ADLC includes review gates at defined points in the workflow. AI can analyze, suggest, build, and do much of the heavy lifting, but a person remains responsible for the important decisions. Nothing gets merged simply because the AI thinks it is ready.
Interestingly, we’ve also discovered that AI-assisted development doesn’t mean one developer can suddenly supervise an unlimited number of tasks.
In our experience so far, two parallel work items per person appear to be the practical optimum.
Why?
Because even when AI can accelerate parts of the work, human attention and review capacity still matter.
That may be one of the most important lessons businesses should remember as they introduce AI:
AI can increase our capacity, but human judgment remains valuable.
AI Learns & Improves development all the time
One of the best things findings have also been that as part of the process it has a memory.
The AI agent systems keeps picking up all the improvements and learning we make. This ensures that all problems, improvements and learnings are not forgotten.
They instead becomes a memory base to be continuously used to ensure that our development improves and problems are avoided
Adapt AI to the business, not the business to AI
There was another important principle behind the project.
We didn’t want to reorganize our entire development process around a new AI tool.
Instead, Clever Spark adapted the workflow around uQualio’s existing project, technology stack, conventions and way of working. That distinction matters.
AI should solve problems for your business. Your business shouldn’t have to create new problems just to accommodate AI.
The implementation also gave Clever Spark new learnings about introducing its workflow into an existing codebase and domain. Their takeaway was simple:
Adapt the workflow to the existing project, not the project to the workflow.
We think that is a useful principle far beyond software development.
Now we’re measuring the results
We’re excited by what we have seen so far, but we’re also careful about declaring victory based on impressions alone.
The agreed implementation is now complete, and our team is using the workflow in daily development.
The next step is to measure what actually happens.
Over the coming months, we are tracking:
- Work in progress
- Throughput
- Cycle time
- Adoption
- Quality
The hypothesis is that we can increase throughput and complete more work while maintaining or improving quality.
But that’s a hypothesis. The data will tell us whether we’re right. And we think that is an important part of adopting AI responsibly.
Small team. Big ambitions.
At uQualio, we know a little about using technology to make knowledge scale. It is, after all, one of the reasons we built uQualio.
Our video learning platform helps businesses capture knowledge once and turn it into structured, engaging learning that can reach employees, customers, and other audiences at scale.
Now we’re exploring a similar question inside our own organization: How can AI help the knowledge and experience of a small team go further?
We’re still learning. We’re still measuring. And humans are still very much part of the process.
But if AI can help a small SaaS team spend less time discovering problems halfway through development and more time building complete, high-quality features, that’s an experiment worth pursuing.
We look forward to sharing what the numbers tell us.
Read Clever Spark’s case study about the uQualio implementation: https://cleverspark.co/case-studies/uqualio-adlc/
