top of page

Power BI vs Tableau vs Looker AI features: Copilot, Tableau Agent and Gemini requirements

2 minutes ago
7 min read
Power BI vs Tableau vs Looker AI features cover comparing Copilot, Tableau Agent and Gemini requirements. Data Studios original cover

Copilot in Power BI, Tableau Agent and Gemini in Looker answer the same request from a business user, a question typed in plain English against governed data, but each vendor gates its assistant behind a different purchase. Microsoft requires a paid Fabric capacity of F2 or higher, or Power BI Premium P1 or higher, and states that a Pro or Premium Per User license alone is not enough. Salesforce limits Tableau Agent in Tableau Cloud to the Tableau Cloud+ edition and the Tableau+ bundle, both quote-only. Google meters Conversational Analytics in Looker against a monthly token pool attached to the Looker (Google Cloud core) edition.


The comparison below is built only on the vendors' own documentation and pricing pages, all checked on November 18, 2026. It covers what each assistant needs in terms of edition, capacity, user role and admin settings, how usage is metered, and what a 50-person analytics team would actually have to buy on each platform before anyone can ask the first question.


The practical difference for a BI lead is where the AI cost sits. In Power BI it sits in the capacity, in Tableau it sits in the edition, and in Looker it sits in a shared token allowance that will be billed for overage once Google turns on enforcement.


··········


COPILOT, TABLEAU AGENT AND GEMINI IN LOOKER: REQUIREMENTS SIDE BY SIDE.


Each row comes from the product's official documentation or pricing page as listed on November 18, 2026.


........


Requirement

Copilot in Power BI

Tableau Agent

Gemini in Looker

Product tier

Paid Fabric F2+ or Premium P1+

Tableau Cloud+ or Tableau+

Looker (Google Cloud core) or Looker-hosted original 25.2+

Trial use

Trial capacities not supported

Tableau Cloud free trial supported

Not stated

User role

Read access for summaries, edit or build access to create

Creator or Explorer site role

access_data and gemini_in_looker permissions

Admin switch

Tenant setting for Azure OpenAI features

AI in Tableau site setting

Conversational Analytics and Looker Assistants settings

Metering

Capacity units on the Fabric capacity

Unmetered, no credits consumed

Monthly token pool per instance

Public price

Pro $14, PPU $24 per user per month

Cloud+ and Tableau+ quote-only

Editions quote-only


........


The rows show three different purchasing models. Power BI publishes per-user prices but requires capacity on top of them for Copilot. Tableau publishes per-user prices for the Standard and Enterprise editions, which do not include Tableau Agent, and keeps the AI editions behind a sales quote. Looker publishes no edition prices at all but does publish its token allocations and overage rates.


··········


COPILOT IN POWER BI: FABRIC CAPACITY, LICENSES AND TENANT SETTINGS.


Microsoft's Copilot introduction page, last updated on August 24, 2026, sets the capacity rule plainly: paid Fabric capacity F2 or higher, or Power BI Premium P1 or higher, with trial capacities and free SKUs excluded. The tenant setting "Users can use Copilot and other features powered by Azure OpenAI" must be on at the tenant level, and Microsoft notes that enabling it only at the capacity level is not sufficient. Tenants or capacities outside the United States and EU data boundary have Copilot disabled by default until an admin allows data to be processed outside the tenant's geographic region.


User licenses still apply on top of the capacity. The Fabric licensing page states that on F SKUs smaller than F64, each user viewing Power BI content must have Pro, Premium Per User or an individual trial, while on F64 and larger a Free user with a viewer role can view content. Microsoft lists Power BI Pro at $14.00 and Premium Per User at $24.00 per user per month, paid yearly. The F2 price is set per Azure region on the Fabric pricing page, and the page did not display a figure when it was checked for this comparison, so no capacity price is quoted here.


Copilot usage is billed from the capacity itself. Microsoft's consumption page lists 100 CU seconds per 1,000 input tokens, 10 CU seconds per 1,000 cached input tokens and 400 CU seconds per 1,000 output tokens, and classifies Copilot as a background job smoothed over 24 hours. The Copilot pane in reports is generally available; the standalone, cross-item Copilot is in preview and needs a separate tenant setting. A detailed breakdown of SKUs and regions is in the Copilot in Power BI requirements reference.


··········


TABLEAU AGENT: CLOUD+ OR TABLEAU+ EDITION AND SITE ROLES.


Tableau's Cloud pricing page lists Standard at $75 per Creator, $42 per Explorer and $15 per Viewer, and Enterprise at $115, $70 and $35, all per user per month, billed annually under an annual contract. Tableau Agent appears only under the Cloud+ edition and the Tableau+ bundle, both marked "Contact Sales for pricing." Tableau Pulse is included in every edition, while Tableau Knowledge, the business-context layer that feeds the agent, is limited to Cloud+ and Tableau+. The full price grid is covered in the Tableau pricing breakdown.


Within an eligible site, the help documentation requires the Creator or Explorer site role and the AI in Tableau site setting, which a site administrator turns on. The full setup for Cloud+ and Tableau+ connects the site to a Salesforce org with Einstein generative AI and a Data Cloud instance. Tableau states that AI in Tableau usage is unmetered and no longer consumes credits, and the FAQ mentions possible hourly request limits only for the free trial. In Tableau Cloud the agent runs on the Einstein Trust Layer, with a zero data retention policy toward third-party model providers and optional PII masking.


Tableau Server is a separate case. From version 2025.3, a Server site with AI in Tableau turned on can use Tableau Agent, but requests go directly to the customer's own LLM provider without the Einstein Trust Layer. Tableau also documents that the agent does not support cubes and cannot build dashboards, format views or model data.


··········


GEMINI IN LOOKER: CONVERSATIONAL ANALYTICS TOKEN POOLS AND PERMISSIONS.


Google's Gemini in Looker overview, updated on November 13, 2026, lists Conversational Analytics, the Visualization Assistant, Looker expression help, Explore summaries and semantic search without a status label, and marks Generate LookML, Insight Assistant and Admin Assistant as Preview. On Looker (Google Cloud core), a user with the Looker Admin IAM role enables the features in the Google Cloud console. On Looker (original), the instance must be Looker-hosted and on version 25.2 or later. Users querying a model need the access_data and gemini_in_looker permissions on it.


The Looker pricing page lists all three editions, Standard, Enterprise and Embed, as "Call sales." It does publish the Conversational Analytics allocations as a pool shared by all authenticated users of an instance: 60 million input and 1.2 million output data tokens per month for Standard, 300 million and 6 million for Enterprise, and 1.2 billion and 24 million for Embed. Non-production instances get 6 million input and 0.12 million output tokens. Overage is listed at $3.00 per million input tokens and $20.00 per million output tokens, and unused tokens do not roll over.


The billing date is where Google's pages currently differ. The Gemini overview states that usage above the allocation will be billed from February 1, 2027. The pricing page, checked the same day, says the promotional period without quota limits is being extended and that Google will give at least 90 days' notice before billing and quota enforcement begin. The earlier per-user allocation scheme reported in the Looker Conversational Analytics billing story has been replaced on the pricing page by these per-edition pools.


··········


WORKED SIZING EXAMPLE FOR A 50-USER BI TEAM.


The example below is illustrative. It assumes a team of 5 report authors, 15 analysts who explore data and 30 business viewers, and a typical assistant request of 3,000 input tokens and 600 output tokens. Token counts per request are an assumption for sizing, not a vendor figure.


........


Item

Power BI

Tableau

Looker

Users who can prompt

All 50 with access

20 (Creators and Explorers)

All 50 with permissions

Published per-user cost

50 x $14 Pro = $700/month

Standard $1,455 or Enterprise $2,675/month, without the agent

Not published

AI tier needed

F2+ capacity, regional price

Cloud+ or Tableau+, quote

Edition, quote

Usage metering

540 CU seconds per request

Unmetered

60M input, 1.2M output pool (Standard)


........


For Power BI, one request at the assumed size consumes 3 x 100 plus 0.6 x 400, or 540 CU seconds. An F2 capacity supplies 2 CUs for 86,400 seconds, or 172,800 CU seconds per day, so 100 Copilot requests a day take about 31.2% of a full day of F2 capacity and 320 requests would use all of it, before any refresh or report rendering. Because all 50 users view content on a capacity smaller than F64, each needs a Pro license, which comes to $700 a month at list price, or $1,200 if the team standardizes on Premium Per User.


For Tableau, only the 20 Creators and Explorers can use Tableau Agent, and the published Standard and Enterprise totals of $1,455 and $2,675 a month describe editions that do not include it. The real AI cost is the difference quoted by Salesforce for Cloud+ or Tableau+.


For Looker on the Standard edition, the 60 million input token pool works out to 1.2 million input tokens per user per month across 50 users. If the team used 75 million input and 1.5 million output tokens in a month after enforcement begins, the overage would be 15 million input tokens at $3.00 and 0.3 million output tokens at $20.00, or $51.


··········


COMMON MISREADINGS OF BI ASSISTANT LICENSING.


A Premium Per User license does not give access to Copilot in Power BI. Microsoft states that Pro or PPU alone is not sufficient and that the report must sit in a workspace on paid Fabric or Premium capacity. Buying or scaling a capacity can take up to 24 hours before Copilot recognizes it.


Tableau Viewers cannot use Tableau Agent for web authoring, since the documentation ties it to Creator or Explorer site roles, and buying Enterprise does not include the agent. On the Looker side, the token pool belongs to the instance, so heavy use by a few analysts draws down the allowance for everyone on that instance.


··········


POWER BI, TABLEAU AND LOOKER AI: WHAT TO CONFIRM BEFORE YOU BUY.


The vendor documentation confirms the structural gates: Fabric F2 or Premium P1 capacity plus per-user licenses for Copilot, the Cloud+ or Tableau+ edition and Creator or Explorer roles for Tableau Agent, and an edition-level token pool with published overage rates for Gemini in Looker.


What depends on the customer is capacity region and price for Fabric, the quoted uplift for Tableau Cloud+ or Tableau+, the quoted Looker edition, and tenant or site settings that an administrator may have switched off. Data processing also differs: Microsoft disables Copilot by default outside the US and EU data boundary, Tableau Cloud routes requests through the Einstein Trust Layer, and Google processes Conversational Analytics in transit through a global service while keeping data at rest in the instance's region.


Still open are the F2 price for your Azure region, which the Fabric pricing page did not display when checked, every quoted edition price, and the confirmed start of Looker overage billing, where Google's two pages give different answers. A team with an existing Microsoft 365 and Power BI estate can size Copilot from published CU rates; Tableau and Looker buyers need written quotes and, for Looker, written confirmation of when token enforcement starts.


··········


FOLLOW US FOR MORE.


·····


DATA STUDIOS


·····


datastudios.org


bottom of page