Spindle AI licence cost: where scenario modelling is billed now

Spindle AI is an awkward product to license because it was bought for a capability, not a price list. Salesforce's own words about the plan give the clue: the technology was to help deliver "custom agentic analytics, ROI forecasting" and continuous improvement for every Agentforce user. That sentence describes a feature of the Agentforce estate. It does not describe a product with seats and a unit price, and that shapes everything below.

#What was bought and what you can buy

The completion notice is a single line.

"Salesforce completed its acquisition of Spindle AI on November 21, 2025."

The announcement that went with it called Spindle AI an agentic analytics platform that combines AI agents and machine learning with data modelling, and gave examples of what business teams use it to simulate: pricing and packaging, go-to-market optimisation, and sales mix shifts. It positioned the technology as complementing analytics platforms such as Tableau, by focusing on using AI to autonomously model scenarios and forecast business outcomes.

we looked for the standalone product on 13 October 2026. The spindle.ai address forwarded to Salesforce's Agentforce page, and none of the Salesforce price pages we read for Sales, Analytics, Data 360, Commerce or Agentforce carried a Spindle line. A buyer who wants scenario modelling today is therefore buying around the capability, through three existing lines, and has to work out which of them applies to the decision at hand. The rest of this guide takes those lines in turn.

#Line one: Tableau Next

Analytics is where a modelled scenario ends up in front of a decision maker, and the Salesforce analytics price page lists the options with billing terms. On 13 October 2026, Tableau Cloud started at US$15 per user per month, Tableau Next at US$40, and CRM Analytics at US$140, each billed annually.

The more important sentence is about bundling: "The full power of the Tableau Next platform is now included in Core, Advanced, and Max editions for Sales, Service, and Industry Clouds." If you already hold a Core, Advanced or Max edition, some Tableau Next capability is in the price you pay, and a separate Tableau Next purchase would be duplication. That makes your first licensing task a check of your own contract, not a visit to a price list. Find out which edition you are on, what the bundled Tableau Next entitlement covers, and which of your intended modellers and viewers fall inside it.

The page also notes that Premier Success Plans are priced as a share of net licence fees, so any uplift in licence value feeds through to the cost of support. It is a small point that is easy to leave out of a business case.

#Line two: Data 360 credits or profiles

A scenario model is only as good as the data it reads. Data 360, which Salesforce formerly called Data Cloud, is where much of that data is unified, and its pricing page is explicit about the two ways to pay. It puts the choice this way: "Choose Credit-based pricing for a flexible, dynamic model that scales alongside your specific consumption, or Profile-based pricing for a predictable, flat-rate model designed for control."

The published rates on the date read were these.

Ingesting data in batch is listed as free; you pay when you act on it, for example in preparation, unification, segmentation, queries and streaming. For a modelling workload, that means the costly steps are the ones that run on every refresh of a scenario, not the one-off load.

#Line three: Agentforce credits, shared with Data 360

The Data 360 page lists a feature that matters here: you can move Flex Credits between Data 360 and Agentforce. That turns the two lines into one pool, which is useful if the spend is uneven, and a trap if two teams draw on the same pool without a budget owner. On the Agentforce pricing page, the unit is introduced like this: "Flex Credits are a new unit of payment for Agentforce that align cost to the business value your AI agents create." A standard action draws 20 credits. If a modelling agent acts on records as well as reading them, that spend lands here.

#A worked estimate method

Because the capability has no list price, the cost has to be estimated from a decision, not from a licence count. The method below uses only published rates and quantities you supply.

1. Choose one decision, for example a repricing review with a defined set of products. Do not start with a platform.
2. Count the records the model needs to read: customers, orders, price books, whatever the decision depends on. That is your data volume.
3. Decide how often the scenario is refreshed. A quarterly review and a daily one differ by about ninety times.
4. For Data 360, work out whether you need profile-based pricing, which suits unification at a stable customer count, or credits, which suit variable use. The pricing page offers a calculator for estimating credits.
5. For Tableau Next, count the people who will build and the people who will only view, then compare the per-seat price with what your edition already includes.
6. For Agentforce, estimate the actions any agent takes as a result, and multiply by 20 credits and US$0.005 a credit.

As a sense check on the last step, 10,000 agent actions a month is 200,000 credits, or US$1,000. That is arithmetic on the published rate, with a volume we chose for illustration. Your own figure will come from a pilot.

#Contract traps

Three traps recur with consumption-priced analytics.

First, credits that expire. The Agentforce pricing page says unused Flex Credits do not roll over into subsequent subscription terms. A year-one pool sized for a launch that slips is a loss.

Second, a refresh cadence that quietly multiplies cost. A model that someone schedules daily because the dashboard looked better that way can cost far more than the quarterly cadence the decision needed.

Third, an ROI forecast that is treated as a licence benefit. A forecast is an output. It does not justify a purchase until the baseline data underneath it has been checked against your own records, and the earlier Spindle AI and Agentforce guide explains why a number without a traceable source is a poster, not evidence.

#Profiles or credits: a break-even sketch

Data 360 offers the clearest example of how the choice of model changes the bill, so it is worth working through with illustrative numbers. Suppose you unify 100,000 customer profiles. These are my figures for the sake of arithmetic, not a forecast.

On the Profiles model at US$240 per 1,000 profiles a year, 100,000 profiles cost US$24,000 a year, and the page says profile-building actions are included, with one Flex Credit per profile to scale. On Enterprise Profiles at US$420 per 1,000, the same profiles cost US$42,000 a year, with two credits per profile and more actions included. On the credit model, the same US$24,000 buys 4.8 million credits at US$500 per 100,000.

The comparison is not like for like, because the profile models bundle a defined set of actions and the credit model charges for each one. But it shows the shape of the decision. If your use is mostly profile building at a stable customer count, the flat model is predictable and likely cheaper. If your use is spiky, heavy on querying and activation, or growing slowly from a small base, credits track what you actually use. A scenario-modelling workload that refreshes often and queries widely leans toward credits, and the pricing page offers a calculator for estimating the number you need.

#Who needs a seat

Seat counts in analytics are easy to inflate. A sensible split has three groups: the few people who build and govern models, the larger group who view results and ask questions, and the executives who only want an answer in Slack or an email. The analytics page lists native Slack integration with Tableau Next, which may let you serve the third group without a full seat, but confirm that in the contract, because an integration that exists is not a licence that covers it. Count the first group carefully, and ask whether your edition's bundled entitlement covers builders or only viewers.

#Exit and portability

A scenario model is a piece of business logic. Before you invest in a hundred of them, ask what you can take with you. Can the model definitions be exported in a readable form? Can the underlying data stay in your own warehouse and be read in place, rather than copied into a vendor store? The Data 360 page lists zero-copy sharing as a chargeable action, which is a reminder that avoiding duplication still has a price. A modelling stack whose logic lives only inside a proprietary format is harder to leave than one that can be rebuilt from a semantic layer you control.

#A thirty-day pilot design

A modelling licence is easiest to justify with a pilot that has a single decision and a clock. Choose a decision your leaders already make, such as how to package two product tiers or which segment to prioritise next quarter. Write down how the decision is made today and who makes it. Then run the scenario model alongside the current process for thirty days and compare the answer, the time taken and the confidence of the people involved.

Keep a ledger as you go. Record the Data 360 credits consumed by each refresh, the Agentforce actions triggered, and the seats used. That ledger lets you extrapolate honestly, and it also reveals which part of the stack is driving the bill. Teams are often surprised to find that data preparation, not the model itself, is the largest consumer.

At the end, ask the plain question: would the people who made the decision pay for this out of their own budget? If the answer is no, the capability may still be useful, but the case for a dedicated licence is weak, and the cheaper course is to use whatever your edition already includes.

#Governing a shared credit pool

Because Data 360 and Agentforce can draw on the same credits, a pool without an owner invites quiet overspend. Name one person who owns the balance, set threshold alerts in Salesforce's Digital Wallet, which the pricing pages describe for tracking consumption, and require teams to register a new use case before it draws credits. Review the balance monthly. Since unused credits do not roll over at the end of a term, a pool that is under-used is as much a problem as one that is exhausted, and the review should be as ready to reduce the next purchase as to increase it.

#Where the neighbouring questions are answered

The close and what the announcement left unsaid are on what happened to Spindle AI. How a figure should be traced back to records is in integrating Spindle AI with Salesforce. Whether to change how you forecast at all is covered in should I migrate from Spindle AI.

#What to ask before you commit

Ask which Tableau Next entitlement your edition already includes, what the Data 360 model is and how credits are metered for the refresh cadence you intend, and whether credits can move between Data 360 and Agentforce under your contract. Ask for the credit rate and the overage rate in writing. Then ask what you lose if you stop: whether models and scenario definitions can be exported and in what form.