How artificial intelligence is changing the work of a business analyst – a practitioner’s perspective

Before we go any further, it is worth establishing one thing: using AI at work is possible, as long as the client has a tool of that kind internally. After work – knock yourself out.

Not every client has already implemented a solution based on artificial intelligence, and data is the crucial issue here – the beginning of everything.

Few year ago, I would say that a business analyst is someone who “glues” IT world with the world of business. Sitting in on the meeting, noting every word, creating tables and charts, so that everybody knows what it was about, and overwatching no detail is gone in the process. A bit of a translator, a bit of a diplomat. Sounds peacefully, calm? That’s the case – those were the days, when the project pace was different, and scopes of tasks were quite predictable.

It’s totally different today, though. Projects look more like a roller-coaster ride without belts fastened: sprints, changing priorities, a few topics at a time, accompanied by a client who sometimes does not know what he/she wants. It is demanded from the analyst not only to note all the requirements but also to understand the context, catch risks, suggest solutions, and sometimes even say, quite plainly: “Look, it is not going to work, because you lack X and Y”.

And in this whole mayhem, artificial intelligence emerges more and more often. Being not a foe of a threat but offering additional pair of hands (or rather an extra brain) ready to help in less challenging tasks or, on the contrary, in even more exciting and thrilling ones.

How does business analyst’s daily routine look like?

Obviously, each and every project varies. Nevertheless, a business analyst’s daily tasks, summed up in a few bullet points, could be the following:

• Gathering needs – meetings, workshops, questionnaires, or sometimes tediously perusing regulations or outdates procedures
• Modelling the process – figures, charts, BPMN, UML, or sometimes improvising at the blackboard during the brainstorming session
• Defining requirements – user stories, backlog, approvals, and sometimes sloppy, quick notes made under the pressure of time and loads of topics which need, in the end, to be carefully organized and written, as befits a good analyst
• Facilitating communication with others – with plenty, various groups of others: developers, clients, teams of DevOps or UX/UI designers, architects, and many many more. Each of them have their own language, and the analyst has to smoothly switch between them.
• Supporting tests – ensuring that what has been created is actually what was agreed on.
• Creating documentation – which means basic BRD (Business Requirements Document) or FRD (Functional Requirements Document). At times, a client has its own drafts of such crucial documents, but it can also be the other way around.

Nowadays, AI can help us in all of the abovementioned areas and domains.

In which areas of an analyst’s work AI actually proves effective?

Theoretically speaking, “everywhere”, but I prefer to provide specific examples. The ones that can be, in fact, used.

1. Meeting notes – there were times when the analyst had to listen to every conversation and note them down, no detail missed. Today, however, if the client agrees to recording and to using specific tools, the analyst can use, for instance, “Otter.ai”, “Fireflies”, or “Zoom AI Companion”. The moment the meeting ends, the analyst has it transcribed, with decisions and questions bolded. Other exceptional tools worth mentioning are “Transcriptor” and “Sound type AI”. Why is that? During the meeting, you can focus on running it effectively (AI fails at that) and not on taking notes.
2. First drafts of user stories – coming back from workshops, you have a sheet of paper full of hand-written notes, and – if you have used AI-approved – it is possible to copy it to “ChatGPT”, “Claude.ai “, or, recently gaining ground, “Gemini”. Within only seconds, you have an initial structure. Then, it is time to concentrate on corrections and customizing them to project standards. Obviously, AI cannot be just trusted; you need to double-check whether the text covers all that you expected.
3. Process diagrams without drawing – no need to click in Visio or Draw.io – the process can be described in simple sentences, and a tool (e.g., Lucidchart from AI) draws a diagram for you. It often happens that we only have incomplete information about the process itself, data that doesn’t provide a full picture and context required to reproduce it in BPMN. Despite that, we are still obliged to offer a proposition based on what we have already known. In projects, analysts are often expected to draw the process even though the business hasn’t shared all the requirements yet. In such cases, it is possible to get support from AI to verify what the process may look like on the basis of speculation and available information. The outcome should be treated as a draft that might help in creating an initial process.
4. Real-time data analysis – usually, you cannot analyse data on the fly, not being in your client’s environment. Only then – and only under the condition that you have tools – you are allowed to use AI or Power BI to run a preliminary analysis.
5. Creating presentations and summaries – AI produces an initial presentation outline – slide layout, main bullet points. Then I add details, correct the language, change graphics. It may not be an arcane knowledge, but is still worth mentioning as an convenience.
6. Analysis of legal or technical documents – once in a while, you need to read 100 pages of a contract or specification in no time and master the details. If you have AI in client’s environment, artificial intelligence will sum up the document in 2 minutes and highlight for example moot points and ambiguities and prepares a recap.
7. Generating test cases – on the basis of the requirements, you can ask AI for a proposal of test cases, which should then be verified with testers.

As a matter of fact, AI will give you a hand in every problem, as long as it is trained in the subject in question. What is also crucial is the tool you choose to use. It is common to be allowed to use only tools given by the client, though.

Which analyst’s tasks are out of AI scope?

In spite of all the help AI can provide, there are still some things out of AI’s reach. It cannot sense that the client has left some things unsaid or does not have enough knowledge in a certain filed. It will not notice that the atmosphere in a team of developers is breaking down and this will result in delays. It will not grasp the subtleties of an internal policy in a corporation.

A business analyst operates a little bit as a psychologist, a tad strategist, and a kind of discussion facilitator. AI can bea tool, but not a partner in making tough decisions. Simply speaking – an analyst is irreplaceable. It is an analyst duty to communicate with various stakeholders and to be the first person to prototype a solution.

Summary

If someone asks me today, “Will AI take your job?”, I would say, “No, but surely it will help me with paperwork and tedious clicks“. For me, artificial intelligence is like an extra assistant who sees no problem in finishing a diagram at 11:00 p.m.

The best part is that the quicker we learn how to use AI, the more time we have left for the part of the job for which we became analysts in the first place: talking to people, looking for solutions, connecting the dots. The rest… let robots and bots do it.

We should concentrate on what a human could only do: design, communication, “problem solving”, implementation and “socialising ideas” basing on client’s needs.