Everyone Becomes the Analyst

Why “will AI take the analyst’s job?” is the wrong question — and what’s actually happening to the people who ask

A question lands in the team’s analytics channel on a Tuesday — did the new onboarding flow actually move activation, or are we imagining it? — and three people glance at it, none of them answers, and by Thursday it has scrolled out of view. Everyone in the channel knew who could have answered it: the one analyst, already two weeks deep in the roadmap’s committed work. So the question quietly joined the pile of things the team wondered about and decided not to find out. Every team has that pile. It’s where most of the decisions you’d want to revisit quietly go to die. When a tool arrives that can answer such questions in plain language, the first thing everyone wants to know is whether it means the end of the analyst. It is the wrong question — asked with real anxiety, and asked about every tool that ever made a hard thing easy. The framing fixes on one role and misses the bigger event: not what happens to the analyst, but what happens to everyone who used to wait on one.

Start with why the role existed at all. A company hires an analyst the way it hires anyone: to convert a recurring need into a function someone owns. The recurring need, in this case, is a steady stream of questions about the business — why did activation dip, which cohort is retaining, what changed after the release — that the people running the business cannot answer themselves, fast enough, on their own. The analyst is the person you route those questions to. They turn an unasked question into a defensible answer. That conversion is the job. Everything else — the SQL, the dashboards, the warehouse — is machinery in service of it.

Notice what that means. The analyst was never primarily a query-writer. The query was the cost of doing the real work, which was reasoning: deciding what the question meant, which dimension to break on first, which alternative explanation to rule out before trusting the number. The typing was the toll you paid to reach the thinking. For most of the history of software, that toll was unavoidable, because the systems holding the data could not reason about it. Someone had to translate intent into instructions the machine could run. So the function of turning questions into answers got concentrated in a person, because concentration was the only way to make the translation affordable.

That is the part worth sitting with: the role was a response to a constraint, not a law of nature. We did not centralize curiosity into a single calendar because that was the ideal way to run a company. We did it because the cost of asking was high, and high costs force rationing, and rationing forces a queue, and a queue needs an owner. The analyst is what a question queue looks like when it has a name.

When the cost of asking falls far enough, the logic that built the role comes apart. Not because the work disappears — the questions are as real as ever — but because the reason to concentrate the work in one person was the expense of doing it, and that expense is what changes. When anyone on the team can ask a question of the data directly and get a reasoned answer back, the function does not vanish. It stops living in a person’s calendar and starts living in the workflow. The conversion from question to answer still happens constantly. It just no longer has to be queued.

This is the distinction the replacement framing misses entirely. There are two loud stories on offer, and both are wrong. The first is the analyst gets automated away — the fearful version, which treats the role as a set of tasks. The second is the analyst gets a copilot — the comfortable, vendor-safe version, which treats the tool as an accelerant that leaves the org chart untouched. Neither describes what actually happens. What actually happens inverts the paradigm the role was built on: cheap asking collapses the queue that justified concentrating the function in one seat, and the capability spreads outward — into the hands of everyone who has a question. The thing only one person on the team could do becomes something the whole team can do. That is not a smaller role. It is a bigger, more capable team.

A capability that belongs to the whole team behaves differently from one rationed through a single queue, and the difference is most visible in what becomes scarce. When asking was expensive, the bottleneck was access — analyst time, the length of the queue, where your request sat in someone’s backlog. You did not get answers because you could not afford them. When asking is cheap, access stops being the constraint. The new bottleneck is curiosity: knowing what to ask, and caring enough to ask it. The scarce resource moves up the stack, from the supply of answers to the demand for them — from whether you can find out to whether you think to.

This sounds like a smaller change than it is. A team that has spent years rationing its own questions has learned, quietly, not to ask most of them. The discipline of not bothering the analyst with a hunch becomes a habit, and the habit outlives the constraint that created it. So the first thing that has to change is not the tooling but the reflex — the team has to relearn that a passing question is allowed to be asked, because asking is finally cheap enough to do on impulse. The decisions that used to get made on unaudited intuition, because finding out was not worth a day of someone’s time, can now be checked in the moment. That is the real shift: not faster answers, but a lower threshold for what is worth finding out.

Now the strongest objection, because it is a fair one. An analyst does not just run queries — they exercise judgment. They know that a metric definition changed last quarter, that a number is suspicious, that the obvious explanation is usually wrong. Surely that judgment is exactly what a tool cannot carry, and surely that is the part of the role that no one else can pick up. The objection is right about the value and wrong about the conclusion. The judgment does not vanish, and the tool does not magically acquire it. The judgment spreads — to whoever is now asking. When a head of product can interrogate a number directly, follow up, and rule out the boring explanation themselves, the judgment that used to live with one analyst starts to live with the person closest to the decision. Some of it is carried by the system, which can reason about a question the way a good analyst would. The rest is carried by the asker, who finally has enough at-bats to develop the instinct. Cheap asking is how non-analysts become analytical. The judgment doesn’t survive in spite of this; the team grows it because asking got cheap.

There is a sharper version of the same worry, and it deserves a direct answer: cheap asking does not guarantee good answering. Open the floodgates and you risk a flood of confidently wrong conclusions — a metric whose definition quietly drifted, a correlation mistaken for a cause, the same number computed three ways by three teams. The fear is that spreading the capability trades a slow, careful function for a fast, careless one. But notice that this is an argument for the analyst’s judgment, not against spreading it. The thing you want to spread is precisely the discipline the analyst carried — define the metric before you trust it, distrust the obvious explanation, check whether the number is even real. Done right, this is not removing that discipline; it is making it the default behavior of the system everyone asks, and the habit of everyone asking. The shared definitions still need an owner. What changes is that defending those definitions, and curating what “trustworthy” means, becomes the high-leverage work — rather than personally running every query that depends on them. Governance is not the casualty here. It is what this looks like when it is done well.

And the analysts themselves? They are not the casualty of this — they are the ones who finally get to do the part of the job that was always the point. When the routine conversion of questions into answers stops needing a dedicated queue, the people who were doing it move to where deep expertise still pays — the genuinely hard investigations, the data model everyone else now relies on, the questions that are hard not because the query is hard but because the framing is contested. The role does not shrink. It climbs. The same thing happened when spreadsheets put financial modeling on every desk: the controllers did not disappear, they stopped doing arithmetic and started doing finance. Spreading the routine work frees the expert to do the part that never should have been a queue.

Step back and the pattern is older than this particular tool. Every time a costly capability becomes cheap, it stops being a department and becomes an ambient skill of the whole team. Typesetting was a trade; now it is a menu. Web publishing was a team; now it is a button. The capability does not leave the company — it leaves the bottleneck. Product analytics is arriving at the same threshold. The capacity to ask the business a question and get a reasoned answer is moving from something you requisition to something you simply do, the way you check a calendar or search a document.

So the real story is not what happens to one role. It is that everyone else gets good at something only one person used to be able to do. The expensive part — asking — is becoming free, and the new scarce resource is the willingness to ask, which the best teams will learn to spend freely. A team that no longer has to ration its own curiosity is a team where everyone, quietly, becomes the analyst.

Key Takeaways

• The analyst’s real job was never writing queries — it was converting unasked questions into trustworthy answers. The query was the toll, not the work.
• That job was concentrated in one person because asking was expensive, and expensive things get rationed and queued. The analyst is what a question queue looks like with a name on it.
• When asking gets cheap, the capability spreads — the whole team gets to do the thing only one person used to do.
• The bottleneck moves from access to curiosity: not whether you can find out, but whether you think to.
• Analysts aren’t the casualty — they climb to the hard problems while everyone else, quietly, becomes the analyst.