Recently I read yet another article claiming that ChatGPT had “killed” a $100k BI analyst role. The argument was familiar: AI can now write SQL, build dashboards, and summarize performance faster than most analysts can finish writing a job description. A six-figure hire replaced by a monthly subscription.

It was dramatic. Convincing. A little terrifying. But as I sat with it, I kept wondering: if AI can now write the SQL, build the dashboard, and summarize performance faster than most of us can finish a coffee… why do so many meetings still end with “Let’s review this again next week.”

If analysis is instant and hesitation remains, the uncomfortable truth probably isn’t about AI at all. It’s about the way our dashboards are (or aren’t) structured to support decisions.

AI can now generate insights and likely drivers in seconds. In many cases, it can explain performance more clearly than a junior analyst. But if those insights don’t translate into a clear decision, then all we’ve really improved is reporting efficiency.

Dashboards were never meant to compete on speed alone. Their job was to help a team decide what to do next.

When AI writes the query, what’s left?

For years, slow decisions were blamed on slow analysis. We said we needed more data, more breakdowns, more charts. “Once the numbers are ready, then we can decide.”

Now that tools like ChatGPT can generate SQL, assemble visuals, and even produce narrative insights in minutes, that excuse disappears. The dashboards arrive on time. The variance is explained. The key drivers are highlighted.

And still, the meeting ends with the same sentence: “Let’s monitor this and review again next week.”

That’s the moment where the story “AI killed the analyst” starts to feel incomplete. Because if the analysis is fast and the dashboards are clear, then the real bottleneck isn’t the analyst. It’s the decision structure that was never designed.

If this feels uncomfortably familiar, you might recognise some of the patterns described in the Symptoms for when dashboards don’t help you decide →

“Actionable dashboards” optimise for noticing

Most teams would describe their current setup as an actionable dashboard:

• KPIs are clean and easy to read.
• Variance is highlighted with colour and icons.
• Trends are visualised with line charts and sparklines.
• Drill-throughs and filters are available for deeper analysis.

These dashboards optimise for noticing: they help people see what changed and how big the change is. That’s valuable, and it has been the focus of most “actionable dashboard” tutorials for years.

But noticing is not the same as deciding. A team can see the numbers clearly and still hesitate, because the dashboard hasn’t answered three uncomfortable questions:

• What matters most? When signals conflict, which metric wins?
• When is it enough to act? At what point does a change move from “interesting” to “urgent”?
• What are we choosing between? Which concrete actions are on the table this week?

A decision-ready dashboard doesn’t just visualise performance. It builds a shared structure around those three questions so that, when AI delivers the insight, the team already knows how to move.

I explore the full distinction in more depth in the guide “Actionable Dashboard vs. Decision-Ready Dashboard” →

The real gap isn’t analysis – it’s decision structure

On paper, the journey looks simple: Data → Insight → Action. But in real organisations, there are several invisible steps in between:

Data → Interpretation → Attention → Trigger → Framing → Decision → Action.

AI is extraordinarily good at the first part of that chain: interpretation. It can summarise, cluster, forecast, and suggest likely drivers.

Where AI does not (yet) operate is in the structural choices:

• Attention: which metric or driver gets looked at first when time is short?
• Trigger: what rule turns “different” into “enough to act”?
• Framing: which set of actions is considered realistic this week?

Without those elements, the dashboard may be accurate and attractive, but it doesn’t create a decision moment. It generates discussion, not movement.

This is why I describe decision-ready dashboards as having a specific backbone: a North Star, driver ranking, weekly triggers, and action framing. You can see that full structure here: Decision Structure Guides →

When analysis becomes effortless, hesitation becomes obvious

Before AI, delays and frustration could be blamed on reporting: “We don’t have the dashboard yet”, “We’re still waiting on the numbers”.

As analysis time shrinks from days to minutes, those excuses evaporate. What we see instead is the quiet reality of many performance meetings:

• The team notices a change.
• Everyone offers a different explanation.
• Responsibilities and priorities collide.
• The safest option is to watch and wait.

AI didn’t create that pattern. It just made it impossible to hide. Once the “data work” is solved, the remaining friction is judgment: how people choose, what they are optimising for, and what they are willing to trade off.

And judgment, unlike SQL, is not something we can fully outsource. What we can do, however, is design dashboards that acknowledge judgment and give it a structure to operate within.

BI isn’t dead. Its center of gravity is moving.

The version of the BI analyst role that focused on building reports and hand-crafting queries will shrink. AI and modern BI platforms are already automating much of that execution.

But another version of the role is emerging – one that looks much closer to decision architecture than dashboard production:

• Clarifying the North Star and what the dashboard is protecting or improving.
• Designing driver tables and ranking rules that focus attention where it matters.
• Defining triggers and thresholds that turn noticing into acting.
• Framing actions and trade-offs so the team can move without re-analysing everything from scratch.

That work is not “killed” by AI. In fact, AI makes it more visible – and more valuable – because the easy parts of analysis are no longer differentiating.

If you want to see how this looks in a real Power BI report, there’s a full example in the Decision-Ready Premium Sales Dashboard template →

Beyond “Problem → Insight → Action”

Many storytelling frameworks for dashboards follow a familiar flow: Problem → Insight → Action. It’s a helpful starting point.

But in live decision meetings, the story rarely plays out that cleanly. People arrive with different priorities, different levels of risk tolerance, and different incentives. An insight that looks obvious on a slide becomes complicated when it runs into capacity constraints, budget ceilings, and political realities.

Decision-ready dashboards don’t ignore that complexity. They are built to surface it:

• The North Star keeps everyone anchored on what the dashboard is protecting.
• Driver ranking shows which levers are actually moving the result this week.
• Triggers and thresholds define when “different” becomes “enough to act”.
• Action framing makes the trade-offs explicit, not implied.

In other words, the framework isn’t just about telling a clearer story. It’s about designing the structure that helps a team move even when the story is uncomfortable.

A harder question – and a more useful one

So perhaps the better question isn’t whether ChatGPT will replace BI analysts.

The better question is this:

When AI can generate the insight in three seconds, what will your organisation do in the next thirty minutes?

If the answer is still “Let’s review this again next week”, then the core issue isn’t the analyst, the tool, or even the data. It’s the absence of a shared decision structure that turns insight into co-ordinated action.

AI didn’t kill the BI analyst. It simply made it impossible to hide behind analysis anymore. And for teams willing to redesign their dashboards around decisions, that’s not a threat. It’s an invitation.