What Allstate’s Azure AI story tells us about the future of ERP

A few years ago, Allstate shared how they were using Microsoft Azure Cognitive Services to transform the way their customer service organization handles claims. The idea was straightforward but powerful: instead of relying on generic speech-to-text, they built custom speech models trained on their own insurance vocabulary—policy numbers, coverage types, claim codes, the messy real-world language customers actually use on the phone.

The result? Faster claims, less friction for the customer, and a quiet but meaningful shift in what frontline employees spend their time on. Agents stopped transcribing and started solving problems.

It’s a great story on its own. But what strikes me now, looking back at it from where we sit in 2025, is how clearly it previewed where ERP is heading.

The Allstate pattern, in plain English

Strip the marketing language away and Allstate did three things most companies are still trying to figure out:

They took a general-purpose AI model and made it speak their language. Out-of-the-box AI is impressive in a demo. It struggles the moment you put it in front of a real business process full of acronyms, SKUs, and tribal knowledge.
They embedded AI inside an existing workflow. Nobody at Allstate had to log into a separate “AI tool.” The intelligence showed up where the work was already happening.
They used the data they already had. Years of recorded calls weren’t an asset—they were storage costs. Until they became training data.

That’s the same recipe being applied right now inside modern ERP systems.

Why this matters for ERP in 2025

For a long time, “AI in ERP” meant a chatbot bolted onto a dashboard, or a forecasting feature buried three menus deep. That era is ending quickly. Microsoft, Oracle, SAP, and the rest of the major vendors are pushing AI into the core of the transaction flow—copilots that draft purchase orders, anomaly detection that flags suspicious AP batches before they’re paid, agents that reconcile sub-ledgers overnight.

The companies getting real value out of these features look a lot like Allstate did:

They have a clean enough data foundation that AI has something useful to learn from.
They tune models to their own context—their chart of accounts, their product hierarchy, their approval policies.
They redesign the job around the new capability, instead of layering AI on top of an unchanged process.

The companies struggling, in our experience, are usually stuck on the first one. Decades of customizations, half-finished integrations, and reports nobody trusts make it very hard for AI to help. You can’t automate what you can’t measure.

What finance and operations leaders should take from this

If you’re evaluating your ERP strategy right now—whether that’s a migration off a legacy system, a move to cloud, or simply deciding what to do with Dynamics, NetSuite, or whatever else is running the business—the Allstate example is a useful reference point. A few questions worth sitting with:

Where in our process is the equivalent of a “recorded call”—data we’re already generating that could power better decisions if we treated it as an asset?
Which workflows would benefit most from AI being embedded directly, rather than offered as a separate tool nobody opens?
Is our current ERP platform a foundation we can build on, or is it actively holding us back from adopting the capabilities our competitors already have?

None of these have a one-size-fits-all answer. But the leaders we work with who get this right tend to start with the business outcome—faster close, cleaner AP, better demand signals—and then ask which platform and which AI capabilities support it. Not the other way around.

The takeaway

Allstate’s story was never really about speech recognition. It was about a company recognizing that the boring, repetitive parts of customer service were full of signal, and that AI—properly trained on their context—could turn that signal into a better experience for everyone involved.

That same opportunity exists in finance and operations today. The technology is more accessible than it’s ever been. The question is whether your ERP foundation is ready to support it.

If you’re working through that question, that’s exactly the kind of conversation we have with finance and operations leaders every day. We’re happy to be a sounding board.