Don't Delete Your Training Data

What's actually happening to analytics roles, and why the conversation your team needs to have is the one nobody's having.

By Helena Turpin, Co-Founder and CEO of GoFIGR

I'm going to start with a confession: I'm not an analytics professional, but I do spend a lot of time with the people who are deciding their future. Leaders trying to work out what AI means for their business, their teams, what roles still make sense, what comes next.

I'm not writing this from some position of smug certainty. I run a SaaS company, a lot of us are questioning our existence a bit right now. But I have spent my career obsessing over what makes organisations work for the humans in them, and that's exactly the lens you need when everyone's drowning in AI hype and nobody's talking clearly about what it means for real live people and their careers.

The thing everyone in this room is feeling but not quite saying

If you're a junior analyst right now, you've probably had at least one moment in the last twelve months where you've wondered whether your career path still exists.

I've felt it myself, that uneasiness that surfaces when you read another headline about AI replacing knowledge workers, or when a senior leader at your organisation makes a throwaway comment about not needing to backfill that role because 'AI can do it.'

That feeling is valid. And it deserves a better answer than 'don't worry, you'll be fine.'

Here's what I do know from working with organisations on AI workforce impact: the picture is more nuanced, more interesting, and if you're willing to lean into it, a lot more hopeful than the headlines suggest. But I'm not going to get there by pretending the disruption isn't real. It is.

The honest version

Some of what junior analysts have historically spent their time on - data cleaning, routine reporting, pulling standard queries, building the same dashboard for the fifth time. AI does that quite well now. Anyone who tells you otherwise probably hasn't used the tools.

If you're terminally online like me, you have seen Matt Shumer's viral post - he's an AI founder who has been unusually blunt about what the technology can do, and he doesn't sugarcoat it. AI is genuinely transformative, and the rate of improvement isn't slowing down. The people who are best served right now are the ones who accept that reality and ask: so what do I do next?

But here's what the doom narrative gets wrong.

It confuses tasks with roles - as if all analysts or data scientists or CDOs in big and small companies do the same job. Two people can have the same job title without doing the same thing.

AI is automating tasks - the routine, specific and often repetitive stuff. It is not, (at least not yet and not at scale), replacing the judgment, the business context, the stakeholder relationships, and the translation work that makes analytics professionals incredibly valuable to an organisation.

The mistake being made in too many leadership teams is treating 'AI can do some of what junior analysts do' as equivalent to 'we don't need junior analysts.'

And if you need data to back that up: researchers at Google DeepMind published a paper this month arguing that intelligent AI systems should deliberately route tasks to junior team members to preserve the learning pipeline - not as charity, but as a design requirement. Even the people building these systems know that cutting the apprenticeship pipeline is a structural mistake.

IBM just proved it. Accidentally.

In February 2026, IBM announced it was tripling its entry-level hiring in the US. Not despite AI, but because of what AI revealed.

Their Chief HR Officer was straight about it: the entry-level jobs of two or three years ago, AI can now largely do. So, IBM didn't freeze hiring (well, you can google what they did initially!). They rewrote every entry-level job description, junior software developers now spend less time on routine coding and more time working directly with customers, building new products, solving novel problems.

IBM didn't delete their training data. They retrained the model.

And here's the workforce intelligence argument that doesn't get made often enough: if you cut your entry-level pipeline today because AI can automate those tasks, you will have a serious problem in three to five years. The senior analysts and data scientists you'll desperately need then, they learn by being junior analysts now, you can't skip that part, and you can't buy that institutional knowledge from somewhere else when you need it.

In machine learning terms, you need the training data to build the model. Delete it, and you've got nothing to learn from.

And some new doors are opening

Here's the part I find exciting, think about what's happened to coding, design, copywriting, financial modelling. Skills that used to require years of specialist training are now accessible to anyone willing to learn the tools. I'm doing things today with design and code that I simply couldn't do even a year ago. The barrier to entry has collapsed across the board.

For analytics professionals, that's not a threat - that's an expansion of your territory.

Analysts have always been the people closest to the business questions, the people who understand what the data means in context. What's changing is that the barrier between 'I understand the question' and 'I can build the thing that answers it' is collapsing.

The analysts who go for it - who use AI to extend their reach into territory that was previously out of reach - are going to be extraordinary. Not replaced but amplified.

The Institute of Analytics Professionals of Australia has sensed this shift for quite some time as reflected in their course curation.

Your unfair advantage

While I was putting this piece together, I shared a draft with Antony Ugoni, a Chief Data Officer and IAPA OG who zoomed straight in on one line: 'understand what the data means in context.'

He gave me the example of banking transaction data: A customer's spending pattern shifts. Purchases start appearing from baby shops, a larger supermarket spend, a change in the kind of restaurants. Nothing in the data explicitly says, 'this person is starting a family.'

Where you’re irreplaceable is your ability to bring your own humanity to the data, to read the signs that don't announce themselves, to ask the question underneath the question. That's the thing that makes an analyst extraordinary, and it's the thing that gets built over years of being close to real business problems with real stakes.

Think about your career, not just your anxiety

So, here's a framework I find useful. In the work we do at GoFIGR helping organisations plan through AI transformation, we borrowed it from how good investors think: safe bets, directional bets, and things to watch. It cuts through the doom and forces a more useful question - not 'will AI take my job?' but 'what do I actually know, and what should I act on now?'

Applied to an analytics career, it looks like this:

Knowing something is a safe bet is only useful if you act on it. Here's what each one means in practice:

Three conversations that need to happen.

If you're a junior analyst feeling uncertain, show this to your manager.

Your anxiety isn’t irrational, but the conclusion that your career is disappearing probably is. The tasks are changing. The role is being rewritten, not deleted. What you should be asking is: what does the rewritten version of my role look like here, and am I being given the chance to grow into it? If your organisation isn't asking that question, that's important information about whether it's the right place to build your career.

(And I promise this isn't a sales pitch - we built a little free app you can try if you need some help thinking it through.)

If you're a manager, have this conversation with your team.

They're wondering. They might not be asking out loud, but the quiet anxiety about relevance is real and it's likely affecting engagement and retention. The most valuable thing you can do right now isn't give them a pep talk. It's being honest about which tasks are changing, involving them in reimagining their roles, and making it clear that their judgment, context and relationships are not the things AI is coming for. Ask them where they want to grow, then make sure you support them.

If you're a leader making decisions about your analytics pipeline, this one's for you.

The short-term economics of not backfilling junior roles are seductive; the long-term economics are terrible. Someone needs to govern, direct and make sense of what the AI is doing. And every senior person in this room knows AI is only as good as the data it's built on - yours is probably not as clean as you'd like to admit. The people who know where the bodies are buried learned that by being junior first. Don't delete your training data.

A final thought

The leaders who see AI as a reason to thin the herd and the practitioners quietly wondering if they have a future are in the same room. The only thing standing between now and a better answer is the conversation. So, what are you waiting for?

About the author

Helena Turpin is Co-Founder and CEO of GoFIGR, a workforce intelligence platform that helps organisations understand how AI will impact their people through task-level analysis. She writes about the future of work, AI, and what it means to build organisations that are good for humans. Find her on Substack and LinkedIn.