ThoughtSpot AgentSpot: Pay per Outcome, Not per Token

ThoughtSpot has taken its biggest step yet toward the agent era. On October 22, 2026, the analytics vendor launched AgentSpot, a workspace where employees across a company build AI agents, automate workflows, and create interactive apps in plain language — and share everything they build so it compounds instead of staying trapped on one person's laptop.

The pricing is the part that will turn heads: ThoughtSpot charges by outcomes, not tokens or seats. You pay when an agent finishes a piece of work — a conversation, a workflow run, or an app session. Building and testing cost nothing, and every plan includes unlimited users and unlimited agents.

Quick answer

• What it is: AgentSpot is a shared "multiplayer AI" workspace from ThoughtSpot for building goal-based agents, scheduled workflows, and data apps — no code required.
• Launch date: October 22, 2026 (Mountain View, California).
• Pricing: Pay per outcome — one credit per agent conversation, workflow run, or app session. Free plan includes a weekly credit allowance; building and testing is free.
• The catch: ThoughtSpot has not published per-credit or per-plan dollar prices. The paid "Fleet" plan's price is a quote-only affair for now.
• Who it's for: Companies already running on ThoughtSpot's data platform get the most value, since agents draw on its governed semantic layer.

What AgentSpot actually does

ThoughtSpot's pitch is simple: the first wave of enterprise AI gave every employee a chatbot, and the result was "single-player" work. A sales rep builds an account-research agent, an analyst builds a variance-check workflow, and neither artifact ever leaves the person who made it. Colleagues duplicate the same work in the next quarter because nobody knew it existed.

AgentSpot moves that work into a shared workspace. It has three building blocks:

• Agents — describe the job in natural language and the workspace assembles an agent with instructions, business context, and tool access. Agents investigate questions, analyze data, recommend next steps, and act within their permissions.
• Workflows — recurring routines that run on a schedule or a trigger: weekly sales forecasts, financial monitoring, employee onboarding steps.
• Apps — live, interactive business applications generated from a prompt, combining data from connected systems with agent-powered actions, with each user seeing only what they are authorized to see.

A well-built agent becomes a company asset: the top rep's account research agent gets reused by the whole sales team, and one analyst's variance workflow runs in every business unit. That "build once, share everywhere" loop is what ThoughtSpot means by multiplayer AI.

The four layers of memory

Memory is AgentSpot's compounding mechanism. Instead of every conversation starting from zero, the workspace accumulates knowledge at four levels:

• Company memory — administrators set key facts about the business, and the shared understanding grows from agent conversations over time.
• Agent memory — creators refine how an agent behaves, and each improvement reaches everyone who uses it.
• User memory — the system learns each person's preferences so answers get more relevant over time.
• Agent-user memory — retains each person's history with a specific agent, so follow-ups don't require re-explaining context.

Underpinning all four is ThoughtSpot's semantic layer — more than a decade of work defining what business terms like "revenue" or "at-risk account" mean inside a given company. Agents query the same definitions that already power the dashboards, so an agent's number should match what a manager sees in a report. That's the structural argument for building agents on an analytics platform rather than on a blank LLM canvas.

Guardrails, not roadblocks

Enterprise IT teams will notice the governance language in the launch: permissions ensure people see only data they are authorized to access, and always-on audit trails log every action an agent takes — what was done, when, and on whose behalf. In practice, governance is what lets a company hand agent-building tools to sales, finance, HR, legal, and marketing staff without the central AI team reviewing every prompt.

The pricing: outcomes instead of tokens

Here's the part enterprise buyers will actually model in a spreadsheet. Traditional AI pricing comes in two flavors: per seat (Tableau Cloud Standard Creator at $75 per user per month, per October 2026 BI pricing data) or per token/credit (where a bill can balloon because nobody knows how much model work a workflow will consume in production).

AgentSpot discards both. The billing unit is the outcome: an agent conversation, a workflow run, or an app session. As ThoughtSpot SVP of product management Francois Lopitaux put it to TechTarget: if a user creates an AI workflow and runs it five times a week, it costs five credits. "It's a nightmare, and it's slowing down adoption," he said of token-based pricing — "that's why we wanted to come out with a pricing that everyone can understand, that is practical, and that is not linked to a model."

The fine print, from the launch announcement and TechTarget's reporting:

• Free plan: available today, no time limit, no credit card required. Includes unlimited agents and unlimited users, with a weekly allowance of agent conversations, workflow runs, and app sessions. TechTarget's reporting puts that allowance at 25 credits per week.
• Fleet plan: for teams that need more usage, reported at 250 credits per week. Dollar pricing for Fleet was not published at launch.
• Building and testing are free — credits are only consumed when an agent delivers work.
• Every plan includes unlimited users and agents — headcount growth doesn't move the bill.

The honest read on pricing: the model is genuinely simpler than token billing, but the two numbers a buyer wants most — the dollar price of Fleet and the per-credit cost of extra usage — are not public. That keeps the launch honest-but-incomplete: you can understand the meter, but you can't price your bill until you talk to sales.

How it stacks up: outcome pricing vs the alternatives

Analysts quoted by TechTarget see the direction, with caveats. Moor Insights & Strategy's Mike Leone argued the big advantage is that "you pay for each piece of work an agent does, not for seats or tokens, so bringing the whole company onto a shared agent doesn't raise the bill" — while noting outcome-based pricing is still "gaining momentum" with unresolved complications about what counts as an outcome. William McKnight of McKnight Consulting called ThoughtSpot "on the leading edge of this" and worth watching, but flagged that the company must keep working on enterprise-grade agent evaluation harnesses to prevent drift and add transactional write-back safeguards for third-party systems.

What to watch next

Two things will decide whether AgentSpot is a pricing revolution or just a pricing relabel. First, the Fleet price: outcome pricing is only "practical" in Lopitaux's terms if the per-outcome cost is public and competitive. Second, the roadmap — ThoughtSpot's stated next step is embedding AgentSpot into ThoughtSpot Embedded, so customers can ship agents and workflows inside their own products rather than only inside ThoughtSpot's workspace. That would extend the model from internal productivity to customer-facing software.

Also worth remembering: vendor claims in the launch — 1,000+ customers, Gartner Magic Quadrant leadership — are company-reported. And the effectiveness of shared memory is untested in production at scale; Leone's condition is the right one to hold: "as long as the underlying data is properly prepared," collaborative agents should work. Poorly modeled data will produce agents that confidently reuse each other's mistakes.

FAQ

What is ThoughtSpot AgentSpot?

AgentSpot is a workspace launched October 22, 2026, where business users build AI agents, workflows, and apps in natural language and share them across their company. ThoughtSpot calls the model "multiplayer AI."

How is AgentSpot priced?

By outcomes, not tokens or seats. Each agent conversation, workflow run, or app session consumes a credit. The free plan includes a weekly credit allowance (reported at 25 credits) with unlimited users and agents; the paid Fleet plan (reported at 250 credits a week) has no published price yet.

Do you need to know how to code to use AgentSpot?

No. Agents, workflows, and apps are built by describing what you need in plain language, guided by an AI assistant. ThoughtSpot positions it for non-technical teams like sales, finance, HR, and marketing.

Is there really no charge for building and testing?

According to the launch announcement, building and testing agents is free — credits are consumed only when an agent delivers work. The free plan requires no credit card and has no time limit.

How is this different from Microsoft Copilot or other enterprise agents?

The differentiators ThoughtSpot claims are the governed semantic layer (agents use your company's own data definitions), four-layer shared memory, and outcome-based pricing with unlimited users. Analysts agree the sharing mechanics aren't unique to ThoughtSpot, but the anchoring to mature BI semantics is a genuine structural edge — if your data is properly modeled.

Key takeaways

• ThoughtSpot launched AgentSpot on October 22, 2026: a shared workspace for AI agents, workflows, and apps built in plain language.
• Pricing is outcome-based — one credit per conversation, workflow run, or app session — with unlimited users and agents on every plan.
• The free plan needs no credit card and reportedly includes 25 credits a week; Fleet's dollar price is not yet published.
• Four memory layers and a decade-old semantic layer are the structural bet: agents reuse your company's own definitions.
• Analysts like the pricing direction but note outcome definitions are still fuzzy across the industry, and production-grade agent evaluation is unfinished work.

Published October 24, 2026. Launch details are from ThoughtSpot's October 22 announcement; pricing allowances are per TechTarget's reporting, which is noted where figures are theirs, not the vendor's.

Internal links: Walmart's AI shopping agents with spending limits and Mastercard Agent Pay for machines. Official source: ThoughtSpot AgentSpot.