Waii implementation cost: Data 360 credits, Tableau Next and the work

Waii built a platform that turns plain-language questions into SQL. Salesforce describes it this way: "Waii’s technology translates plain-language questions into complex, production-ready SQL queries, helping users access and act on data more naturally and efficiently." The deal closed quickly, and the record is simple: Salesforce completed its acquisition of Waii on August 15, 2025. The announcement said Salesforce planned to integrate Waii into Data Cloud, and that its metadata knowledge graph would power agentic workflows and insights across the platform, including Agentforce and Tableau Next.

So the honest answer to "what does a Waii implementation cost?" is that you no longer buy Waii. You buy the Salesforce products the technology was announced for, and you pay for the work that lets a plain question produce a correct number. Salesforce has not published a page saying which shipped feature contains Waii code, so this guide does not claim one. It prices what you would actually buy today to get natural-language access to governed data inside Salesforce: Data 360 and Tableau Next.

For the history, see what happened to Waii. For how the technology fits Data 360 in practice, read Waii and Data 360. This page is about the bill.

#A note on names

Data Cloud is now Data 360. Salesforce's help documentation puts it plainly: "As of October 14, 2025, Data Cloud has been rebranded to Data 360." The Waii announcement predates the rename, so it says Data Cloud. Everything below uses the current name, and the pricing pages use it too.

#What you pay Salesforce, as at October 2026

Two pricing pages matter. The Data 360 pricing page sets out how data work is charged, and the Analytics pricing page lists Tableau Next. All figures are in US dollars and come from those pages as at October 2026.

A few details on those pages change how you should budget.

Ingestion into Data 360 is free. The pricing page says bringing data in, through batch ingestion and zero-copy integrations, costs nothing; you pay when you use the data. For a natural-language query tool that is the important line, because every question that runs against Data 360 data is a query, and queries are charged against credits. The cost of answering questions scales with how many questions people ask and how heavy each one is.

You can start for free. The page states: "Salesforce customers can provision Data 360 for free with a limited amount of storage and consumption credits." That is enough for a proof of concept on a sample of data, but not for a production rollout.

Tableau Next may already be in your contract. The Analytics pricing page says: "The full power of the Tableau Next platform is now included in Core, Advanced, and Max editions for Sales, Service, and Industry Clouds." If your Sales or Service users are on one of those editions, the question is not whether to buy Tableau Next, but whether the users who need it are on the right edition.

Flex Credits move between products. Salesforce says credits can be transferred between Data 360 and Agentforce. If an agent will also answer data questions, plan one credit pool rather than two budgets.

#What a natural-language analytics implementation is made of

Turning questions into reliable SQL depends far more on the data underneath than on the model. Waii's own pitch was that a knowledge graph of tables, relationships, metrics and governance rules is what makes the SQL accurate. Whatever engine you use inside Salesforce, the work packages look like this.

Discovery and question inventory. Collect the questions people actually ask: revenue by region last quarter, open cases older than a week, renewals at risk. Note who asks them and who is allowed to see the answer. This list becomes the specification and, later, the test suite.

Source mapping into Data 360. Decide which systems feed Data 360, through which connectors, and whether you ingest or use zero-copy access to a warehouse you already run. Ingestion is free on the price list, but someone still has to configure streams, map fields to data model objects, and agree refresh timing.

Semantic modelling. This is the heart of the job. Define metrics once: what counts as revenue, which date a booking uses, how a customer is identified across systems. Tableau Semantics is listed as part of Tableau Next, and the Waii announcement said its technology would form the basis of Tableau's next-generation semantic engine. A thin model gives fluent answers that are wrong. A well-built one makes the plain-language layer almost boring, which is the goal.

Identity resolution and harmonisation. If questions span customers who exist in several systems, the unification rules decide whether "how many active customers do we have" returns one number or three. On the Profiles models, profile-building actions are included in the flat price; on Flex Credits they draw down credits.

Access and governance. Decide who can ask about what. Natural-language access does not change the rules on who may see salaries or margin; it makes it easier to ask. Permissions must be set before the question box goes live.

Testing. Run the question inventory, compare answers to known figures from finance or operations, and fix the semantic model where they differ. Repeat after every model change. Keep the bank and grow it.

Adoption and monitoring. Train people to ask well, show them how to check an answer, and watch credit consumption in the first months.

#What drives cost up or down

The biggest variable is the state of your metric definitions. An organisation where finance, sales and operations already agree what "bookings" means will finish semantic modelling in a fraction of the time of one that has four spreadsheets with four answers. That is a business decision, and it often needs a steering meeting more than a developer.

The second is the number of sources and how they join. Two well-keyed systems are quick. Six systems with different customer identifiers need identity resolution, and that adds design time and credit consumption.

The third is the pricing model you pick in Data 360. Profiles give a flat, predictable cost for profile-heavy uses such as marketing segments. Flex Credits suit query-heavy, variable use. A natural-language analytics rollout is usually query-heavy, and querying is charged as you use it under every model, so you will hold Flex Credits whichever model you choose. Salesforce's FAQ notes that profile pricing still uses Flex Credits for non-profile work such as querying data.

The fourth is how many people get a seat. Tableau Next is priced per user. If the aim is that every manager can ask questions, the seat count may matter more than the implementation fee.

#How to get comparable quotes

Give every partner the same pack: the question inventory with fifty or so real questions, the list of source systems and their record volumes, the editions you already hold, and the people who will own metric definitions. Ask each partner to quote against it, not against a generic "Data 360 and Tableau Next deployment".

Then ask the same questions of each. How many questions from our list are in the acceptance test? Which metrics will you model in phase one? Are connectors and data streams in scope, or do you assume data is already in Data 360? Who resolves disagreements on metric definitions, and what happens to the timeline if that takes a month? What credit consumption do you expect in the first quarter, and how did you estimate it? Will you set up monitoring in Digital Wallet?

A partner who cannot estimate credit consumption has not thought about the running cost. The pricing page links a Data 360 pricing calculator; ask the partner to use it with your volumes and show the inputs.

#Costs outside the services quote

Credits after go-live. This is the one most often forgotten. Every question costs something. Salesforce offers Digital Wallet, a free tool for monitoring consumption-based products, and it should be set up before launch, not after the first surprise.

Edition upgrades. If Tableau Next comes through Core, Advanced or Max, moving users onto those editions has its own price, set out on the relevant cloud's pricing page.

Data quality work in source systems. Duplicate accounts and inconsistent picklists in the CRM are fixed in the CRM, not in Data 360. That work is usually done by your administrators.

Sandbox and testing environments. Confirm with your account executive what testing environments come with Data 360 and how consumption there is charged.

A metric owner. Someone in the business has to own definitions after go-live. If nobody does, the semantic model drifts and the answers drift with it.

#How to shape the first-year budget

A useful first-year budget for natural-language analytics has five lines, and only one of them is a services fee.

The first is seats. Count the people who will ask questions, check which of them already sit on an edition that includes Tableau Next, and price the rest at the published per-user rate or the upgrade to a bundled edition. Do this per team, because the cheapest answer for sales may not be the cheapest answer for finance.

The second is credits. Start from the question inventory: how many people, how many questions each per week, and roughly how much data each question touches. Feed that into the calculator and add headroom, because usage in the first month after launch is rarely typical. Novelty drives questions up, then it settles.

The third is the build: discovery, source mapping, semantic modelling, testing and training. This is the line partners quote, and it should be broken down by work package so you can see where the effort sits.

The fourth is internal time. Metric owners, source system administrators and a product owner will each spend real days on the project. Put their time in the plan even if it never appears on an invoice.

The fifth is run cost: monthly credit review, model maintenance as the business adds products or regions, and periodic re-testing of the question bank.

#Phasing that keeps spend predictable

The safest rollout starts with one domain where the data is already clean. Sales pipeline is a common choice, because the objects are well understood and the people asking are used to dashboards. Model a dozen metrics, test them hard, open the question box to one team, and watch both accuracy and credit use for a month.

Only then add a second domain. Each new domain brings new sources, new definitions and new arguments about definitions, so it is better treated as its own small project with its own acceptance test. That rhythm also gives finance real consumption figures from the first domain to forecast the next, rather than a guess made before anyone had asked a single question.

#Where to go next

If you have a Waii contract from before the acquisition, the migration guide covers timing. For the licence side alone, see the Waii licence cost page. For integration design, read integrating Waii with Salesforce, and if you are choosing a delivery firm, hiring a Waii partner lists what to test them on.

The short version for a budget holder: you are buying credits and seats, and paying for a semantic model. The credits are published, the seats are published, and the modelling can be quoted precisely once you hand partners your real questions.