Every enterprise we talk to today has an AI initiative running somewhere — a chatbot pilot in IT support, a copilot embedded in the CRM, vendor invoices being read and created, etc. Ask what problem it solves, and the answer usually centers on effort: fewer hours spent searching, faster ticket resolution, less manual data entry. All these are cases of task automation. Useful, often. But there’s a second category that requires intelligence to be put into, one that comes up less often in these conversations — making decisions.

Part of why tasks get more of the attention is simply that they’re easier to see. A task lives inside one screen, one workflow step — easy to point at, time, and show progress against. A decision often runs across several systems and several people, visible only in the outcome it eventually produces. It’s natural that conversations gravitate toward what’s easiest to observe, even when it isn’t where the larger value sits.

The difference is worth assessing. A task is something a person does. A decision is something a person chooses. A task, once automated, saves the time it used to take — a real but bounded source of value, fixed at roughly what the task cost before. A decision, once improved, changes what happens next, and that effect travels forward. The same decision gets made again next week, and the week after, each time carrying the improvement with it. Effort saved has a ceiling. Better choices, repeated at scale, generally don’t.

This is where a more deliberate lens helps — thinking in terms of decision intelligence rather than task automation. In practice, that means looking at the decisions a business makes regularly and asking a handful of questions about each one.

  • How often does it get made.
  • How reversible is it once made.
  • How complete is the information available at that moment.
  • How long does it currently take.
  • What does a poor version of it cost.
  • And does getting it wrong tend to cascade into other decisions downstream.

A decision made often, hard to reverse, made on incomplete information, and expensive when it goes wrong is usually the kind worth close attention. A decision made twice a year, on complete data, with a low cost of error, is probably well served by a straightforward tool (could even be excel) already in place. The exercise is short, but it tends to reorder a list of priorities fairly quickly — from “what could be automated here” to “which of these decisions, improved even modestly, would actually move something the business cares about.”

Decision intelligence also has its own maturity, and it’s worth being precise about how it develops. There are four stages (we call them 4 Ds):

  • Discover - simply seeing what’s happening as it happens;
  • Diagnose - understanding why it happened;
  • Decide - working out the best course of action among the options available; and
  • Do, carrying that action out.

The first two stages mostly produce insight — a clearer picture of a situation, without much support for the choice that follows. Real decision support only enters at the third stage, when the system starts weighing options (even doing simulations) rather than just describing what’s in front of it, and it’s fullest at the fourth, when the chosen action is carried out as well.

An enterprise that has built only the “Do” — the execution step — without the discovering, diagnosing, or deciding that should come before it, has automated a task, not improved a decision. That’s still worth having. But there isn’t much intelligence in a system built this way beyond how faithfully it carries out an instruction, and the benefits it produces tend to stay modest, capped by the volume of that task.

Task automation and decision intelligence aren’t competing priorities, and most organisations will keep doing a fair amount of both. But it’s worth occasionally turning the question around — not only “what tasks could we automate” but “which decisions, made a little better, would move something that matters.” The second question gets asked far less often. It’s usually where the larger opportunity has been sitting all along.