Why every university deserves its own innovation lab
Written by Sam Serrien ·

A university or university college is a big, slow machine. That is not a criticism: every day it has to serve tens of thousands of students, keep exams running and handle their data with care. But that same care makes it almost impossible to try something quickly. In this article I explain why I believe a small internal lab can make the difference, what other organisations already do and where it often goes wrong.
Stability is the job, and that is the problem
An IT department in education is judged on what does not happen. The learning platform stays up during exams, no student data leaks, the procurement was done by the book. That is exactly what you want from them.
The catch is that a small idea goes through the same mill as a large project. A lecturer wants to test an AI tutor in one course, a programme wants to know whether an app helps students plan. That idea gets an analysis, a budget request, a slot on the roadmap. By the time anything can be built, the academic year is over or the technology has moved on.
What happens next, I know from experience: lecturers build it themselves, in the evening and between classes. That is how the tools I built for education came about too. It works, but it depends on one person, and nobody else in the organisation learns from it.
Shadow IT is a signal, not sabotage
There is a name for this: shadow IT. Technology that staff buy, build or modify themselves, out of sight of the IT department. A free quiz tool, a shared spreadsheet that slowly turned into a database, a script that pulls grades out of the learning platform every week. And lately also an AI chatbot that a lecturer pastes student assignments into.
It is not a fringe phenomenon. Gartner predicts that by 2027, 75% of employees will acquire, modify or create technology outside IT's visibility, up from 41% in 2022.[1] With AI it moves even faster: according to Microsoft and LinkedIn, 78% of people who use AI at work bring their own tools.[2]
I am not innocent here either. Anyone who builds a tool in the evening because the official route takes too long is doing shadow IT, even with the best intentions. And I see what it costs: nobody else knows those tools, and if I stop, they stop too. Still, it almost always starts with good intentions. The UK National Cyber Security Centre writes that shadow IT is rarely malicious, but arises because staff struggle to complete a task with the sanctioned tools or processes. Its advice is therefore not to lock everything down, but to look at the need behind it and offer a simple process for anyone who needs something.[3] Higher education shows the same pattern: lecturers reach for their own tools to move faster, and the best remedy is a supported path that is just as easy.[4]
So the problem is not that lecturers experiment. The problem is that it happens in the dark. Nobody knows which data sits where, who maintains it when the builder leaves, or whether two programmes are building the same thing.
AI makes this more urgent. Shadow IT used to be a spreadsheet or a free tool. Today someone without any programming experience builds a working app with an AI assistant in an afternoon. That is great, but AI lowers the bar to build something, not the bar to build it safely. Veracode had more than a hundred AI models complete programming tasks: in 45% of the cases the code contained a security flaw.[5]
This is not a theoretical risk. In early 2026, an AI built platform for creating exam questions and viewing grades turned out to expose the data of almost 18,700 user accounts, belonging to lecturers and students. A check that was supposed to let signed in users through did the exact opposite: it blocked anyone who was signed in and let anonymous visitors in. According to the researcher who found it, a mistake a human reviewer would have caught in seconds.[6]
We recently got a request like this ourselves. Someone had built a few handy applications with an AI tool and wanted to put them into real use. When we looked at them, we saw that user data could leak and that they were nowhere near GDPR compliant. Not out of carelessness, but because you only see those risks when you know where to look. That is why a lab needs experience: whoever builds may move fast, but before anything goes live with real students, someone who recognises those mistakes takes a look.
Controlled chaos
A lab brings that energy out of the shadows. Experimenting is allowed, and fast, but within agreed boundaries, with a contact person in IT who knows what is running, with documentation from day one and with a review before real students use it. That way every pilot ends with a clear handover or a deliberate shutdown, not with a tool nobody remembers.
Other sectors already solved this
The idea of a small team that is allowed to experiment outside the regular operation is not new.
- Lockheed Skunk Works. In 1943 a small team led by Kelly Johnson built the XP-80 jet fighter prototype in 143 days, seven fewer than required. It worked in a rented circus tent next to the factory. One of Johnson's rules: work with a small number of good people.[7]
- VRT. The Flemish public broadcaster has VRT Innovatie, an interdisciplinary team of project leads, developers and researchers, and an Innovatielab that tests new ideas, technologies and formats through polls, interviews and prototypes.[8] In 2018 VRT also described an internal digital start-up of four full time staff that builds and tests prototypes with users and turns the insights into formats, products and ways of working.[9]
- BBC News Labs. A team in which developers and journalists build prototypes together that make newsroom work easier, from speech to text to chatbots.[10]
- The UK government. In 2011 a small team built alpha.gov.uk in three months, a prototype for a single government website. It later became GOV.UK.[11]
What these examples share: a small team, a clear mandate, close to the people who have the problem, and the freedom to build something before it is perfect.
For me this is not a new idea. Back in 2009 I was already presenting on skunkworks projects at barcamps: start with an idea, put a small team on it, build a prototype within one sprint, then decide whether to keep it running, shut it down or turn it into a regular project. That is still exactly how I see a lab.
And in education?
It exists there too, though it is less well known. My favourite example is the Digital Innovation Greenhouse at the University of Michigan. A team of developers, designers, behavioural scientists, data scientists and students helps tools that lecturers came up with grow "from innovation to infrastructure". In the first year, more than 22,000 students used those tools.[12]
The smartest part of that approach: every project gets an exit plan at the start. It is commercialised, handed over to the university's IT department, or wound down.[12] Stopping is a planned outcome, not a failure.
Swinburne University in Australia has an Education Innovation Lab where lecturers can try new technology, such as generative AI, in a safe environment and are allowed to fail fast.[13]
Lab, intrapreneurship, skunkworks
The names differ: innovation lab, skunkworks, internal start-up, intrapreneurship. The principle is the same. A small team is allowed to experiment outside the regular processes, so the rest of the organisation can keep running reliably.
How I see such a lab
No building, no sixty page innovation strategy. A small team with a mandate:
- Two or three creative technologists who quickly build a working prototype, from web app to AI application
- A creative educational support specialist who guards the didactics and bridges the gap to lecturers and programmes
- A fixed contact in IT from day one, not only at handover
- A sponsor in senior management who backs the lab when an experiment fails
The way of working is simple. A lecturer, programme or department brings a problem. The lab builds a prototype in a few weeks and tests it in one course or one programme. Then comes an honest decision: throw it away, try again or hand it over.
Throwing things away matters. A lab that never stops anything is not really experimenting.
The handover is the most important part
A lab is not a second IT department working around the rules. It is an antechamber. What works in a pilot moves into the regular operation, and that needs a clear agreement.
The lab
- Tests ideas quickly and cheaply in one course or programme
- Is allowed to make mistakes and stop projects
- Proves with a pilot whether something actually helps
- Delivers documentation and a handover plan
IT and educational support
- Set the boundaries up front: privacy, security, sign in
- Take over what works and scale it up
- Provide management, support and continuity
- Help decide whether something fits the regular operation
Without that bridge, a lab produces what INSEAD calls "orphan products": solutions nobody in the organisation owns.[14] That is why I want IT involved from the start. Not as a gatekeeper afterwards, but as a partner who helps draw the boundaries and knows what is coming their way.
This is also where a lab differs from shadow IT. A tool a lecturer builds in the evening rarely has documentation, because everything lives in the builder's head. A lab delivers what IT needs in writing with every handover. Not a thick binder, but an answer to at least these questions:
- What problem does it solve, and what did the pilot show?
- Which data does it use, where is it stored and for how long?
- Where does it run, how do people sign in and which systems is it connected to?
- Who maintains it after the handover, and what does that cost per year?
- How do you shut it down when it is no longer needed?
Fast does not mean sloppy
A prototype still works with real students. Privacy, security and accessibility apply from the first pilot. Use as little student data as possible, host in the EU and sign in through the institution's existing accounts.
Where it goes wrong
I believe in this strongly, but I am not naive. Many innovation labs do not deliver on their promise. According to a 2015 Capgemini report, up to 90% of labs fail to deliver the value expected of them.[15] The causes mentioned most often:
- No connection to the organisation. The lab builds nice things nobody asked for or nobody can take over.[14][16]
- No clear measure. If you judge a lab on the number of demos, you get demos. Measure how many tools stay in use and how much time lecturers and students save instead.[16]
- An island. A lab with only technologists misses the didactics. A lab with only educationalists builds nothing. You need both.[16]
- The first victim of budget cuts. Even Google cut most of Area 120, its internal incubator, in 2023.[17]
There is one more trap specific to IT: the idea of two speeds, a fast team next to a slow one, is rightly criticised. A fast prototype is only as fast as its slowest connection to existing systems.[18] So a lab only works if IT benefits from it too, for example because integrations with the learning platform or the timetable become reusable for other projects.
Start small
You do not need a new department to get going. Give one or two people a one year mandate, one clear question and a fixed contact in IT. After that year you know whether it works, which is exactly how a lab should think.
Why I am writing this
This is what I already do today, just without the "lab" label. Building small prototypes, testing them in my own classes, improving processes that cost lecturers a lot of time, and throwing away what does not work. I see what one person can achieve that way. With a small team, a mandate and a good working relationship with IT and educational support, a whole institution can benefit.
Sources
- [1]Gartner Unveils Top Eight Cybersecurity Predictions for 2023-2024 (Gartner, March 2023). Predicts that by 2027, 75% of employees will acquire, modify or create technology outside IT's visibility, up from 41% in 2022.
- [2]Microsoft and LinkedIn release the 2024 Work Trend Index on the state of AI at work (Microsoft, May 2024). 78% of people who use AI at work bring their own AI tools.
- [3]Shadow IT (National Cyber Security Centre, July 2023). Guidance from the UK NCSC: shadow IT is rarely malicious, look at the need behind it and avoid unnecessary lockdowns.
- [4]What Shadow IT Can Teach Higher Ed Tech Leaders About the Faculty Experience (EdTech Magazine, September 2026). Why lecturers in higher education reach for their own tools and how IT can offer an easier supported path.
- [5]AI-Generated Code Poses Major Security Risks in Nearly Half of All Development Tasks (Veracode, July 2025). More than a hundred AI models were given eighty programming tasks. In 45% of the cases the code contained a security flaw.
- [6]Lovable-hosted app littered with basic flaws exposed 18K users, researcher claims (The Register, February 2026). An AI built exam and grading platform exposed the data of lecturers and students through missing access control and an inverted sign in check.
- [7]Skunk Works origin story (Lockheed Martin). How Skunk Works started in 1943 and how a small team built the XP-80 in 143 days.
- [8]Innovatie (VRT (in Dutch)). About VRT Innovatie, the Innovatielab and VRT Sandbox.
- [9]VRT takes a three-pronged approach to innovation (EBU, May 2018). VRT Innovatie, VRT Sandbox and an internal digital start-up of four full time staff.
- [10]The role of innovation labs in advancing the relevance of Public Service Media: the cases of BBC News Labs and RTVE Lab (Communication & Society, 2019). Research into how BBC News Labs and RTVE Lab innovate within a public broadcaster.
- [11]Story 2011 (Government Digital Service). How a small team built alpha.gov.uk in 2011, the forerunner of GOV.UK.
- [12]How Universities Can Grow a Culture of Academic Innovation (EdSurge, October 2016). About the University of Michigan's Digital Innovation Greenhouse, its team, its exit plans and student use.
- [13]Education Innovation Lab (Swinburne University of Technology). A lab where lecturers can test new technology in their teaching.
- [14]Is your innovation process a corporate illusion? (INSEAD Knowledge, January 2020). About the distance between innovation labs and the core organisation, and orphan products.
- [15]Three reasons innovation labs almost always fail (Canadian Underwriter, July 2019). Refers to the 2015 Capgemini report on innovation labs that do not deliver on their promise.
- [16]Why Innovation Labs Fail, and How to Ensure Yours Doesn't (Harvard Business Review, July 2019). Simone Bhan Ahuja on alignment with the business, metrics and team composition.
- [17]Google Slashes Most Jobs at Area 120 Incubator as Part of Cuts (Bloomberg, January 2023). Google cuts most of its internal incubator Area 120.
- [18]The flaw at the heart of bimodal IT (Jez Humble, April 2016). A critique of running a fast and a slow IT side by side.
Frequently asked questions
An idea for your course or programme?
We build prototypes and tools for education, from AI tutors to Canvas integrations. We are happy to think along about a first pilot.
See what we do for education