Workiva has put AI throughout its platform: a chat assistant that follows the user across documents, spreadsheets, presentations, Wdata and files, domain extensions for finance, GRC and sustainability work, an agent builder called Agent Studio, and a connector that lets an outside AI client read Workiva content. The licensing question for an asset manager is not a seat count for these features. It is the contractual gate that switches them on, the controls that limit who can use them, and the early access programs whose terms can change.
Editions
Workiva describes its AI offer in layers.
- Workiva AI, the integrated assistant. Its chat is available from anywhere in the platform and understands the context of Documents, Spreadsheets, Presentations, Wdata and Files.[1]
- Intelligent extensions, which the FAQ names as Intelligent Finance (disclosure creation, peer research and alignment with accounting standards), Risk and Controls Intelligence for GRC, and Intelligent Sustainability (materiality assessments, benchmarking and disclosure drafting aligned to standards such as ISSB and ESRS).[3] The SEC page adds that Intelligent Finance unifies AI at the core of the platform.[10]
- Agent Studio, where customers create agents or customise pre-built templates, described as no code, with rules for human review, approval and oversight.[6]
- Workiva MCP, a connector through which an AI client reads Workiva documents and data, offered in an early access program.[8]
The pricing page lists “Integrated Workiva AI Intelligent Companion powered by leading LLMs and pre-built prompt library” among the items included with every plan.[9] At the same time the Support Center requires a signed Workiva AI Terms of Use before Workiva AI is available.[1] The two statements are compatible if the AI terms are a legal acceptance rather than a separate fee, but Workiva’s public pages do not say whether any AI feature, including Agent Studio and the intelligent extensions, is charged separately. Catalog rows: Workiva AI module; Workiva AI companion included in every plan; Agent Studio runs agents under human oversight.
Metrics
Workiva states no AI metric such as credits, prompts or tokens in the sources used here. The only counted construct is the person: users with the Workiva AI User organization role can use Workiva AI in the workspaces where it is enabled.[5] By default that role is assigned to all users in workspaces where AI is enabled, and Org User Admins can switch to an assignment-only model.[2] Catalog rows: Workiva AI User organization role; Workspace member (role-based user).
Counting and floors
The Terms of Use gate
Workiva AI “is available to any customer who has signed the Workiva AI Terms of Use”. After approval it is enabled by default for the whole organization, including every existing and newly added workspace and user.[1] The FAQ states that these terms “cannot be customized and are non-negotiable”, and it sends data-security questions to the organization’s Master Services Agreement.[3] A customer’s inventory should therefore record the date on which the AI terms were accepted, because from that date AI is on for every user unless an Org Admin restricts it. Catalog rows: Workiva AI Terms of Use enablement; Workiva AI requires signed AI Terms of Use; Workiva AI Terms of Use are non-negotiable.
Controls available to Org Admins
An Org Admin can switch the Workiva AI module off for the organization, enable it for chosen workspaces only, limit access to users holding the Workiva AI User role, and enable or disable individual model providers.[2] Supported models are Google Gemini, Microsoft and OpenAI, and AWS and Anthropic, with provider options that may vary by region, and Workiva routes to the best model among those enabled.[2] The Org User Admin who changes user access must themselves hold the Workiva AI User role.[2] Catalog row: Workiva AI model providers.
Data and permission rules
Workiva states that Workiva AI is fully encrypted and never trained on Workiva content or customer data.[2] Its approach article describes a contractual commitment that customer data is not used to train large language models, with encryption in transit using TLS 1.2 or later.[4] The FAQ adds that Workiva may store limited usage and operational data, such as identifiers, prompts, document context and performance metrics, to operate, secure and improve the service, and that this data is not used to train Workiva AI or third-party models.[3] The platform page states that actions, whether built into the platform or connected through Workiva MCP, run under the customer’s existing permissions and access controls.[7] Catalog rows: Workiva AI is not trained on customer data; Workiva AI usage data is not used for training; Platform actions run under existing permissions.
The FAQ also states that Workiva AI is not always accurate and should not be relied on as a source of up-to-date laws, regulations or accounting guidance.[3] That is a use caution, not a licence term, but it is relevant to a regulated filer deciding how far to extend AI use.
Virtualization and partitioning
AI enablement follows the structure of the platform. The organization-level toggle covers all workspaces, and individual workspaces can be enabled or disabled beneath it.[2] Workiva AI works with the content of a workspace, and Workiva describes the Workiva MCP as acting only within the signed-in user’s existing permissions.[8]
Cloud and BYOL
Workiva AI is delivered by Workiva, using models from the providers listed above, so no customer licence for those model providers is needed to use the integrated assistant. The Workiva MCP is different in one respect: the customer’s AI client is provided by the customer. Workiva’s guide uses Claude as the supported client at the early access launch, and the client’s own subscription terms are outside Workiva’s documentation.[8]
Programs
Workiva MCP Early Access Program
The Workiva MCP connector lets an AI client read Workiva documents and data and respond with the sources used. At launch it is read-only: it cannot create, edit, update or delete artifacts.[8] The connector is a custom connector that the customer’s Claude administrator must add, and a Workiva security admin must then grant the agent access to specific workspaces through the Connected AI agents console.[8] The feature is available through the Early Access Program until 2026-11-01, is provided for technical testing and evaluation, and may be extended by Workiva at its discretion; interested teams contact their Customer Success Manager.[8] Catalog rows: Workiva MCP Early Access Program; Workiva MCP early access until 2026-11-01; Workiva MCP is read-only at launch.
Other early access features
The custom roles feature follows a similar pattern: it is part of an early adopter program that only Workiva can activate, and it is provided for technical testing and evaluation until 2026-10-28.[11] Catalog row: Custom roles early adopter program. For both programs, an inventory should record the end date and who at the customer has enabled the feature, because the terms after the testing period are not stated.
Out of scope
This article does not evaluate model quality, the legal basis of any AI feature, or the contents of the Workiva AI Terms of Use, which are not published on the pages cited. It does not cover third-party AI client licences, such as those of the client that connects through the MCP, or any consumption charges that may be agreed in an order form.