AI and Business Analysts: What's Actually Happening and What to Do

The honest assessment

Business analysis sits in an odd spot with AI. Some of the job is heavily exposed. Some of it is barely touched. And the profession itself has always been a bit vague about where it starts and ends, which makes this harder to pin down.

Here's what AI can already do in business analysis. Document requirements from meeting transcripts. Generate process maps and workflow diagrams from descriptions. Write user stories and acceptance criteria. Produce gap analyses between current and future state. Draft business cases with financial modelling. Analyse data to identify process inefficiencies. Create stakeholder maps. Write change impact assessments. ChatGPT can produce a perfectly serviceable requirements document from a good brief. Claude can take a transcript of a stakeholder workshop and extract, categorise, and prioritise the requirements mentioned. Copilot can build the supporting financial model.

What AI can't do is the translation work that sits at the heart of business analysis. When the operations director says "we need a better system" and the IT architect says "the integration layer needs refactoring" and the end users say "it's just really slow"... a good BA knows these are all the same problem described from different angles, and can synthesise them into a coherent set of requirements that everyone agrees on. That translation between business language and technical language, between what people say and what they actually mean, between what they ask for and what they need... that's a fundamentally human skill.

The shift is that the documentation side of BA work is being automated rapidly. Writing requirements, producing reports, creating presentations, and maintaining traceability matrices... all of this can be accelerated by AI. The elicitation, facilitation, and stakeholder management side remains firmly human. If you're a BA who spends 60% of your time writing documents and 40% in rooms with people, the 60/40 split is about to change.

Your exposure level: Medium

Medium exposure. The analytical and documentation work is automatable. The human interface work is not. Your exposure depends on which side of that divide you spend more time on.

The BAs most at risk are those in heavily process-driven roles where the primary output is documentation. If your job is essentially to attend a meeting, write up what was said, and format it into a requirements template... AI handles all of that faster than you. The BAs least at risk are those who facilitate, negotiate, challenge, and shape. The ones who walk into a room, ask the question nobody else was willing to ask, and change the direction of a project because they understood something everyone else missed.

There's also a split between different types of BA work. Data analysis and reporting are highly automatable (see the data analysts page). Process analysis and improvement are moderately automatable. Requirements elicitation and stakeholder management are barely automatable. If your BA role is heavy on the first category, your exposure is higher. If it's heavy on the third, you've got more time.

Here's the thing that actually concerns me about BAs and AI though. Business analysis has always been a role that's hard to define and easy to cut. In good times, companies hire BAs because they improve project outcomes. In tough times, companies ask whether they really need a dedicated BA or whether the project manager and the developers can cover it between them. AI gives companies another option: skip the BA and let AI handle the documentation while the PM manages stakeholders. That's not how it should work. But it's how some organisations will try to make it work.

The 90-day action plan

1. This week: let AI write your requirements. Take a recent set of requirements you produced. Give Claude the source material (meeting notes, stakeholder inputs, existing documentation) and ask it to generate a requirements document. Compare it to yours. The structure will be solid. The content will be about 70% right. The 30% it gets wrong is where your expertise lives.

2. Week two: automate your documentation workflow. Use Otter AI or Copilot to transcribe your next stakeholder workshop. Then use Claude to extract and categorise the requirements from the transcript. Edit and refine the output. You've just cut your documentation time from a day to an hour.

3. By day 30: build a process analysis toolkit. Use ChatGPT to map current-state processes from descriptions. Feed it process narratives and ask for swimlane diagram descriptions, bottleneck identification, and improvement recommendations. It's not as good as observing the process yourself, but it's an excellent first draft.

4. By day 45: develop your facilitation skills. This is the un-automatable core. Get better at running workshops, managing conflicting stakeholders, and drawing out requirements from people who don't know what they want. Take a course. Read a book. Practice. The best BAs are the best facilitators, and that gap between human and AI facilitation is enormous.

5. By day 60: learn to use AI for impact analysis. When a change is proposed, feed the current system documentation, process maps, and stakeholder landscape into Claude. Ask it to identify the potential impacts. Use it as a starting point for your change impact assessment. It catches things systematically that you might miss intuitively.

6. By day 75: become the AI requirements expert. As companies adopt AI tools, someone needs to define the requirements for AI implementation. What data does the AI need? What decisions should it make? What should be escalated to a human? What are the ethical considerations? This is BA work, and very few BAs are doing it yet. Claim the territory.

7. By day 90: reframe your value proposition. Update how you describe your role, both internally and on your CV. Move from "I document requirements" to "I bridge the gap between business needs and technical solutions, using AI to accelerate the documentation while focusing on the stakeholder engagement and strategic analysis that determines project success." That's not spin. It's accurate, if you've done the work.

AI tools you should be using this week

• ChatGPT for Work — Your daily driver. Requirements documents, user stories, process narratives, business cases, and stakeholder communications. Give it context about your project, your stakeholders, and your organisation's standards. Also useful for brainstorming solutions and challenging your own assumptions by asking it to argue the other side.

• Claude for Work — Excellent for longer, more complex analysis. Paste in a full project brief, existing documentation, and meeting transcripts. Ask it to identify gaps, contradictions, and missing requirements. i find it particularly good at spotting inconsistencies between what different stakeholders have asked for.

• Microsoft Copilot for Work — If you work in a Microsoft environment, Copilot helps with everything from meeting transcription to document drafting to data analysis. The integration means you don't have to copy-paste between tools. It just works where you already work.

• Notion AI — If your project team uses Notion, the AI features help with requirement tracking, documentation maintenance, and project wikis. Summarises pages, generates content, and helps keep documentation current without manual effort.

What to say in meetings

In project kickoff meetings: "I'll be using AI tools to accelerate our documentation process. That means faster turnaround on requirements and impact analyses. It also means I can spend more time in workshops with you, which is where the real requirements emerge."

When a PM suggests the team doesn't need a BA: "AI can generate requirements documents. What it can't do is sit in a room with your stakeholders who disagree about everything and get them to a shared understanding of what we're actually building. That's what determines whether the project succeeds or fails. That's what I do." Be direct. Not aggressive. Direct.

If someone presents AI-generated requirements: "This is a solid structure. Here are the three things it's missed that will cause problems in development if we don't address them now." Being the person who improves AI output is a strong position.

If the worst happens

If you're made redundant from a BA role, your core skill... understanding complex situations and translating between different groups of people... is transferable to almost any professional context. Product management, consulting, change management, UX research, solution architecture, and programme management all use the same fundamental capability.

The most natural adjacent move is product management. PMs and BAs do similar work but PMs typically have more authority, higher visibility, and better pay. If you've been doing BA work well, you already have most of the skills needed for a product role. The main addition is commercial thinking: understanding market positioning, pricing, and competitive dynamics. That's learnable.

i'll tell you something about BA redundancies specifically. They often happen when projects end rather than as deliberate cost-cutting. If you've been on a fixed-term project and it's wrapping up, that's not the same as being replaced by AI. But the effect is similar, and the preparation is the same. Have your AI skills current, your portfolio of successful projects documented, and your network active. The BA market is cyclical but it rewards people who can demonstrate concrete impact on project outcomes. If you've got that evidence, preferably with numbers, the next role is usually findable. Just maybe not as fast as you'd like.