LO
Information systems
Trains you to run and change systems other people wrote: to find out what an organisation already has, what depends on what, what a record in it is allowed to mean, and what would break if you switched one of them off.
Pick one system you can actually reach — a society's, a part-time employer's, your department's — and draw what it is connected to by asking people rather than by reading the documentation. Count the connections nobody could name. That count is the one thing on this page nobody can generate for you.
A single number for a whole major would hide the thing that matters: this degree trains several separate competencies, and they are not all moving in the same direction. Automation acts on tasks, so any assessment lives on the occupation pages below — not here.
Written for undergraduate degrees variously called information systems, management information systems or information management — the ones usually housed in a business school and built around systems an organisation buys rather than systems it writes. A programme with the same name inside a computing faculty is often computer science with a business minor, and its graduates land somewhere else. One thing this page cannot generalise across matters more than the curriculum: whether you end up inside the organisation that runs the systems, or at a vendor, a consultancy or a managed-service provider, because those three sell the work the first one is trying to keep. Note also that most people doing this work are employed by a bank, a hospital or a factory rather than by an IT firm, so statistics about technology-industry employment largely do not see them.
What this degree actually trains#
Not the course list — the competencies underneath it, and whether each one is worth more or less than it was.
Knowing what the organisation already runs
Worth more than beforeWhich systems hold which records, which one is authoritative when two of them disagree, what is wired to what, and that one integration somebody built in 2019 and left.
This is the input that did not get cheaper while everything around it did. Writing the integration code is one of the things these tools are best at; knowing which of four systems holds the authoritative version of a record, and which field is quietly lying, lives in people who have been there for years and is written down almost nowhere. A text-to-SQL benchmark built from real corporate warehouse query logs rather than public datasets makes the same point from the other side: a plain language model scored zero there, against eighty to ninety per cent on the public benchmarks, and the reasons given were structural — real schemas rot into several columns with the same name, and every estate carries local idiom.
Translating between the business side and the technical side
Worth less on its ownTurning what a manager says they want into something an engineer can act on, and turning a technical constraint back into a choice a manager can actually make.
This is the degree's oldest selling point and it is squarely what a drafting tool does best: describing a workflow in words and getting back the objects, the rules and the specification is inside these tools' range. What did not get cheaper is knowing which of the described steps people actually skip, and why — that is learned by watching a team work, not by reading the process document. So the translation still has to happen and it no longer pays on its own. It pays when the person doing it also knows the estate.
Producing the requirement, the process map and the report
Worth less on its ownThe written output of the degree: requirements documents, process diagrams, test plans, queries and dashboards.
Each of these is structured text with a checkable outcome, which is the property that lets a tool draft, fail and redraft with nobody watching. Natural-language querying and chart generation are already built into the products analysts use. Two consequences follow and they are not the same one: drafting stops being the thing you are paid for, and the number of drafts goes up — so the hours move from producing these documents to finding the one that is now wrong and arguing with the person who is quoting it.
Saying what a record is allowed to mean
Worth more than beforeDeciding which system's version of a customer is the real one, what a status value means, who may set it, and writing that down in two sentences that the two people who use it both accept.
Self-service querying only produces right answers on top of definitions somebody maintains, so the scarce work moves from answering questions to curating the layer everyone else answers from. It has also stopped being only an internal good habit: since August 2026, a deployer of a high-risk system in the EU that controls the input data must ensure that data is relevant and sufficiently representative. That duty lands on the party running the system, not the party that built it, and this degree's destinations are all on the running side.
Holding a system somebody else built
HoldingConfiguring a bought platform, and then living with the configuration: the rule added last year that is wrong now, the field nobody fills in, the permission granted once for a reason nobody recorded.
The building half got much cheaper and the holding half grew by exactly the same mechanism: every rule added is a rule that will later be wrong and have to be found, so faster configuration produces more configuration rather than less work. The hours do not disappear, they move from building to untangling, and untangling is invisible on a roadmap. That is why this is holding rather than a saving — and why it is the part of the job that gets budgeted as though it had been removed.
Where it can lead#
Several directions, never one. Each says what your training reuses, what graduates typically lack, the real entry conditions, and one thing you can test this term.
Owning the system the business runs inside
task-level analysis →- What transfers
- Nearly all of it, and this is the closest thing this degree has to a direct path: configuring a bought platform, keeping its records worth trusting, deciding what its automation may do with no person involved, and being the one who knows which of the numbers coming out of it are solid and which are held together by an assumption.
- What graduates typically lack
- Graduates arrive able to configure and unable to hold a line. That occupation's own constraint is that the role usually has the knowledge and not the authority: when the team carrying the revenue number wants the automation to send more, the person who can say no sits further up. Being overruled repeatedly is how a boundary erodes without any decision ever being recorded, and nothing in a degree rehearses that.
- Entry reality
- A genuine graduate-level door, usually advertised under a title like systems administrator, sales operations or business analyst rather than this one. In practice a certification in the specific platform an employer runs decides the interview more than the degree name does — which cuts both ways: it is cheap to obtain and it attaches you to one vendor's product.
Find a system a society, a lab or a part-time employer actually runs on. List every action it takes today with no person involved: assignments, reminders, status changes, messages sent. Mark each one fine, should be reviewed, or should never have been automatic, and take the three counts to whoever owns it. The argument you get back is the job.
In through the service desk, which is a door and not a destination
task-level analysis →- What transfers
- The desk is where estate knowledge is actually acquired: what the organisation runs, in which version, wired to what, and which of it breaks on a Monday morning. Nobody hands a graduate that map. You assemble it out of other people's problems, which is why a year here is worth more to this degree than to any other.
- What graduates typically lack
- Graduates arrive able to describe a process and not to work a queue where the user's account is confident, well-meaning and wrong about the one detail that decides everything. They also underestimate what the entry rung consists of: the repeat half of the queue — resets, unlocks, access requests, the printer — is the clearest automation target in an office, because the problem is narrow and a machine can check whether the fix worked.
- Entry reality
- This has been one of the main doors into technology work for thirty years, and it is narrowing for three separate reasons of which only one is a machine. Singapore's infocomm regulator assessed stand-alone applications and systems support roles as high-impact and named all three: tools answering help-desk questions and routing tickets, the development and DevOps function absorbing the maintenance under Agile and CI/CD, and a SaaS or PaaS provider taking over maintaining and updating the software. The second is a reorganisation and the third is a purchasing decision, and neither is affected by how good you get at the work. Plan for a door that narrows for reasons no technical skill of yours can address.
Work or volunteer a shift somewhere real requests arrive — campus IT, a society's systems, a part-time job. Log ten of them and sort them into two piles: the ones where a machine can tell whether the fix worked, and the ones where only a person can. The ratio is what the tooling takes first, and you produced it rather than read it.
Where that work went: platform and operations
task-level analysis →- What transfers
- The operational half of this degree — thinking in terms of what runs, what depends on what, and who gets called — applied where it is the entire job rather than one module. The same workforce study that expected stand-alone support to shrink expected these duties to land here: it says service-level and system-performance oversight moves onto DevOps teams, and lists automation and orchestration engineers among the roles growing in demand.
- What graduates typically lack
- Graduates from this degree have written configuration for an assignment and have never held anything through a bad night, which is the part that cannot be studied. The second gap is reading: more of this work is code and configuration you did not write, generated or not, and a large practitioner survey found adoption of these tools relating positively to throughput and negatively to delivery stability. More gets shipped and more of it breaks, so the value sits with whoever can say why.
- Entry reality
- Not a graduate door in most places. The same study named DevOps engineer as an easy-to-moderate move from support roles, which is the honest route: two or three years at a desk or in development first. Coming from this degree rather than computer science, expect to be the weaker programmer in the room and to be paid for knowing what the thing is connected to.
Take one service your university or a society runs that you did not build, and find out two things without opening the documentation: who would be called if it stopped at three in the morning, and what it depends on that nobody has written down. How long that takes you is the measurement, and the answer nobody can give you is the finding.
Deciding what an organisation buys, switches on and retires
task-level analysis →- What transfers
- Knowing the estate, which that occupation page names as the binding input and the one nobody can retrieve: which step is actually the bottleneck, which system holds the authoritative record, whose numbers move if it changes, and who will be angry. It is collected by someone who can walk around and ask, and this degree is the one that teaches you what to ask.
- What graduates typically lack
- Graduates from this degree can run the meeting and usually cannot build the check. The task that decides whether a tool is worth adopting is taking real cases out of the business — not the vendor's demo — running the tool on them, and counting how often it is right, wrong, and confidently wrong. That is closer to designing an experiment than to managing a project, and deciding what counts as correct for this particular company is a business judgement wearing a technical costume.
- Entry reality
- Rarely a first job, and hard to read from outside: that occupation page marks its confidence low because almost everything published about the role is written by people selling adoption services. Two rules now name a person rather than a team — a US federal memorandum requires each agency to identify a chief AI officer, and since August 2026 an EU deployer of a high-risk system must assign oversight to natural persons with the necessary competence, training and authority. Read those as evidence that the duty exists, not as evidence that the post is funded where you live.
Take one tool a society, a department or your employer has switched on, and build the check nobody built: collect twenty real cases from before it was turned on, run it on them, and count three things — right, wrong, and confidently wrong. Twenty cases is a small number and it will still tell you something the vendor's material does not.
What to add outside the classroom#
This is about what graduates commonly lack in practice — not a claim that your school failed to teach it.
Time inside an estate you did not set up — an internship in an IT function, a part-time job, a society's systems. What coursework cannot reproduce is inheritance: a system you built last month is one you understand completely, and no system you will ever be paid to run is like that.
Reading a system's documentation against the running system and finding where the two disagree. They always disagree. Where they disagree is where the job is, and being able to state the difference in one sentence is what separates this degree from the certification that gets you the same first interview.
The part of the law that names the operator rather than the builder — in the EU the duties the AI Act places on a deployer, and wherever you are, its local equivalent. Every destination in this degree is on the operating side, which is the side those obligations have been landing on, and they are short and public.
This term#
One or two actions, each producing something you can show someone. Not a reading list.
Pick one system you can actually reach — a society's, a part-time employer's, your department's — and draw what it is connected to by asking people rather than by reading the documentation. Count the connections nobody could name. That count is the one thing on this page nobody can generate for you.
Write the definition of one field that decisions depend on: what it means, who may set it, and when it is wrong. Show it to two people who use it. If they disagree with each other, keep both versions — that disagreement is the artefact, and producing it is the work.
Common questions#
This site will not answer that with a score, and for this degree a single verdict would be especially misleading, because its parts are moving in opposite directions. Read the four occupation pages it reaches — the business systems owner, the service desk, platform and operations, and the person accountable for AI adoption inside a company — and compare them task by task. What holds across all four is the same thing: knowing what an organisation already runs, what depends on what, and what a record is allowed to mean. What is genuinely thinner is the first rung, because the service desk is narrowing for three reasons of which only one is a machine. So the question worth deciding is not whether to leave this degree. It is whether you can reach that estate knowledge without the entry job that used to hand it to you, and the paths on this page are written to be read as four different answers to that.
They answer to different things. Computer science trains you to build a system precisely enough that a machine executes it. This trains you to run and change systems other people built, inside an organisation that will not stop working while you do it. Where the two overlap — writing routine code and configuration — the computer science graduate is better trained and the tools are strongest, so competing there is the one mistake this degree makes easy to make. Compete where it is different: no model knows which of your four systems holds the real customer record, and the people who do know are paid for it.
The bridge as usually described — taking what a manager wants and restating it so an engineer can act on it — is close to what a drafting tool does best, and that is the honest problem with this degree's oldest pitch. What did not get cheaper sits underneath it: knowing which of the described steps people actually skip, and which system's version of a number is the real one. A benchmark built from real corporate warehouse query logs rather than public datasets found a plain language model scoring zero where the public benchmarks read eighty to ninety per cent, and the reasons were structural — real schemas rot into several columns with the same name, and every organisation carries its local idiom. So translating on its own is worth less than it was, and translating by someone who knows the estate is worth more.
Probably narrower, and the useful part is why. Singapore's infocomm regulator assessed stand-alone support roles as high-impact and named three routes: tools answering help-desk questions and routing tickets, the development and DevOps function absorbing the maintenance under Agile and CI/CD, and a SaaS or PaaS provider taking over maintaining and updating the software. Only the first is automation. The other two are a reorganisation and a purchasing decision, and neither is affected by how good you get at the work. Two things about that study make it more useful than a fresh headline, not less. It is a forecast rather than a measurement, published in 2022 on analysis dated 2020 — so the route it calls automation was written before generative tools reached the public, while the two organisational routes are the ones its authors could already watch happening. And its horizon has now run out, which makes it one of the few predictions about this job you can actually settle. Go and settle it on the desk you are applying to: ask how many people work it now against three years ago, and whether maintenance of the main systems still sits in-house.
For the first job, often yes, and that is worth knowing rather than resenting: a certification in the platform an employer runs gets you an interview for the role that administers that platform. What it does not do is travel. It is specific to one vendor's product, it expires, and the employer who values it values it because you will hold their instance — not because you can judge which instance is worth holding. The judgement this degree can produce has no certificate behind it, which is exactly why it has to appear as something you did: a system you mapped, a definition you wrote, a count you produced.
Enough to read, and the target is a specific one rather than a level. The work this degree leads to is full of code and configuration you did not write, generated or otherwise, and the question you are paid to answer is what it touches and what happens if it is wrong — not whether you could have produced it faster. In practice that means SQL you can read against a schema you have never seen, a scripting language you can follow, and the habit of finding out what a query joins before you repeat what it returned. Trying to out-program people from a computer science degree is the losing version of this degree, and it is also the version the tools are best at.
Method#
Assessments live on tasks, not on majors. Follow any direction above to its occupation page to see which tasks are changing, how strong the evidence is, and what it does not yet show.