A senior BA at a services firm asked me last month what's changed about the job in the last five years. He'd been heads-down on client work since 2020 and was preparing for a career conversation. He wanted a clean list.

The clean list exists. Five things changed visibly. Three things didn't change at all. The interesting part is which list matters for your next five years, and most career advice for BAs gets this exactly backwards.

What actually changed (the surface)

One. AI tools became normal in the workflow. In 2020 a BA who used GPT-style tools at work was rare and slightly suspicious. By 2024 it was standard. Claude, ChatGPT, Notion AI, Miro AI — most BAs now use at least one daily. The companies that banned them in 2023 quietly unbanned them in 2024. The shift happened faster than any previous tool adoption I've watched.

What the tools are actually used for, honestly: drafting acceptance criteria first-pass, summarising stakeholder transcripts, generating user-story candidates from prose, restructuring rambling requirements documents, and producing meeting notes. Not requirements analysis itself. Mostly the boring documentation overhead that used to eat a third of the week.

Two. Async work permanently changed how requirements get captured. Remote and hybrid teams meant that the spoken stakeholder workshop became the recorded one, then the Loom video, then often nothing — replaced by Slack threads and shared documents. Requirements gathering got faster and noisier. Fewer formal workshops, more continuous capture across channels.

The good BAs adapted by treating Slack and Loom as raw input that needs analysis, the same way they used to treat workshop transcripts. The mediocre BAs treated the messages as the requirement itself, copy-pasted them into tickets, and shipped confusion.

Three. "Product Owner" became the title services firms preferred. In 2020, most Indian IT services firms hired BAs. By 2025 the same role had been quietly relabelled as PO on most job postings. The work didn't change. The title did, because clients running Scrum wanted to see "PO" on the staff aug invoice. Anyone who saw this as a real role transformation rather than a relabelling exercise wasted two years chasing certifications for a job they were already doing.

Four. Tools got cheaper, more numerous, more interchangeable. Jira's monopoly weakened. Linear, ClickUp, monday, Notion, Coda, Asana all became viable for requirements work. The good news: BAs got more leverage to push back on tool decisions. The bad news: a lot of BA energy went into tool migrations that didn't change outcomes.

Five. Stakeholder expectations sped up. "When can you have this back to me" used to mean a few days. By 2025 the expectation in many orgs is hours. Partly because AI tools made faster turnaround possible, partly because async work made stakeholders impatient with waiting on a workshop they're not in. The BAs who responded by getting faster shipped more low-quality work. The ones who held the line on analysis time got better outcomes but had to negotiate harder for it.

What didn't change (the underlying craft)

One. Requirements still come in wrong. Stakeholders still describe their preferred solution and call it a requirement. They still leave out the trigger that generated the request. They still believe the priority they assigned is meaningful. They still misremember what they said in the workshop last week. The job of decoding what they actually need from what they actually said is identical to what it was in 2020, or 2015, or 2005.

Two. The work that matters is still problem understanding, not story writing. The BA who can articulate why a feature should exist, who's affected, and what changes if it ships still outperforms the BA who writes beautiful INVEST-compliant stories about unvalidated requirements. AI tools made the story-writing part faster. They did nothing for the problem-understanding part, which is judgement and remains stubbornly human.

Three. Stakeholders still don't know what they want until they see what they don't. The iterative discovery process — show a prototype, get a real reaction, refine, repeat — is the same as it ever was. The cycle time shrank because tools got better, but the cycle itself is unchanged. The BAs who treat the first stakeholder statement as the requirement, even now, ship the wrong thing.

The gap that nobody talks about

Here's the actual point of this piece. The gap between "what changed" and "what didn't" is widening, and most BAs are on the wrong side of it.

A BA who has spent five years getting better at the surface changes — faster AI-assisted documentation, more tool fluency, better Slack-thread synthesis — has a CV that looks current. They can write the right keywords. They can talk about AI tools confidently in interviews. They've earned a few certifications in the new tools their company uses.

What they have not done is gotten meaningfully better at the underlying craft. Their problem-understanding is the same level it was in 2020. Their ability to interrogate a requirement is the same. Their stakeholder management is the same. They are faster at producing documents that nobody reads with the same level of insight as before.

This is fine for the moment because the market still values surface signals. It will not stay fine.

Two forces are closing the gap:

The first is AI itself. Surface BA work is exactly what AI tools are best at. Drafting stories, summarising transcripts, formatting acceptance criteria, generating documentation. Any BA whose value is concentrated in those activities is competing directly with a tool that costs $20 a month and gets better every quarter. The market is going to notice within the next few years.

The second is that companies are starting to feel the cost of BAs who can document fast but can't think well. The features ship, the documentation is clean, the metrics don't move. Eventually someone asks why, and the answer points back to requirements that were well-formed but aimed at the wrong target. The teams that get burned by this start hiring differently.

What to actually invest in for the next five years

If the surface changes are dangerous to over-index on, what should a BA invest in instead?

Problem decomposition. The ability to take a vague stakeholder statement and break it into the actual underlying needs, with evidence, in a way that survives challenge. This is the muscle that doesn't atrophy and that AI cannot replicate. It's also the hardest to teach, which is why most BA training avoids it.

Stakeholder interrogation. Not asking what they want. Asking what they're trying to do, who else is affected, what changes if you ship, what happens if you don't, what they've tried before. The questions are the work. The BAs who can ask these questions well, in real time, in a room with a defensive stakeholder, will be valuable in five years the same way they're valuable now.

Judgement under ambiguity. Knowing when a requirement is real versus when it's someone's preferred solution to a misdiagnosed problem. Knowing when to push back and when to absorb. Knowing what to document formally versus what to handle in a hallway conversation. This judgement is built case by case over years and cannot be downloaded.

Domain depth. A BA who understands payments deeply, or insurance underwriting, or regulatory compliance, or healthcare workflows, has something AI cannot easily acquire and most other BAs don't have. Generalist BAs are increasingly replaceable. Specialist BAs are increasingly necessary.

None of these things are new. They were the right things to invest in in 2020 too. The difference is that the cost of NOT investing in them is rising faster now, because the surface skills are getting commoditised.

The honest summary

Five years of change in the BA role is real but mostly cosmetic. AI tools, async work, the PO rebranding, tool proliferation, faster turnarounds — these changed how the work happens, not what the work is.

The work itself — turning vague stakeholder statements into shipped features that actually solve the underlying problem — is the same job it was in 2020. It will be the same job in five years. The BAs who recognise this and invest accordingly will compound. The ones who chase the surface changes will find themselves doing increasingly commoditised work for stakeholders who increasingly notice they could have used a tool instead.

If you're preparing for a career conversation, that's the honest framing. The market signals are changing. The actual job mostly isn't.