Tungsten Automation connects document AI to business processes

Tungsten Automation helps businesses turn documents into usable information and then into operational work. Its TotalAgility platform joins document processing, process automation and knowledge discovery. That makes it relevant when extracting a number is only the beginning: the business must validate it, reconcile it with a system of record, route an exception and retain the evidence. The right evaluation follows that entire chain instead of stopping at a persuasive extraction demo.

01 / ProductThe former Kofax portfolio now leads with document intelligence

Tungsten's rebrand FAQ confirms that Kofax became Tungsten Automation and that offerings including ReadSoft, Ephesoft and Tungsten Network remain in its portfolio. The company-level identity matters because legacy documentation and product installations may still use Kofax names. Those are not separate new companies to evaluate or independent competitors within this blueprint.

The current TotalAgility overview describes document automation, process orchestration and knowledge discovery as connected capabilities. DocAI Studio handles incoming document information; process automation coordinates work across people and systems; knowledge discovery helps users ask questions of managed content. The relationship is sequential as well as complementary: poorly understood documents can produce unreliable downstream actions.

Document Automation sets out ingestion, classification, extraction, validation and delivery. Its validation step includes rules, matching, confidence and exceptions. That is the important product distinction for a buyer: the output should be a usable business record with an explainable review path, not merely a collection of strings that looks plausible on a screen.

02 / AudienceA fit for operations where documents determine what happens next

TotalAgility is most relevant to document-intensive operations such as shared services, onboarding and claims administration. The common problem is variable evidence: documents arrive in different formats, contain incomplete information and require different handling depending on the case. Staff must interpret the material and then coordinate with another system or team.

It is less naturally a first choice for a small application that only needs one extraction API and already has a well-managed workflow engine. A larger automation platform introduces configuration, licensing and operational ownership. The evaluation should establish whether bringing capture and process handling together removes meaningful friction or merely moves existing integration work into another environment.

ABBYY is a useful comparison for document intelligence, while UiPath provides a broader automation comparison. Put the same representative packets, exception types and downstream tasks in front of both approaches. Otherwise, a capture-only benchmark may favor one system while leaving the most expensive operational step outside the assessment.

03 / WorkflowProposed workflow: process a supplier invoice with visible exceptions

For a proposed pilot, choose a supplier-invoice process with a clear handoff to an existing finance system. Build a reference collection containing purchase-order invoices, credit notes, duplicates and invoices that mix several orders. Identify the fields that matter to the decision and retain the original file and its source. This prevents a visually clean extraction from being mistaken for a valid payable transaction.

Use DocAI Studio to classify each item and extract the relevant values. Then apply a separate validation layer: match the supplier, reconcile currency and totals, and compare the purchase-order reference with the system of record. The pilot should distinguish unreadable information, missing information and business-rule conflicts. They require different corrective actions and should not all be represented by an undifferentiated low-confidence flag.

The Process Automation page describes reconciliation, checks, approval routing and archival as connected stages. In the proposed design, a conflicting amount should reach a reviewer with the supporting document and matched record. A clean extraction may still need approval under the organization's spending rules; the model should not decide that authority simply because the document was easy to read.

Knowledge discovery can be investigated as a later step, such as answering which agreement governs a recurring invoice dispute. Its product page describes an AI Knowledge Base and answers linked to sources. Require a reviewer to open the agreement and verify the cited clause. Retrieval can assist the decision without replacing the business owner responsible for interpreting it.

Measure completed, usable records and total exception-handling effort. Keep separate counts for extraction corrections, failed matches, approval delays and downstream posting errors. This reveals whether the integrated process improves work after the documents have been read. No throughput, accuracy or time-saving result is asserted here from testing.

04 / PricingLicence tiers make workflow scope a commercial decision

The TotalAgility 2026.1 licensing guide distinguishes Standard, Advanced and Enterprise. Standard provides DocAI document-processing capability; Advanced adds the wider TotalAgility product suite for orchestration and automation; Enterprise adds knowledge-base capability. A proposal that includes source-grounded knowledge discovery should not be costed as a Standard-only extraction deployment.

The current product pages offer demonstrations and contact routes rather than a single universal subscription price. Request a quote tied to the intended document volume, environments, users, deployment and process features. Clarify which components are covered by the selected tier and whether existing licences can be reused. Public access to a feature guide does not establish that a customer's licence permits the feature.

Generative AI capacity also needs explicit treatment. The licensing release note documents token-based LLM licensing and a Tungsten AI provider, including regional capacity and fallback configuration. Because that note is release-specific, confirm the applicable current contract and installed release instead of transferring its terms indiscriminately to every deployment.

05 / DistinctionsThe distinction is the connection from extraction to governed action

Tungsten's combination is useful when documents and process exceptions have to be managed together. The document pipeline produces information, the workflow decides who handles it and the knowledge layer helps interpret related evidence. A team can therefore evaluate the full operational path inside one platform family, while still choosing explicit boundaries for each stage.

The company's longstanding capture portfolio also makes version and naming discipline important. An existing Kofax installation is not automatically equivalent to a current TotalAgility deployment. Ask which capabilities are native to the selected release, which need an upgrade and which depend on another service. This keeps an attractive current demonstration from becoming an inaccurate description of the installed estate.

Source-linked answers are another useful distinction when staff repeatedly read related policies or agreements. Their value depends on selecting the correct corpus and maintaining it. The proposed knowledge base should distinguish approved, expired and draft material, and the evaluation should include questions for which the documents provide no answer. A citation is a navigation aid, not proof that the interpretation is correct.

06 / QuestionsResolve document quality, licensing and recovery behavior

A representative evaluation must include the hard documents, not just common templates. Test poor scans, handwriting where relevant, tables spanning pages and multiple documents joined into one file. Record whether failures are detected and how quickly a reviewer can recover. The product pages describe broad capabilities, but they do not establish accuracy for a particular archive or supplier population.

Check recovery at the system boundary as well. If a downstream connection fails after a transaction is created, can the process retry without duplicating the invoice? If an approver changes a value, is the correction retained with a clear history? These are proposed acceptance tests for the implementation, not claims about behavior observed during this research.

Finally, validate the plan and release behind every important feature. Knowledge-base access, model capacity, on-premises responsibilities and cloud service boundaries can alter both cost and architecture. A quote, deployment diagram and supported configuration should agree with the demonstrated workflow before the team commits to a broader migration.

07 / DecisionEvaluate a complete document-driven outcome

Tungsten Automation is worth investigating when document understanding and subsequent operational work need a common process. Start with one measurable journey, keep human authority explicit and test exceptions through to completion. The purchase decision becomes clearer when the team knows which tier and services are required to deliver a usable outcome, rather than merely to extract data.