Most organisations don’t lack data. What they lack is a clear way to turn it into a decision. Using an insurance repricing room as the example, this blog explains why traditional BI stopped at insight. It sets out how decision intelligence closes the gap: map the decision first, show the options and their trade-offs, and match each decision to the right mode (human deliberation, human oversight or full automation). It also argues that explainability is what makes accountability possible, and that trustworthy data is the work itself rather than a precondition. The organisations that lead will be the ones that close the loop between decision and outcome.
From Insight to Decision
Every year, inside insurance companies across the country, pricing decisions get made, often, unfortunately, still the hard way.
Claims data. Provider inflation figures. Actuarial models. Spreadsheets versioned badly and in many cases partially understood by those who open them. The pricing manager, the actuary, and the CFO are all technically looking at the same data. They are not looking for the same thing.
Underneath it all sits a hard constraint: the government caps how much premiums can increase year on year. Which means being right about risk is not enough. The question is not “what changed” but “given what changed, given what we are allowed to charge, given the levers available, what do we actually do.” Adjust excess levels. Tighten category limits. Accept margin compression on some cohorts to protect retention on others. Every lever has downstream effects. None of it sits in a single report.
This is not a data problem. The data exists. It is a decision problem.
BI delivered the insight. It was never designed to deliver the decision. Everything that happened between the two, the meetings, the conflicting reports, the institutional knowledge locked in one person’s head, was left to the organisation to figure out. Most still are.
Why It Stopped There
The limits of traditional BI were not a failure of ambition. They were a product of architecture and timing.
Data lived on-premise. Building platforms that could collect and query it at scale was slow and expensive. Organisations built what they could afford: reporting focused on the questions they already knew to ask. When cloud removed that constraint, the tooling paradigm did not change with it. Dashboards multiplied. Self-serve BI promised democratisation and delivered fragmentation: more people with access to more disconnected views, still manually stitching together data from different reports to get anywhere near a decision.
The models that actually drove decisions lived with specialists. Everyone else held a report and waited for a meeting. Machine learning raised the analytical ceiling but kept the same gap between model and decision-maker. Natural language interfaces are now beginning to close that gap, allowing analysts to iterate conversationally without writing reams of queries. That is real and accelerating. But it only works if the data underneath is trustworthy and well-governed. In most organisations, that remains the gap.
What Closing It Actually Looks Like
When you design around the report, you ask: what data do we have and who needs to see it. When you design around the decision, you ask something different: what are we trying to decide, what constraints apply, what options do we have, and what does each one cost us.
Those are not the same questions. They produce fundamentally different systems.
A decision-oriented system maps the decision before it builds anything: who owns it, what it requires, what constraints apply, what levers are available. It does not produce a dashboard. It produces options, with trade-offs made visible. In the insurance pricing room that means: here are three approaches within the regulatory cap:
- Option one optimises for margin.
- Option two optimises for retention.
- Option three balances both, with these implications for claims exposure. The actuary’s model stops being a black box that produces a number. It becomes a shared surface that produces choices.
There is not always a right answer. That is the point. The job of a decision intelligence system is not to tell you what to do. Yet. It is to make the trade-offs legible so that the people accountable for the outcome can decide with clarity rather than confusion.
In practice this means conversations with business leaders that start differently. Not “what do you want to see” but “what do you need to decide, and what is getting in the way.” Instead of mapping KPIs, you map decision flows: who decides, on what cadence, with what inputs, against what constraints. The data infrastructure is then built to serve those decisions, not to feed a reporting layer that stops short of them.
What It Takes To Do It Properly
Not all decisions are the same, and decision intelligence should not treat them as if they are.
- Some decisions require human deliberation by design. High stakes, competing objectives, no single right answer. The insurance repricing room is a good example. The system’s job is not to decide but to make the trade-offs legible: here are the options, here are the constraints, here are the consequences. Human judgement is the point, not the bottleneck.
- Some decisions carry enough downside risk that a human should review before anything executes. Fraud flags that lead to account suspension. Credit decisions. The system forms a view and recommends; a human remains the gate.
- Some decisions should be automated entirely. Volume too high, speed too important, stakes per decision too low for human review to add value. Dynamic pricing, inventory reorders, personalisation. Here accountability sits in the design of the system and the governance around it, not in approving individual decisions.
Most organisations have never consciously mapped their decisions to these three modes. They keep too much in committee out of habit, or automate things that warranted more oversight. Getting that mapping right is one of the most valuable things decision intelligence work can surface.
Accountability And Explainability
Across all three modes, explainability is not optional. It is the condition under which accountability is possible.
In committee mode, you need to see the assumptions to challenge them. In oversight mode, you need to understand the recommendation to approve or override it. In autonomous mode, you need to understand the system well enough to answer for its behaviour in aggregate. The form explainability takes is different in each case. The need for it is not.
“We did not know what the model was doing” is not a defence. It is an abdication.
The Foundation
None of this works without trustworthy data. In a world where data drives decisions directly, the cost of getting it wrong is not a misleading slide in a monthly review. It is a bad decision made in good faith, based on data nobody knew was wrong or assumptions that were never made explicit.
Most organisations are not there yet. Closing that gap is not a precondition to starting. It is the work itself.
The Loop
The organisations that will lead are not the ones with the most data or the most sophisticated models. They are the ones that have built the tightest loop between decision and outcome.
Every decision generates data. Every data point refines future options. Every outcome improves the next decision. That loop, consistently closed, is what separates organisations that are genuinely decision-intelligent from those that are still, despite everything, just reporting.
The frontier is not better dashboards. It is a fundamentally different relationship between data and action: one where the annual repricing cycle stops being an exercise in managing conflicting reports, and becomes a question the organisation is actually equipped to answer.
What To Do Next
If this resonates, then carry on reading! Our Analytics Modernisation and the Modern Data Leaders Blueprint will be of interest.
If your business is ‘drowning in data’ and has plenty of insights…but its not systematically driving decisions and outcomes – let’s talk. Book a free, no-pressure strategy session with one of our senior consultants. We’ll talk through where you are, what’s blocking progress, and what a practical path forward could look like.
