Tungsten told you where the platform is going. Nobody told you how to run the two or three years it takes to get there.— The Reveille Perspective
If you run Tungsten Capture and Transformation, 2026 has been a loud year — and almost none of the noise has been about you. Tungsten Automation shipped TotalAgility 2026.1 in March and 2026.2 in May, adding generative classification, email-based intake, trainable document separation, and Model Context Protocol support for third-party AI models. The roadmap is unambiguous, and to be fair to Tungsten, it’s a good one.
Meanwhile, in a data center you actually own, a Capture batch class built in 2014 is still splitting a lockbox scan into eleven document types at 4:40 in the morning. It hasn’t read the roadmap. It will be running the October quarter-close, and the one after that.
The distance between the platform the vendor is building and the platform the business is running isn’t a gap you close on a weekend. It’s a place you live in, for years. Call it what it is: the Unmigrated Middle.
Core Tension
Every migration story is told as a before and an after. The install base lives in the during — where the legacy capture stack and TotalAgility both run, both matter, and neither one is measured against the other.
Quick answers
Is Tungsten Capture being discontinued in 2026?
What does the Capture and Transformation converter not convert to TotalAgility?
What is Tungsten’s Move Up to TotalAgility program?
How do you observe Tungsten Capture and TotalAgility at the same time?
01 — The Setup
What actually changed in 2026
Two real shifts — neither of which turned your servers off.
The destination got specific. For a few years after the Kofax rebrand, “move to TotalAgility” was a direction. Now it’s a product plan with dates on it. TotalAgility 2026.1 added large-language-model classification for low-confidence and variable-format documents, email-based intake, and Model Context Protocol support so external AI models wire in without custom code. 2026.2 followed in May with deeper extraction and AI governance work. This is a platform being invested in, not parked.
The path got formal. “Move Up to TotalAgility” is a named program with its own Tungsten curriculum aimed squarely at “experienced users of Tungsten Capture and Tungsten Transformation.” The mechanics are a converter utility that takes Capture batch classes and Transformation projects across, plus a Capture Starter Pack for rebuilding on the other side. Tungsten has stopped implying and started instructing.
None of which is a shutdown notice. Tungsten Automation has not published an end-of-support date for the current Capture and Transformation releases, and the Tungsten Capture 11.x documentation set is still on Tungsten’s own product portal. Anyone selling you a hard cliff date is selling you something the vendor hasn’t said.
02 — The Constant
What didn’t change at all
The plumbing under your capture operation is exactly where you left it.
Nothing about the vendor’s 2026 changed a single component of the environment you’re accountable for on Monday morning.
Your license server still answers license requests or it doesn’t. Your Remote Sites still check in on their synchronization interval, still download batch classes on a polling schedule, and still quietly stop doing both when a certificate rolls. TCN Server still holds a queue of in-progress batches that can climb all afternoon unnoticed. Unattended modules still process too slowly; batches still age past the point where the 5:00 PM commitment is recoverable. In Transformation, the Real-Time Transformation Interface still tracks request duration and worker processes available, and field accuracy — the rejection and correct-read rates that tell you extraction quality is drifting — is still a Transformation Modules measurement. On the scanner side, VirtualReScan still reports roller health, clarity and sheets scanned between feeder errors, still the earliest warning that image quality is about to poison everything downstream.
And the business rules — validation logic, foldering, document review methods — still sit inside the Transformation project where a decade of institutional knowledge quietly accumulated. The vendor’s roadmap moved. Your operational surface area did not.
03 — The Gap
Why the middle lasts years, in the vendor’s own words
The converter moves the software. It doesn’t move the decade of business logic on top of it.
The most useful document in this entire conversation isn’t a competitive analysis. It’s Tungsten’s own migration guide — specific, honest, and almost entirely unread by the install base. On what happens when a Transformation project is imported:
- “All the validation forms in the Kofax Transformation Modules project are deleted when the project is imported.” You “create replacement forms in the TotalAgility Form Designer.”
- “All the validation rules … are deleted when the project is imported.” You “create replacement rules in TotalAgility.”
- “Kofax Transformation Modules foldering, folder nodes and scripts are deleted from the project file during import.” You “must re-configure them in TotalAgility.”
- Document review methods and rules “are deleted from the project and related settings are not accessible in the Transformation Designer.” You “must manually replicate them.”
- “All scripts … remain, but not all of them are executed in TotalAgility.”
Source: Kofax TotalAgility Migration From Kofax Products, v7.7.0. Verify against the current edition in your Tungsten support portal.
Read that as an operator rather than an architect. The converter isn’t failing — it’s doing exactly what it says. It carries the structure across and leaves the judgment behind. Fields and locators travel; the accumulated logic of how your organization decides an invoice is wrong does not.
Which is why the honest project plan isn’t six months. It’s a rebuild of business rules, running in parallel against production, validated against what the old system was already producing, for as long as that takes. The Unmigrated Middle isn’t a symptom of a badly run migration. It’s what a competently run one looks like.
04 — The Ledger
Changes, constants, and what each means for you
The whole argument on one page.
| Dimension | Changed in 2026 | Didn’t change | What it means for you |
|---|---|---|---|
| Product direction | TotalAgility 2026.1 and 2026.2 shipped, positioned to replace legacy capture | No published end-of-support date for current Capture and Transformation releases | Plan the move on your evidence, not a deadline nobody issued |
| Migration path | Move Up formalized, with converter utility and Capture Starter Pack | Validation forms, rules, foldering and review still rebuilt by hand | Scope the project around business-rule rebuild, not software conversion |
| Your components | Nothing | License server, TCN Server, Remote Sites, module queues, RTTI, VRS, Import Connector | Every legacy failure mode is still live and still yours to catch |
| Extraction quality | Generative classification and trainable separation in TotalAgility | Field accuracy measurement remains a Transformation Modules capability | Don’t assume accuracy telemetry follows you across the seam — verify it |
| Service level | Two platforms now carry production work | The business still expects one number for the whole process | You need one measurement spanning both, or you can’t answer for either |
05 — The Practice
Four things that break in the middle
Quiet failures, attributed to the wrong platform, found by the business first.
The Remote Site That Stopped Checking In
A branch station keeps accepting work but stops synchronizing, so it never downloads the batch class revision the central team published. Documents keep being captured — against last quarter’s index definition.
The Split Ledger
Half the invoice volume runs on the legacy stack, half on TotalAgility. Each has its own console and its own definition of “processed.” Finance asks one question — did everything clear? — and nothing can answer it.
The Aging Batch Nobody Owns
During parallel run, an unattended module slows under contention from the new environment. The platform is up and the queue is moving, so no threshold fires — and the exception surfaces at close.
The Rebuilt Rule That Isn’t The Same Rule
A validation rule is faithfully recreated in TotalAgility — but the edge case it was tuned for in 2019 was never written down. Extraction still “works.” The rejection profile shifts a few points, invisibly.
Same root cause every time, and it isn’t a Tungsten problem: during the middle, the two halves of your document pipeline are measured by two tools that were never designed to reconcile — while the business process spans both.
Forward Principle
You cannot prove a migration succeeded if you never measured what the old system was actually delivering. A baseline you didn’t capture is an argument you can’t win.
06 — The Point
One layer over both platforms
The difference isn’t the migration plan. It’s what you could see while it ran.
Content Observability is the continuous visibility, assurance, and optimization of the content and document workflows that drive business outcomes — and feed AI. Mid-migration, that means one agentless layer watching the legacy stack and TotalAgility with the same instrumentation, so the service level is a single number instead of an argument between two consoles. It matters more here than in steady state for a structural reason: vendor-native tools are scoped to the vendor’s own boundary, and during a migration that boundary runs straight through the middle of your process.
Reveille for Tungsten Automation
Agentless, deployment-agnostic Content Observability across the whole Tungsten estate — over 100 pre-built tests and over 180 dashboard metrics spanning Capture, Transformation Modules, TotalAgility and the supporting servers, with proactive actions that clear bottlenecks before they become tickets. See Reveille for Tungsten Capture →
In the version that ends well, the team treats the middle as a first-class operating state. They baseline what the legacy environment actually delivers before anything is rebuilt, then run one layer across both so the parallel run is a comparison rather than a hope. The cutover is boring, because the evidence arrived months earlier. When someone asks whether the migration worked, the answer is a chart. The same layer follows onto TotalAgility and proves the rebuilt process performs at least as well as the one it replaced.
In the other version, the middle is a temporary inconvenience. Nobody baselines the old system because it’s “going away” — and monitoring on it is quietly deprioritized for the same reason, while it runs production, unwatched, for another two years. The rebuild drifts a few points from the original and nobody can prove it, because the original was never measured.
The difference isn’t the roadmap, the converter, or the vendor. It’s whether the organization treated the years in between as an operating state worth measuring — or as a waiting room.
Your capture platform still works. The only real question is whether you can prove it — before, during, and after the thing that replaces it.
See both platforms in one view — before the middle starts.
One agentless layer across Tungsten Capture, Transformation and TotalAgility.