I’ve spent a lot of time this past year in conversations with controllers and finance leaders who are living in an uncomfortable middle ground. Their teams are already using AI. Claude for financial statements. ChatGPT to read contracts. Automation scripts running parts of the close. The productivity gains are real and the teams love it.
But when I ask whether that work is ready to survive an audit, the room gets quiet.
That’s the problem this release is built to solve. We’re already seeing that AI can help finance teams move faster. The harder question is whether that work is actually defensible to your internal team, to external auditors, and to your CFO when they ask how you know the output is right.
The Spring 2026 Release is our answer to that question. Here’s what your team can do with it.
Automatically Govern Every Workflow
According to Grant Thornton’s 2026 AI Impact Survey, 78% of business executives lack confidence that they could pass an independent AI governance audit within 90 days. I’ve seen why firsthand.

The most common workaround I see in finance organizations is a manual compliance regime that lives outside the platform. A spreadsheet tracking which workflows ran. An email chain documenting approvals. A manually reconstructed change log when auditors ask what the process looked like three months ago.
That approach made sense when work was done manually. It was simple to collect audit evidence, step-by-step, as the work was done. Automation changes that. The workflow runs on its own, but someone still has to manually track what it did, when it was done, how it was done, what changed and why. The time saved on the work itself gets spent reconstructing the evidence trail around it. You haven’t eliminated the manual effort. You’ve just changed what it looks like.
That’s what Version Controls, Sign-offs and Approvals, and Evidence Logs are built for.

When a financial process runs, your team needs to be able to answer a precise set of questions. What business logic was applied? What inputs did it use? What did it produce? And if any process step changed between this month’s run and last month’s, what exactly changed and when? That’s not just an audit requirement. It’s the foundation of having any confidence in your financial processes at all. Without it, we’re back to taking someone’s word for it. Back to the pre-SOX days of Enron.
Version Controls make that possible. At any point in time, you can point to the specific business logic that was used and exactly what changed between this month’s version and last month’s. Combined with Savant’s Run History, you can also answer when it ran, who ran it, what inputs were used, and what the results were. This is how finance teams build confidence in their own processes. And how external auditors build confidence in them too. When auditors ask what the intercompany reconciliation flow looked like at Q3 close, you pull up that version. One click. The answer already exists.
Once you have a healthy version control process in place, you need controls over who is working on those versions. That’s where Sign-offs and Approvals come in. An accountant, tax specialist or analyst can build or edit a workflow, whether it’s v1 or v10, and send it off to their manager for approval before it moves into production. True and simple segregation of duties between the process builder and the process approver. Finance organizations get real control over how their automation is developed, modified, maintained, and deployed.
Finally, where would SOX controls be without evidence collection? Every time someone logs in, edits a process, makes an approval, changes a role, your auditor needs proof it happened. That’s where Savant Logs come in. Detailed, immutable logging is automatically built into the system from day 0. No need to design custom, complex, head-splitting ticketing and IT systems to manage and collect these details, because they’re natively in Savant as datasets. Queryable and reportable like any other data source, with prebuilt templates covering the most common SOX and ICFR requirements. And as a bonus, they’re great for platform usage analytics too.

Process Unstructured Data From Finance Documents Without Prompt Engineering
Document-heavy processes are still where some of the biggest manual effort lives in finance teams. Financial statements. Lease agreements. Vendor invoices. Bank statements. Tax filings. Most teams have experimented with LLMs to read them. Few have operationalized extraction at scale.
The barrier isn’t model capability. It’s the setup.
Getting an LLM to reliably extract structured fields from a thousand invoices with different layouts requires difficult prompt engineering that most finance teams shouldn’t have to develop. So experimentation stays as experimentation, and analysts keep doing the extraction by hand.

Vision Agent Skills remove that barrier entirely. We’ve built a library of prebuilt skills for the document types finance teams actually work with: invoices, bank statements, purchase orders, contracts, receipts, and more. Each skill is pre-engineered to extract the structured fields your workflows need. Pick a skill, adjust the fields if needed, and run it at scale against tens of thousands of pages on day one.
For anything that doesn’t fit a standard skill, AnyDoc handles extraction without requiring you to engineer a prompt from scratch. The logic stays visible and editable. Nothing runs as a black box.
Run Your Largest Workloads Inside the Close Window
Processing speed doesn’t sound like a governance problem. But in practice, it is.
When a tax provisioning workflow takes six hours to run, teams work around it. They run it less frequently. They run it on smaller datasets. They make decisions based on prior runs because rerunning to validate a change isn’t practical. That’s how automation that technically works ends up contributing to control risk.
The Savant Lightning Engine is a ground-up reinvention of Savant’s compute architecture. Across typical finance workloads, it delivers approximately 5x faster processing than our prior generation. Our benchmark testing shows it runs more than 10x faster than leading cloud Spark engines on standard data prep tasks.

Workloads that previously required overnight runs now finish before the workday ends. Large consolidations that created close-cycle pressure finish faster and more predictably. The engine scales automatically, so small workloads get fast lightweight processing and massive workloads get full compute resources without manual configuration. For teams running under compressed reporting timelines, that reliability matters as much as the speed itself.

Add Location Intelligence to Your Workflows
Some of the most important business decisions have a geographic dimension that standard analytics tools were never built to answer. Where should we put the next warehouse or factory? Which neighborhood makes sense for the next store? How does our supply chain exposure change if we shift distribution? These are questions finance and supply chain teams work through together, and the answers live in location data that most platforms can’t touch.
Geospatial Analytics brings that capability directly into your Savant workflows, without custom development or specialist tools. Turn raw location data into real-world geographic analysis in a few clicks. Model drive-time trade areas, eliminate overlap, assign nearest locations, and join datasets by proximity, using no-code templates designed for finance and operations teams. Results export directly to Tableau, Power BI, or any external visualization tool your team already uses.

See How It Works in Your Environment
The teams I talk to aren’t waiting for AI to be ready. They’re already testing and piloting it. What’s been missing is the infrastructure to make that work defensible to auditors, to leadership, and to themselves.
That’s what this release delivers. Governed workflows, document extraction at scale, and faster processing, all built for the way enterprise finance teams actually operate.
For a closer look at everything in this release, visit: https://savantlabs.io/new