Summer 2026 Updates

Finance teams have stopped asking whether AI can do the work. What they keep hitting is the distance between an answer and a process – between something AI produced once, in a chat, and something that runs every cycle with a record an auditor will accept.

Savant's Summer 2026 release closes that distance with three major additions:

The theme isn't more AI. It's shortening the trip from "the AI figured it out" to "this runs every month, and we can show our work."

Every controller tells the same two-part story. The first part is genuine enthusiasm: someone on the team wired up an AI assistant, and work that used to eat three days now takes an afternoon. Nobody wants to go back.

The second part rarely makes it onto a slide. The person who built it is the only one who understands it. The steps live in a chat history. When the auditor asks how a number was derived, the answer is a screenshot and a good memory. The team got faster and less certain at the same time.

Speed was never the hard part. Defensibility is – to your own reviewers, to your external auditors, and to your CFO when they ask how you know the output is right.

The Summer 2026 Release is our answer to that. Here's what your team can do with it.

Your team already knows AI is fast. Hand it a reconciliation, a tax calculation or a forecast and something usable comes back in minutes. That part is settled.

What AI still leaves on your desk is everything that comes after the first answer. The number has to be right. You have to be able to show how you got it. And it has to hold up next month, and the month after that, when a different person runs it on a deadline.

That is the gap Savvy closes. Savvy connects Claude and Codex to Savant, so the work you talk through with your AI assistant becomes a governed Savant workflow – built in plain language, reviewed by a human, and run the same way every cycle with a full audit trail behind it.

Savvy is a plugin for the AI client your team already uses. You describe the work; Savvy builds it in Savant.

There is no new tool to learn and no pipeline to configure. If you can explain how the work should happen the way you would explain it to a new analyst, Savvy can build the process that does it – and will tell you where your explanation left something out.

Two things define this release.

Savvy is where the process gets built and refined. Savant is where it runs in production.
​

Accurate. Workflows run in Savant's deterministic engine, not in a language model. The same inputs produce the same outputs, run after run, with human review at the points that call for judgment.
​

Defensible. Every step is documented in business language as it is built, and every run is attributed to the signed-in user in Savant's audit trail. Nothing about your SOX (Sarbanes-Oxley) control environment changes, because what changed is how the workflow was authored, not how it executes.
​

Repeatable. Your process knowledge gets captured once and reused every cycle. The person who built it does not have to be the person who runs it.
​

Savvy is a new way to reach the Savant platform you have already approved – not a new system.

Full detail, including the questions security teams raise on a first plugin approval, is in the Getting Started with Savvy guide. Certifications and compliance documentation are on the Savant Trust Center.
​

Savvy is in limited availability. It is open to Savant customers today, with general availability to follow. Savvy requires a Savant account and the desktop version of Claude or Codex – the web clients are not supported.
​

Talk to your Savant team to get your organization enabled, or reach us at savantlabs.io.
​

Our summer release webinar walks through a live build – from a described process to a running, documented Savant workflow – in under an hour. Watch it below.

Lightning Builder gives you a faster, simpler way to build and validate workflows in Savant.

The biggest change is the new mental model. Instead of juggling Analyze, Test, and Run, you now have:

You can also refresh your data right from the canvas and test only part of a workflow while you build.

Test replaces the old Analyze experience. Before, Analyze only processed a limited amount of data. That meant you were not always seeing the full picture of how your workflow would behave.

Now, when you click Test, Savant processes the entire dataset.

That gives you a much better way to validate your workflow before writing anything to a destination.

Every time a Test runs, Savant stores those results in cache. That matters because if nothing has changed, Savant can pull those results back up instead of recalculating the whole workflow again.

So the first Test does the work, and the next one can be much faster. The cache also stays available across sessions. If your browser closes, you log out, or you open the workflow somewhere else, your Test results are still there.

Right now, cached Test results are retained for up to 7 days, as long as the workflow has not changed.

This is one of the biggest additions in Lightning Builder.

If you already ran a Test, Savant has already processed the data. So when you’re ready to send the results to a destination, you do not need to run the whole workflow again.

Instead, Run from Test uses the results from your most recent Test and writes them to the destination.

That means:

This is especially useful for larger workflows where a full Test can take a while.

A common pattern now is:

Run from Test always uses the most recent Test. It does not let you pick from older Test versions. So if you Test again after making a change, that latest Test becomes the version Run from Test will use.

You can now Refresh your data directly in the canvas. That makes it easier to pull in updates while you’re building, without having to leave the workflow first.
​

Refresh checks for changes in your data, including:

You can also see when the data was last refreshed.

Sometimes you do not want to test the whole workflow.

When you click Test in the preview window, Savant only runs the workflow up to that specific tool. So if you’re working on a Filter and want to check the results at that point, you can open the preview window and click Test.

Savant processes everything up to that tool without running the rest of the workflow downstream.

This makes it much easier to build and validate your workflow one step at a time.

Here’s the simplest way to think about it: Option Use it when you want to... Test Process the full dataset and review results Run Recalculate everything and write to the destination Run from Test Use your latest tested results and write them to the destination

For a lot of workflows, the new pattern will be: Build → Test → Review → Run from Test
​

Lightning Builder is meant to make workflow building feel faster and more straightforward.

You can now:

That means less waiting, less back-and-forth, and a much cleaner workflow-building experience.