Business analyst

United States. The statistics bureau projects that employment of management analysts will grow 10 percent from 2025 to 2035, much faster than the average, with about 94,100 openings a year. O*NET lists business analyst among the job titles reported for this occupation, which also includes management consultants; the page itself does not mention business analysts or AI. It counts jobs, not tasks.

A named person with standing publicly predicted something, on a date, in an attributable statement. It is recorded so that who said what, and when, stays checkable — and it never moves a task's assessment, because a prediction is not an observation. Its value arrives later: the record sits on the same page as the evidence about that occupation, so anyone reading the forecast reads the record of what happened next beside it. That is the reckoning; this site publishes no verdict on whether a forecast came true.

United States. The statistics bureau projects that employment of computer systems analysts will grow 8 percent from 2025 to 2035, much faster than the average, from 544,400 to 587,200, with about 32,900 openings a year, and says that as organisations continue to rely on and expand IT, including artificial intelligence, computer systems analysts will be hired to design and install new systems. O*NET files IT business analyst among this occupation's job titles. It counts jobs, not tasks.

A named person with standing publicly predicted something, on a date, in an attributable statement. It is recorded so that who said what, and when, stays checkable — and it never moves a task's assessment, because a prediction is not an observation. Its value arrives later: the record sits on the same page as the evidence about that occupation, so anyone reading the forecast reads the record of what happened next beside it. That is the reckoning; this site publishes no verdict on whether a forecast came true.

United States, Veterans Health Administration. The agency's inventory entry, marked deployed and developed in-house, says its human-centered design team uses large language models and custom agents, and that the system streamlines tasks such as analysing user feedback, generating user stories, summarising workshop outcomes and refining user personas, with faster insights and less manual labour for staff. It is the agency describing its own system; it does not report effects on staffing or the quality of the stories.

An employer has put it into production. Can move the baseline — weighted by scale and how similar the setting is.

One project at an IT consulting company, where language models generated functional design specifications and user stories from elicitation summaries and templates, assessed by the company's expert analyst. The analyst estimated drafting time savings of 10% to 15%; some user stories stayed unaddressed across all models because knowledge owned by analysts can be tacit and absent from the elicitation documents; and the authors conclude that the models can enhance early documentation but cannot replace human intervention, with expert analysts still needed for accuracy and completeness.

Small-scale trial in a real setting. Tells us the deployment conditions are being tested, not that they hold — so one pilot is never enough on its own; two independent ones are.

An academic chatbot evaluated in 33 simulated stakeholder interviews, with students playing stakeholders. It made a similar number of common mistakes as a human interviewer and elicited up to 73.7% of all requirements. The authors write that human-led interviews are still valuable and necessary for highly sensitive or complex elicitation, where interpersonal dynamics play a crucial role in uncovering unspoken requirements.

A demo, benchmark or paper shows the task can be done. Updates what the technology can do — not what employers will do.

United States, Social Security Administration. The inventory entry, marked deployed with an operational date of 1 October 2024 and a purchased vendor platform, says the generative component accelerates legacy modernisation by automatically analysing COBOL and other legacy code, generating consistent business requirements documentation and refactoring code into modern languages, which reduces manual effort for document creation. It recovers requirements from existing code rather than eliciting them from people.

An employer has put it into production. Can move the baseline — weighted by scale and how similar the setting is.

A study comparing GPT-4 and CodeLlama with human benchmarks in drafting a requirements specification for a university student-club management portal, graded on eight criteria. It reports that the models can match the output quality of an entry-level software engineer, and that human documents took 4 to 24 hours once the requirements were specified while the model drafts took 7 to 47 times less time, though they were hard to get right the first time. The requirements were given, not elicited.

A demo, benchmark or paper shows the task can be done. Updates what the technology can do — not what employers will do.

Austria, the IT division of a postal group. Researchers and company staff implemented an autonomous LLM-based agent system to improve user story quality and evaluated it with 11 participants across six agile teams. The paper says the system's outputs currently require manual validation by the product owner to align with project goals and stakeholder expectations, and six survey participants found the rewritten descriptions too long. It is a small pilot at one company.

Small-scale trial in a real setting. Tells us the deployment conditions are being tested, not that they hold — so one pilot is never enough on its own; two independent ones are.

If you are starting out, expect language models to draft requirement documents and user stories faster than you can type them, and expect your value to lie in what they cannot do: getting the real need out of people, noticing what was left unsaid, and checking the draft against it.

Expect less time writing and more time eliciting, negotiating and validating. The analysts in demand will be those who can run the hard conversations and judge whether a machine-drafted specification is right.

Stay and strengthen

Stay, and move your time from writing to eliciting and validating

Drafting is where tools are strongest; the research recorded here still leaves interviews on sensitive matters and validation with people.

Reshape the role

Move towards process and change work

Getting departments to agree and people to adopt a new way of working is negotiation that tools summarise but do not do.

Adjacent move

Move into product management

Deciding what to build and why sits next to the analyst's work and depends on the same judgement about what people need.

Not on present evidence. Language models draft requirement documents and user stories, and US federal agencies use AI to generate them, but the studies recorded here still have people validate the drafts and lead interviews on sensitive or complex needs, and US projections expect the occupations that include business analysts to grow.

We do not answer that with a number of years. A signal to watch instead: whether machine-drafted requirements start going to developers without an analyst validating them, and whether interview bots are trusted with complex, sensitive elicitation. Neither is what the evidence shows today.

It can draft them. A study found GPT-4 drafts of a requirements specification comparable to an entry-level engineer's and 7 to 47 times faster once requirements were specified; at a postal group's IT teams, an agent's improved user stories still needed the product owner to validate them.

US projections expect management analysts to grow 10 percent and computer systems analysts 8 percent from 2025 to 2035; both include business analysts. The work is shifting from writing documents towards eliciting, negotiating and validating.