What is Decision Intelligence?

The clean definition of Decision Intelligence comes from what it must produce. A platform earns the “DI” label when its output ships with all four of these:

1. A recommended action.Not a chart, not an insight, not an anomaly flag. A specific next move — sized, scoped, and assigned. “Markdown the 200 units of Category A in the Northeast stores by 15% by Friday” is an action. “Inventory is up 12%” is not.
2. A quantified impact.Sized in dollars, ex-ante, before the action ships. The projection should be grounded in segment-level historical data, not pulled out of an LLM’s imagination. Reconciliation against actuals after the action becomes the next training signal.
3. A validation record.Three or more alternative strategies, scored against the operator’s proposal. Diwo uses a High Confidence / Maximum Reach / Optimized triad. The point isn’t that the AI picks for you — it’s that the human picks knowing what they’re trading off.
4. An execution pathway.The approved decision lands in the system of record — Salesforce, Slack, Microsoft Teams, Mailchimp, ERP, ticketing — not in a slide deck or a meeting follow-up email. Without execution, every preceding step is theatre.

A platform that produces three of these but stops at the fourth is augmented analytics. A platform that produces all four is Decision Intelligence. The category boundary is that precise.

The lineage matters because every category that came before DI still exists, still works, and still gets confused with it.

• Business Intelligence (1990s)— aggregate, model, visualize. Output: dashboards, cubes, scheduled reports. The user interprets.
• Predictive analytics (2000s)— forecasts, segmentation, classification. Output: probabilities and scores. The user decides what to do with them.
• Augmented analytics (2017, Gartner)— AI accelerates the analyst’s discovery loop with NL search, auto-insights, smart visualization. Output: faster insights. The user still has to act on them.
• Decision Intelligence (early 2020s, formalized 2024 by Gartner + IDC) — everything above, plus the action layer. Output: a decision, executed.

Each generation kept what came before. DI doesn’t replace BI any more than augmented analytics replaced BI: it adds a layer that closes the loop between insight and outcome. The reason it took until the early 2020s for DI to emerge is that the prerequisite technologies (modern data warehouses, vector databases, large language models, agentic orchestration, semantic knowledge graphs) only matured into enterprise-ready form recently.

Most DI platforms organize work around a six-stage flow. Different vendors use different names; the architecture is consistent.

1. Detect. The platform scans every metric in your warehouse against seasonal baselines, segment patterns, and industry signals, and ranks what matters in dollars. Output: a queue of opportunities, ordered by impact.
2. Explain.Each opportunity arrives with an AI-authored briefing — what it is, why it matters now, what specifically to do. The operator reads a position, not a chart.
3. Understand. The opportunity is decomposed into the channels, products, segments, and stores driving it. Concentration risk and long-tail upside both surface instantly. Strategy decisions get anchored in the segments that matter, not averages.
4. Simulate.A what-if engine lets the operator move levers — conversion rate, segment mix, promotion depth, impact range — and watch the quarterly dollar impact update in real time.
5. Choose.Three AI-generated alternative strategies (High Confidence, Maximum Reach, Optimized) are scored against the operator’s proposal. The operator picks the one they’ll defend.
6. Certify. The chosen decision is logged to the audit trail with the alternatives considered, the projected impact, the approver, and the executed action. A print-ready brief is generated for the standup or board deck. Outbound agents push the action to the system of record.

The output of every run isn’t a tile on a dashboard — it’s a decision in the system, traceable from hypothesis to outcome.

The DI pattern is industry-agnostic, but the decisions it’s used to make are concrete.

• Retail. Daily merchandising decisions on markdowns, replenishment, assortment, regional pricing. Catalyst alerts a regional merchandiser that home goods turnover dropped 8% QoQ because of replenishment lag at the new Atlanta DC, recommends re-routing via Memphis for 30 days, projects $285K recovery, and pushes the routing change to the WMS.
• CPG. Trade promotion lift, distribution expansion, pricing-pack architecture. The system flags a promotion underperforming its forecast in three retailers, proposes a one-week reset of the promo mechanic, and pushes the brief to the brand manager and the broker.
• Financial services. Cardholder spend activation, attrition, fraud, credit risk. The platform surfaces a Gen Z card-activation lift opportunity, scores three campaign approaches, and pushes the approved audience to the marketing automation tool.
• Healthcare. Patient flow, no-shows, capacity. The system identifies clinics with rising no-show rates, proposes targeted reminder sequences scored against control, and pushes the segment to the patient engagement platform.
• Manufacturing & ops. Yield, demand forecasting, supply chain. The system detects a supplier risk pattern from upstream signals, proposes a dual-source realignment, and tickets it to procurement.

Use these six tests when shortlisting platforms. Any platform that fails any of them is augmented analytics, not Decision Intelligence.

1. Does the output include a recommended action with quantified dollar impact? Or does it stop at an insight or chart?
2. Does it produce alternative strategies, AI-scored, before commit? Or is the operator left to come up with options manually?
3. Is there a built-in what-if simulator? Or does scenario modeling require a separate tool?
4. Does it push approved decisions into operational systems? Or do operators export to spreadsheets and ticket them manually?
5. Is every decision logged to an audit trail tied to outcomes? Or is there no trace of what happened next?
6. Is the vendor named in the Gartner Market Guide for Decision Intelligence Platforms (2024) or the IDC MarketScape for Worldwide Decision Intelligence Software (2024)? Both reports established the category and named representative vendors. Diwo is named in both.

These aren’t arbitrary tests. Each maps to one of the four required elements of a decision (action / impact / validation / execution) plus the audit + analyst recognition signals that prevent vendor over-claiming.