What are the types of decision support systems?
A widely used taxonomy sorts them by what drives the analysis.
| Type | What drives it | Example |
|---|---|---|
| Data-driven | Queries over company data | A sales dashboard showing margin by region |
| Model-driven | Financial or optimization models | A pricing simulator |
| Knowledge-driven | Rules and expert knowledge | A drug interaction checker |
| Document-driven | Search over documents | A contract clause finder |
| Communication-driven | Group collaboration | A shared scenario planning tool |
Most real systems mix several types, and the AI-era versions lean heavily on the document-driven kind, because language models finally made unstructured text usable as input.
How does AI change a decision support system?
Three ways. It reads what older systems could not, such as emails, scanned forms and call transcripts, through document AI. It forecasts, through predictive analytics. And it explains, drafting the reasoning behind a recommendation in plain language. AI makes decision support broader and faster, but it also makes it easier to trust a recommendation without checking it. A system that writes a confident paragraph invites the reader to skip the numbers underneath.
Decision support vs decision automation: where is the line?
The line is who commits the action.
| Decision support | Decision automation | |
|---|---|---|
| Who acts | A person, after reviewing | The system |
| Person’s role | Decides every case | Reviews samples or exceptions later |
| Error caught | Before it happens | After it happens |
| Fits | High-stakes or rare decisions | Frequent decisions with a low override rate |
| Move the line one decision at a time, not one system at a time. Start with support, measure how often people overrule the recommendation, and automate only the cases where they almost never do. An override rate is the best evidence you have that a decision is ready to automate. The control that keeps this safe is a human in the loop on everything above a risk threshold. |
What makes a decision support system trustworthy?
It shows its sources, states its uncertainty and records what was recommended against what was decided. Without that record, no one can tell whether the system helps or just agrees. AI explainability covers how a recommendation can carry its reasons, and AI decision-making covers who stays accountable when it is wrong.