.svg){kind=logo}
Research & Insights
How our engineers built 7 AI Tools that save thousands of hours
Inside Lucidya's Quality of Work Initiative: how six engineers built seven AI tools in six months and saved an estimated 16% of engineering capacity.
Inside Lucidya's Quality of Work Initiative: how six engineers built seven AI tools in six months and saved an estimated 16% of engineering capacity.
Between December 2025 and May 2026, a six-person team at Lucidya built the Quality of Work Initiative: seven AI-powered tools that changed how our engineers find knowledge, document code, review security, onboard new joiners, and track delivery.
The early results: documentation that used to take a full day now takes five to ten minutes, security review of deployment plans went from a manual step to an automated gate, and the team's launch estimate puts the system at around 2,500 engineering hours saved per month per 100 engineers, roughly 16% of engineering capacity back to real work.
Lucidya builds AI products for enterprises across the region. This is the story of what happened when we pointed that same discipline at ourselves, told through the words of the people who built it. It includes the parts most companies leave out: the hallucinations, the features that got cut, and one detail about this team that changes how you will read every number in this article.
.png)
Every growing tech company knows this scene.
An engineer needs to understand how a service deploys. The answer exists, but it lives in a Confluence page written two years ago, a video someone recorded once, a Google Doc nobody can find, and the head of one senior engineer who is in back-to-back meetings.
So the engineer searches. Asks. Waits. Searches again. An hour disappears. Multiply that hour across every engineer, every week, and you get a quiet tax that most engineering organizations pay without ever putting a number on it.
At Lucidya, the tax showed up in four ways: knowledge scattered across tools, time lost to repeated questions, dependency on a few key people who held the answers, and blind spots for leaders trying to understand delivery health.
"Engineers were spending time trying to understand how a service works, where the documentation is, what the deployment flow looks like, or who has the right answer. This is normal in engineering teams, but over time it becomes expensive."
– Alaa Alharthi, AI Engineer and Lead of the QoW Initiative
The idea came from inside the engineering team. Alaa Alharthi, our AI Engineer, went to Lucidya's CTO, and asked to take on work beyond her daily engineering tasks, something that would improve how the whole engineering team works, not just one project.
At the same time, Lucidya had accepted a group of university students who wanted to build their graduation project inside a real company. They needed a serious project. Alaa's idea needed a team. The CTO put them together.
The group started with a long, unorganized list of ideas about automation and engineering workflows. Their first weeks went into narrowing that list down to what would actually save engineers time. By December 2025, the plan was set, and the team of six was formed:
Ameera Milibari - Technical AI Manager, Alaa Alharthi - AI Engineer, Abdullah Alameer - SRE intern, Abdulrahman Alassaf - DevOps Engineer Intern, Hassan Barmandah - Software Engineer Intern, and Yazan Alkamal - AI Security Engineer Intern.
The goal fit in one line: build better, track smarter, reduce engineering effort.
The deadline: six months.
Tools for documentation, knowledge search, and code review exist off the shelf, so the obvious question is why build at all. The team's answer came down to one word: context.
Lucidya has its own services, repositories, deployment flows, documentation style, and engineering habits. A generic tool would need heavy customization before it became useful, and it would still not know the codebase the way an internal system can.
Building in-house also meant the team could start small, test with real engineers, and improve from real feedback every week.
"If we build AI systems for others, we should also use AI to improve our own engineering experience."
– Alaa Alharthi
By May 2026, the initiative had shipped seven capabilities that work together as one system:
Every project has a moment when it stops being a plan and becomes real. For this team, it was the first time Doc Agent documented an actual repository. A task that normally takes hours, sometimes a full day depending on the repo, was finished in five to ten minutes.
The first reactions came in two languages: "fantastic" and "تمزحين؟ أسطوري", which roughly translates to "Are you joking? Legendary."
The speed was impressive. The team's next step was making sure the quality matched it.
"At the beginning, we saw a lot of AI hallucinations. Sometimes the text looked clean and confident, but the details were not fully accurate. A repo is not just files. It has structure, logic, dependencies, naming, hidden assumptions."
– Alaa Alharthi
Behind that caution sat the hardest design question of the whole project: how much should the AI actually do? Should it only summarize? Suggest? Generate full documentation? Flag risks? The team kept coming back to one rule.
The breakthrough was not a clever model trick. It was a decision about roles. Doc Agent became valuable the day the team started treating it as a strong first draft that engineers review, rather than a final source of truth. The AI removes the blank page. The engineer keeps the judgment. Documentation went from a writer's job to a reviewer's job, and that is how the day became ten minutes.
The team set one rule at the start: ship what saves engineers time now, stage everything else in order. The roadmap was not an afterthought. It was part of the plan.
First in line is the most significant addition: a full agent version of LuQua that executes tasks, not just answers questions. The team designed it from day one and scheduled it for phase two deliberately. A tool that acts autonomously on an engineer's behalf needs systems it can rely on. Those systems now exist, tested in daily production use across the entire engineering team.
The rest of the system is advancing in parallel. The Deployment Hub is rolling out across teams. The KPI dashboard is being updated with real usage data so its metrics reflect the full context of an engineer's work, not just raw activity counts.
Phase one shipped. Phase two is already scoped.
The initiative launched recently at full scale, so the team is careful with its numbers, and this article is too. The estimates above will be measured against real data in the months ahead. What is already measured: the Security Bot is a required step in every deployment approval, leadership uses the bots daily, and the bot has already caught vulnerabilities in pipelines and security checks that manual review missed.
The behavioral proof is already in.
Anand Saurabh, Lucidya's VP of Engineering, and Mostafa Asaad, Security Delivery Manager, use the bots daily to approve deployment plans.
And the AI Security Bot has crossed the line from helpful tool to required step: if a deployment plan did not go through the bot, the security team will not approve it. The bot has already caught vulnerabilities and mistakes in pipelines and security checks that manual review could have missed.
“Seeing our VP of Engineering and our Security Delivery Manager using our bots daily to approve deployment plans is the best way to describe the success of this project."
– Yazan Alkamal
Zoom out from Lucidya for a moment, because this story is landing at a very specific point in time.
2026 is the year AI agents stopped being a curiosity and became a race.
And shipping is where things break.
Gartner places the technology at the peak of inflated expectations, noting that most deployments remain narrowly scoped and that fully autonomous agents are not ready for the majority of enterprise use cases.
Read those numbers next to this story, and the lesson writes itself.
The Quality of Work Initiative sits on the right side of every one of those statistics, for three reasons:
The team did not build AI tools because agents are the moment. They built them because engineers were losing hours to scattered knowledge, and they could point to exactly where.
While one in five companies has mature oversight for autonomous agents, this team built oversight into the architecture from day one. Doc Agent produces a first draft that engineers review. The Security Bot feeds a human security team that makes the final call. Nothing runs unsupervised, which is precisely why leadership trusts it enough to use it daily.
The team cut the most exciting feature, a fully autonomous agent, because it did not fit the timeline. In a year when inflated expectations are the defining risk, knowing what not to build is the skill that separates the 17% who deployed from the 60% still planning.
Timing is the last piece. Right now, most organizations are still planning their first AI agents, which means the teams already running them are learning faster than everyone else: what to measure, when a human needs to review, how narrow a tool should be. Those lessons take months of real use to learn, and they cannot be bought or copied later. That is the real advantage of starting now.
The onboarding platform has a personal origin. When the students joined Lucidya, their own first week felt like being lost in a maze: resources scattered across Confluence, videos, docs, and drives, with no map.
"We felt like we were missing something. There were a lot of resources across Confluence, YouTube, other websites, Google Docs, and Google Drive. We felt like we were lost in a maze."
– Abdullah Alameer
So they built the map they wished they had. A new engineer at Lucidya now gets every learning path and resource in one place, plus an assigned peer buddy who tracks their progress and helps them connect with the team faster. The tool exists because its builders refused to let the next person feel what they felt.
Here is the detail we promised at the start, the one that changes how you read every number in this article. Most of the people who built these seven tools were university students. This was their graduation project, built inside a real company, shipped to real users, while they were still studying, guided by the two people Abdullah calls "our great leaders," Ameera and Alaa.
They worked through midterms. They worked on weekends. They delivered the project during their final exams.
"We would be lying if we said we didn't feel pressure. We did, but we loved what we were working on and what we were building."
– Abdullah Alameer
Read the estimate again with this in mind. Thousands of engineering hours per month, projected to be saved by a system that a handful of students and their mentors built in six months, at a company confident enough to let them try.
The Quality of Work Initiative is still growing. The Deployment Hub is expanding through team trials, the KPI system is being refined toward fairer, more context-aware metrics, and the postponed agent version of LuQua is waiting for its turn.
For any engineering team looking at this and wondering if they could do the same, the team's answer is simple:
"Just start what you want to do, and you will find a way to do it."
– Abdullah Alameer
The core shift has already happened. The same discipline Lucidya applies to the AI it builds for its customers now runs its own engineering floor. And the people who built it were the newest people in the building.
That's a win worth telling.
.svg)
.svg)
%20with%20competitive%20intelligence_640x774.avif)
