The next competitive divide will not come from who buys the best model. It will come from who converts their own business judgment into intelligence.

The last great conversion

We have seen a version of this before. Financial services firms became fintech firms when they learned to turn a business process into a technology service. Payments, lending and onboarding stopped being back-office procedures and became products.

Intelligence is a far more powerful conversion. Fintech codified the process. Intelligence codifies the judgment inside the process: which claim to fast-track, which batch will miss yield, where order-to-invoice is quietly leaking. A codified process stays roughly the same once built. Codified judgment compounds, because every decision and outcome makes it sharper.

Same platforms, different winners

Most business domains already run on a small set of common platforms. Wealth management onboards clients on shared wealth platforms. P&C insurers run claims and policy on Guidewire. Retailers run on Manhattan and similar suites. Manufacturers run on PLM, SCADA and Oracle order management.

The processes are common too: onboarding, claims, policy administration, order-to-cash, production scheduling. Development teams reuse the same open-source components across all of them.

Yet the outcomes are wildly uneven. Facebook figured out social networking better than MySpace. Fuji navigated the shift Kodak missed. Same era, same technology, different results.

The platform was never the differentiator. How an enterprise runs its processes on that platform is. That difference is visible in the data, if anyone looks.

Master intelligence

When we built an AI-native operations platform, historical intelligence surfaced something we had not set out to find. Across enterprises on the same platforms, some had already reached a far higher level of efficiency. They got there through changes in process and configuration, not by buying different technology.

That best-observed efficiency became a master intelligence for the domain. It worked as a loop:

1. Learn the benchmark. Historical data (tickets, logs, workflows and more) reveals the most efficient way a process actually runs on a given platform.

2. Evaluate new data against it. Each new set of operational data is compared with the master intelligence to find the gap.

3. Generate actionable intelligence. The gap becomes a concrete recommendation: eliminate the work, hand it to an agent, or assist a human.

4. Learn from human feedback. People explain when a recommendation is not practical for their enterprise, and the intelligence absorbs that context.

This held across every layer we worked in: business process, application, data and infrastructure. The same approach drove 60–80% work elimination and 35%+ cost takeout in production.

Two examples from the field

The order that couldn’t ship. At several manufacturers on Oracle order management, users were not permitted to delete order line items. Every deletion needed an IT ticket, and those tickets sometimes took two to three days. Meanwhile, orders sat unready for shipment.

Historical intelligence spotted the same pattern across companies. It also found one enterprise that had solved it with a change to how the underlying Oracle platform was configured. Applying that pattern eliminated the ticket entirely. Resolution time dropped to zero and orders were ready to ship immediately. This was a business process problem and an application problem at once, and no amount of faster ticket handling would have fixed it.

The slow data build. In another case, the intelligence quickly traced a sluggish data build to how batch jobs were scheduled and the fact that they were not monitored. Nobody had connected the two. Fixing the schedule and adding monitoring removed the slowness at its source.

We saw many more like these in the infrastructure layer. The pattern was the same each time: the fix already existed somewhere, hidden in the data.

Your own history already holds it

You do not need a cross-industry dataset to start. Every enterprise has its own “good is not good enough” cases buried in its historical data.

They show up as recurring tickets that should never exist, as manual approvals that add days without adding control, and as workarounds teams have normalized. They also show up as the best-performing plant, region or team, whose way of working never spread to the rest.

That internal best is your first master intelligence. Measure everything else against it, and the gap becomes your roadmap.

Market intelligence: answer “faster and cheaper” before it’s asked

In almost every engagement, the business eventually came back with the same request: can you do this faster and cheaper? When we looked closer, that request was rarely arbitrary. It usually meant a competitor had already achieved it.

So we started building market intelligence for each domain alongside the historical intelligence. It turned out to be one of the most practical data points we could offer. We could show a wealth management firm that a competitor was onboarding assets in half the time, a P&C insurer that rivals were settling claims faster, and a manufacturer that peers had much shorter order-to-delivery times.

That changes IT’s posture from reactive to proactive. Instead of waiting for the business to turn a competitive gap into a complaint or a requirement, intelligence can surface the gap first, show the efficiency level that is already possible, and arrive with the path to close it. In an age when everyone is talking about super intelligence, this is the minimum standard: know where you stand before the market tells you.

The benchmarks have to be credible. Draw them from public disclosures, industry studies, customer-facing service levels and aggregated, anonymized operational data, never from any single company’s confidential records.

Why boards aren’t driving it yet

Few boards or CXOs are treating intelligence as a business asset to build. Four framing problems keep it small:

  • Wrong owner. AI sits with the CIO as a tooling program, not with the P&L leader as a business capability.
  • Wrong metric. Success is counted in copilot seats and hours saved, not in combined ratio, yield, DSO or revenue leakage.
  • Wrong starting point. Companies pick a model first and hunt for use cases later, instead of mining their own decisions to find where intelligence pays.
  • No product mindset. Nobody owns intelligence as a living asset that is versioned, measured and improved.

Where to start

1. Pick one KPI that matters to the board. Claims cycle time, yield, order-to-cash leakage or DSO. Make the P&L leader the owner.

2. Build historical intelligence first. Mine tickets, logs, workflow and transaction data to establish your internal best and the gap to it.

3. Benchmark against the market. Know where competitors already are on that KPI before the business asks why you aren’t.

4. Eliminate before you automate. Remove work that should not exist, hand repeatable decisions to agents, and assist humans on the rest.

5. Lead with change management. Actionable intelligence only pays when the operating model changes with it: transform, then transition.

6. Treat the intelligence as a product. Version it, measure it against the KPI, and feed human feedback back into it.

The moat is not the model

Models are becoming commodities. Open-weight or frontier, they are increasingly interchangeable engines. What no competitor can download is the intelligence built from your own decisions and outcomes.

Every company became a software company. The next winners will become intelligence companies, but only the ones that own and build on their decision data. The platforms are the same for everyone. The question is who starts building first.

This article is part of my Enterprise Intelligence series on building intelligence into every business process, application, data and infrastructure layer to drive peak business performance. Earlier: why the forecast is free now but your history isn’t and why enterprise intelligence hasn’t started yet.