AI agents for business: a buyer's guide to 10 platforms

Buying an AI agent starts with two questions: what work will it finish, and what event appears on the bill?

Platforms sell seats or workspaces, meter usage, or combine both shapes. A monthly price alone tells you very little.

This guide compares ten third-party platforms through screenshots of their public product and pricing pages.

Short on time? The 60-second shortlist names two candidates for each situation.

Key takeaways

• Cassidy AI, CrewAI, Relevance AI and standalone Microsoft Copilot Studio publish no usable dollar rate.
• Gumloop's $37 is a slider floor, Zapier's $33.33 requires annual prepayment, and n8n shows euros with annual billing selected.
• Tasklet and Lindy publish paid entry rates without a billing qualifier. Salesforce publishes seven figures across two buying models and routes every paid option through Contact us.

Shortlist in 60 seconds

• Free start for a team of 3: Cassidy AI publishes 3 seats and 10,000 monthly credits; Zapier Agents publishes 400 monthly activities and web browsing. Disqualifier: Cassidy's fourth seat requires the unpriced Business path.
• Shared access for a team of 10 to 50: Gumloop publishes unlimited seats from $37/month at its 20k-credit slider floor; n8n publishes unlimited users on Starter at 20€/mo billed annually. Disqualifiers: Gumloop puts RBAC and SCIM/SAML in its unpriced Enterprise tier; n8n puts SSO, SAML and LDAP first on Business at 667€/mo billed annually.
• Paid workflow work below $50/month: Tasklet is $25/month with an agent browser; Gumloop starts at $37/month with unlimited seats. Disqualifier: Gumloop puts RBAC and SCIM/SAML in its unpriced Enterprise tier.
• Inbox or calling work: Lindy documents email and meeting tasks; Relevance AI lists calling and meeting agents. Disqualifier: Relevance AI has no published price or self-serve tier.
• Technical deployment control: n8n publishes execution billing and a self-hosted Business plan; CrewAI documents GitHub plus cloud, VPC and own-infrastructure deployment. Disqualifier: CrewAI publishes no paid dollar rate.
• An existing enterprise suite: Microsoft Copilot Studio connects its buying paths to Microsoft 365 and Azure; Salesforce Agentforce documents CRM objects and customer conversations. Disqualifier: standalone Copilot Studio has no published rate on the captured page.

Take the pair that matches your situation to the three-step comparison.

What an AI agent for business actually is

Here, an AI agent for business works toward an assigned outcome across connected tools. It keeps context, chooses permitted actions and returns a result for review.

Chat assistants answer in a conversation. Workflow automation follows a preset trigger and action path. An agent can use both, with chat supplying the request and fixed workflows setting guardrails.

Ask what consumes a unit, whether retries count, whether units expire and whether the team shares one pool.

AI coworker, AI teammate, AI agent: what the category is being called

"AI coworker" and "AI teammate" are category labels. Cassidy AI calls its agents AI teammates, and Relevance AI calls its product an AI Workforce.

Compare where it runs, what it remembers, which actions need approval and how work returns.

How we evaluated

This guide is built from what these vendors publish. We read each platform's public product pages, pricing pages and docs, and we compare billing shape, published limits and documented fit. It is not a head-to-head bake-off, and nothing here ranks the products by output quality.

Each platform was read against four weighted criteria, applied only to what its own public pages and docs disclose:

• Billing clarity (40%). Is a price published, is the billable unit defined, and can a buyer model a real month from the page alone?
• Published limits (25%). Seats, credit allowances, execution caps, concurrency and rollover terms, stated as numbers rather than adjectives.
• Governance disclosure (20%). Role controls, audit logs, SSO and SAML, retention and deployment options, and whether the price of those controls is published or gated.
• Documented fit (15%). Which stack, team size and job the vendor's own material says the product is for.

The weights describe how much each criterion moved our read of a platform. We publish no composite score.

All prices were checked on each vendor's own pricing page on 30 July 2026. Every other figure comes from the vendor pages listed under Sources.

No affiliate links. No sponsored placements.

The tools at a glance

Every table cell below comes from the vendor's own pages, linked in each entry.

Jump to a platform:

• Cassidy AI
• Zapier Agents
• Tasklet
• Lindy
• Gumloop
• n8n
• CrewAI
• Microsoft Copilot Studio
• Salesforce Agentforce
• Relevance AI

Four of ten vendors publish no usable price

Cassidy AI has two plans, both labelled Start for free, with no dollar figure. CrewAI publishes one hard number, a 50-workflow-execution monthly cap on Free, while Enterprise is Custom. Relevance AI has one Enterprise plan, a Talk to sales button and no price.

Standalone Microsoft Copilot Studio is credit-metered, with no per-credit rate on the captured page. Its only list price is $30.00 user/month, paid yearly, for the separate Microsoft 365 Copilot licence that includes Copilot Studio access.

Three published numbers need a qualifier. Gumloop starts at $37/month with its slider at the 20k-credit floor. Zapier Agents shows $33.33/month against a $400 annual prepayment. n8n shows euros with annual billing selected.

Exactly two platforms publish a paid entry price a buyer can read and act on without a billing qualifier: Tasklet at $25/month and Lindy at $49.99/month. Salesforce Agentforce publishes seven figures across two tables; every paid option has a Contact us button, while the $0 Foundations card says Try for free.

A missing price is a data point about the sales process, not an oversight.

How to choose: six questions to ask before you buy

1. What creates a billable unit? Define the credit, activity, execution or conversation, then price a real weekly workflow.
2. Which systems can the agent read and write? Request permission scopes, credential ownership, retention terms and actions that change a system of record.
3. Where does human approval happen? Ask to see the action preview, approver, timeout and decision record.
4. How are identity and governance handled? Confirm role controls, audit logs, account removal and the price of required controls.
5. What happens on failure or retry? Establish whether failed work consumes usage, who sees errors and what prevents duplicate actions.
6. Can the company leave cleanly? Ask how to export prompts, workflows, knowledge, logs and credentials.

Patricia

Patricia is made by Better Group, which also builds Runflow.

The three product screenshots here come from our own workspace.

Patricia's pricing file and agent skill put the published plan at $45/workspace/month, including 20,000 monthly credits and covering every seat, client and channel.

Our workspace runs on non-standard founding-plan credit terms and consumes well beyond that allowance. Its daily use doesn't show what the published plan buys.

Limitation: Patricia documents no public API, MCP server, programmatic access, SSO, SAML, SCIM, image generation or video generation.

Compared with CrewAI, CrewAI publishes GitHub integration; Patricia documents no programmatic access.

Ten AI agent platforms for business

The order carries no quality verdict. Each entry separates list prices, qualified rates, free labels and custom quotes.

1. Cassidy AI

Cassidy AI is an agent and workflow builder with a knowledge base and meeting tools. Its homepage calls the agents AI teammates anyone can build.

The Cassidy pricing page shows a free Starter plan with 3 seats, 1 workspace, 10,000 AI credits per month and 30,000 pages of storage. Business moves to custom seats, workspaces, credits and storage, with premium onboarding and support. Neither plan carries a dollar price.

Best for: A small team that wants several agents and workflows inside one shared workspace before entering a sales process.

Limitation: The free Starter plan caps a team at 3 seats and 10,000 credits per month. A fourth person means entering a sales process because Cassidy publishes no Business price.

Compared to Zapier Agents: Zapier publishes a free 400-activity pool with web browsing, which Cassidy does not publish.

2. Zapier Agents

Zapier Agents extends Zapier's app-connected automation model to agent behaviors, chat, web browsing and attached knowledge.

The Zapier pricing page lists Agents Free as free forever with 400 activities per month. Agents Pro is $400 billed annually, shown as $33.33/month, with 1,500 activities per month. The pay-monthly rate was not selected in the captured page.

Best for: Teams that already understand Zapier and want agent actions close to their existing app connections.

Limitation: Pro caps usage at 1,500 activities per month, and $33.33/month is available only as a $400 annual prepayment in the captured plan.

Compared to Tasklet: Tasklet documents an isolated cloud terminal and agent browser, and its $25 Starter rate has no annual-prepay qualifier.

3. Tasklet

Tasklet connects services to a sandboxed command-line environment and adds a dedicated web browser on paid plans.

The Tasklet pricing page lists Free at $0/month with 300 bonus credits per day, 10 executions per trigger and a limited file upload size. Starter costs $25/month for 10,000 monthly credits, while Pro costs $100/month for 40,000. Custom shows $250/month at the dropdown's default 100,000-credit option.

Best for: Teams that want agents to work across web pages, connected services and command-line tasks under a credit budget.

Limitation: The move from Starter to Pro multiplies both the allowance and the price by four. Free also limits each trigger to 10 executions and restricts file upload size.

Compared to Lindy: Lindy documents email, calendar and meeting work, with inbox caps published on its plan cards.

4. Lindy

Lindy focuses on inbox, meeting, calendar and follow-up work. Its plans add inbox capacity and usage as the price rises, while Pro adds computer use and model choice.

The Lindy pricing page lists Plus at $49.99/month with standard usage and up to 2 inboxes. Pro is $99.99/month with 3x more usage than Plus and up to 3 inboxes. Max is $199.99/month with 7x more usage than Plus and up to 5 inboxes. The page has no monthly or annual billing toggle.

Best for: Individuals or compact teams that want one product centered on email, scheduling, meeting preparation, notes and follow-ups.

Limitation: The page states multiples against Plus without defining the base unit. Plus caps capacity at 2 inboxes, while SSO, SCIM and HIPAA sit behind the Enterprise Talk to sales path.

Compared to Gumloop: Gumloop publishes a shared multi-agent canvas, unlimited Pro seats and 5 concurrent workflow runs.

5. Gumloop

Gumloop combines credit-metered workflows, agent chats and team access. Its Pro plan includes unlimited seats and teams.

The Gumloop pricing page is slider-driven. The captured slider is at its lowest stop, so Pro starts at $37/month for 20k credits per month and the number rises with the slider. Its published stops run from 20k to 1.5M credits per month. At the entry setting, Pro includes unlimited seats, 5 concurrent workflow runs, 25 concurrent agent chats and one hosted MCP server.

Best for: Operations teams that want shared agent and workflow capacity without buying a seat for each collaborator.

Limitation: Only Pro and Enterprise are shown. Pro caps workflow concurrency at 5 runs. RBAC, SCIM/SAML and audit logs are Enterprise-only, and the public page does not publish that tier's price.

Compared to n8n: n8n publishes workflow-execution billing and a self-hosted Business option.

6. n8n

n8n is a workflow automation platform for technical teams. It bills each completed workflow execution regardless of step count, and all captured plans include unlimited users and workflows.

The n8n pricing page had its Annually toggle selected with a stated 17% saving. Starter is 20€/mo billed annually for 2,500 workflow executions, Pro is 50€/mo billed annually for 10,000, and Business is 667€/mo billed annually for 40,000 on a self-hosted plan. Enterprise pricing is custom. We did not capture the monthly-billing rate.

Best for: Technical teams that want workflow automation with hosted and self-hosted deployment options.

Limitation: Starter is capped at 2,500 executions per month. SSO, SAML and LDAP first appear on Business at 667€/mo billed annually, more than 13 times Pro's 50€/mo annual-billing rate.

Compared to CrewAI: CrewAI documents GitHub integration and deployment to CrewAI cloud, a buyer's VPC or the buyer's own infrastructure.

7. CrewAI

CrewAI joins a visual editor and AI copilot with GitHub integration, then places governance and deployment controls in Enterprise.

The CrewAI pricing page lists Free with 50 workflow executions per month. Enterprise is Custom and adds governance, SSO, RBAC, and deployment on CrewAI cloud, a buyer's VPC or the buyer's own infrastructure.

Best for: Engineering-led teams that need a path from agent workflow design to a centrally governed runtime across several deployment environments.

Limitation: Free is capped at 50 workflow executions per month, and no dollar figure is published anywhere on the captured pricing page.

Compared to Microsoft Copilot Studio: Microsoft publishes a Microsoft 365 Copilot licence that includes Copilot Studio access for internal agents.

8. Microsoft Copilot Studio

Microsoft Copilot Studio is Microsoft's platform for building and managing agents. Its buying paths connect agent building to Microsoft 365 licensing or a separate credit commitment and Azure billing.

The Microsoft Copilot Studio page lists Microsoft 365 Copilot at $30.00 user/month, paid yearly. That is a separate licence which includes Copilot Studio access for internal agents. Standalone Copilot Studio has Pre-purchase and Pay-as-you-go options, and neither shows a price. Both require an Azure subscription.

Best for: Organizations whose users, identity and daily work already sit in Microsoft 365 and Azure.

Limitation: The captured page publishes no per-credit rate for standalone Copilot Studio, an Azure subscription is required, and the only list price belongs to a different SKU.

Compared to Salesforce Agentforce: Salesforce documents agent actions beside CRM objects and customer conversations.

9. Salesforce Agentforce

Salesforce Agentforce places agents beside Salesforce users, CRM objects and customer conversations. Its pricing page uses Flex Credits, conversations and user licences.

The Salesforce Agentforce pricing page publishes seven price points across two tables. The consumption table lists Salesforce Foundations at $0, Flex Credits at $500 USD per 100,000 credits, and Conversations at $2 USD per conversation on a pre-purchase-only model. The licence table lists Agentforce add-ons at $125 user/month for Sales, Service and Field Service; Agentforce Industries add-ons at $150 user/month; Agentforce 1 Editions from $550 user/month with 2.5M Flex Credits per org per year; and the Agentforce User License at $5 USD/user/month, which requires Flex Credits.

Best for: Salesforce customers that want agent actions close to CRM records and can govern usage inside an existing Salesforce buying relationship.

Limitation: The page has seven price points across two buying models. Every paid option says Contact us, and Salesforce says the page is provided for information purposes only and is subject to change.

Compared to Relevance AI: Relevance AI publishes unlimited agents, tools, users and projects plus calling and agent evaluations on Enterprise.

10. Relevance AI

Relevance AI describes its product as the home of the AI Workforce. Its Enterprise tier includes unlimited agents, tools, users, projects and workforces.

The Relevance AI pricing page shows one Enterprise tier with a Talk to sales call to action and no published price. The page lists 2,000+ integrations, calling and meeting agents, enterprise triggers, agent evaluations, A/B testing, analytics, SSO, RBAC and audit logs.

Best for: Companies seeking an enterprise agent platform with workforce management, calling, meetings and governance in one sales-led package.

Limitation: Relevance AI publishes one Enterprise plan, no price and no self-serve tier, so a buyer cannot size or trial it before a sales process starts.

Compared to CrewAI: CrewAI publishes a Free 50-execution allowance and cloud, VPC or own-infrastructure deployment paths.

Types of AI agent platforms for business

This classification groups each third-party platform by the product shape described in its linked entry. Each platform appears once.

• Packaged AI teammates: Cassidy AI puts agents, workflows and a knowledge base in a shared workspace. Lindy centers its product on inbox, calendar and meeting work.
• Computer-using task agents: Tasklet connects services to an isolated cloud terminal and adds a dedicated browser on paid plans.
• App-connected no-code agents: Zapier Agents adds agent behaviors to Zapier's app connections. Gumloop combines a visual multi-agent canvas, agent chats and shared team access.
• Technical workflow and agent runtimes: n8n bills completed workflows and publishes hosted and self-hosted options. CrewAI joins visual agent design, GitHub and several deployment environments.
• Suite-native agent builders: Microsoft Copilot Studio sits inside Microsoft 365 and Azure buying paths. Salesforce Agentforce sits beside Salesforce users, CRM objects and customer conversations.
• Enterprise AI workforces: Relevance AI publishes one Enterprise plan with unlimited agents, tools, users, projects and workforces.

How billing shape and stack affect the budget

The 60-second shortlist is the single source of truth for candidate pairs and disqualifiers. These checks are narrower: they show how headcount and an incumbent suite change the budget after you have a pair.

• Seat economics at 25 users: The separate Microsoft 365 Copilot licence at $30.00 per user each month, paid yearly, totals $9,000 a year for 25 users. Gumloop's Pro slider starts at $37/month with unlimited seats at its 20k-credit floor. RBAC and SCIM/SAML remain in Gumloop's unpriced Enterprise tier.
• Incumbent-suite billing: Microsoft 365 Copilot is a separate paid-yearly SKU that includes Copilot Studio access; standalone Copilot Studio publishes no rate and requires Azure. Salesforce's $5 User License requires Flex Credits, and every paid card routes through Contact us. Use the existing suite as an implementation tie-breaker.

AI agents for sales

Candidate selection and pricing disqualifiers stay in the 60-second shortlist. These sales examples explain the documented capabilities behind its entries.

• Lindy documents inbox, scheduling and meeting work. The shortlist lists it under inbox or calling work.
• CrewAI shows a sales meeting preparation flow tied to CRM, email and documents. The shortlist lists it under technical deployment control, with no published paid dollar rate.
• Salesforce Agentforce documents CRM objects and customer conversations. The shortlist lists it under an existing enterprise suite.
• Relevance AI lists calling and meeting agents. The shortlist lists it under inbox or calling work, with no published price or self-serve tier.

Define who approves outbound messages, and when a billable conversation or execution starts and ends.

AI agents for finance

Candidate selection and pricing disqualifiers stay in the 60-second shortlist. These finance examples describe documented functions only.

• Tasklet shows analysis and reconciliation commands in an isolated terminal. The shortlist lists it under paid workflow work below $50/month.
• n8n bills completed workflows and publishes hosted or self-hosted paths. The shortlist lists it under technical deployment control and shared access; SSO, SAML and LDAP first appear on Business at 667€/mo billed annually.
• Microsoft Copilot Studio connects to Microsoft 365 and Azure buying paths. The shortlist lists it under an existing enterprise suite; standalone Copilot Studio has no published rate on the captured page.
• Zapier Agents documents app-connected actions and metered activities. The shortlist lists it under a free start for a team of 3.

For ledger changes, payment messages or reports, require separate read and write permissions, a preview, named approvers, an audit trail and a retry policy. Ask to see the pause and completed output.

What we could not verify

• We did not verify the security or compliance posture of any product here. Establish every governance claim during procurement.
• Cassidy AI's Business tier and the Enterprise tiers from Zapier Agents, Lindy, Gumloop, n8n, CrewAI and Relevance AI have no published price in the source material. Standalone Microsoft Copilot Studio also has no rate on the captured product page.
• Zapier's pay-monthly Agents Pro rate and n8n's monthly-billing rates were not selected in our captures.
• Microsoft's Power Platform pricing page at https://www.microsoft.com/en-us/power-platform/pricing/ returned HTTP 403 to our automated capture on 2026-07-30, so we read no per-Copilot-Credit rate from it. Microsoft's reachable billing and licensing guide describes the credit mechanics without a rate.
• Our automated captures could not read prices from Sintra, agent.ai, Manus or Bardeen, so we publish no figure from those pages.

FAQ

What belongs on a list of AI agents for business?

The list covers Cassidy AI, Zapier Agents, Tasklet, Lindy, Gumloop, n8n, CrewAI, Microsoft Copilot Studio, Salesforce Agentforce and Relevance AI. Compare billing shape, then read each limitation.

Best AI agents for business

The 60-second shortlist is the candidate list and keeps each pricing disqualifier attached. Match your situation there, then apply the six buying questions.

Looking for AI agents examples?

The vendor screenshots show Tasklet running terminal work, Lindy triaging email, n8n routing employee onboarding and CrewAI laying out meeting preparation. They document interfaces and demo flows.

Custom AI agents for business

A custom agent makes sense when company-specific systems, permissions or output formats exceed a standard template. CrewAI documents several deployment paths, while Microsoft Copilot Studio and Salesforce Agentforce keep the build close to their respective suites. Define the finished artifact, allowed actions, review step, failure behavior and credential owner first.

What does Agent AI mean?

"Agent AI" is a search phrase for the product category covered here. Product names such as AI coworker, AI teammate and AI workforce say little about permissions or billing, so compare the documented limits in Cassidy AI, Gumloop and the other entries.

What can this guide answer for an AI agents McKinsey search?

A reader entering that phrase likely wants a category map and a way to turn adoption interest into a buying process. Use the platform types to separate product shapes, then the six buying questions and shortlist to compare billing units, controls, failure handling and workflow fit. This guide cites no McKinsey finding, statistic or report. McKinsey's own publications are the source for its research.

Which AI agent is best for a small team?

Use the free-start row in the 60-second shortlist for a shared team of up to 3. It keeps Cassidy AI's fourth-seat sales path attached. Price one weekly workflow because each product meters a different unit.

What is the best free AI agent for business?

Use the free-start row in the 60-second shortlist. It keeps the team limit and next paid step beside its two candidates. Price the real task against those caps before building around either product.

Are AI agents safe with sensitive company data?

Use the six buying questions and the governance criterion to request role controls, audit logs, SSO, retention, credential ownership and an approval record. The Lindy, Gumloop and n8n entries show where some controls first appear in published plans; verify the required control in your own procurement process.

Where to start

1. Pick two candidates from the 60-second shortlist.
2. Run both through the same real workflow with the same inputs, permissions and approval requirement.
3. Record every billable event, the approval step, failure behavior and which required controls cost extra. Compare those records before choosing.

Sources

1. Cassidy AI, homepage: https://www.cassidyai.com/.
2. Cassidy AI, pricing: https://www.cassidyai.com/pricing.
3. Zapier, Agents product page: https://zapier.com/agents.
4. Zapier, Agents pricing tab: https://zapier.com/pricing.
5. Tasklet, homepage: https://tasklet.ai/.
6. Tasklet, pricing: https://tasklet.ai/pricing.
7. Lindy, homepage: https://www.lindy.ai/.
8. Lindy, pricing: https://www.lindy.ai/pricing.
9. Gumloop, homepage: https://www.gumloop.com/.
10. Gumloop, pricing: https://www.gumloop.com/pricing.
11. n8n, homepage: https://n8n.io/.
12. n8n, pricing: https://n8n.io/pricing/.
13. CrewAI, homepage: https://www.crewai.com/.
14. CrewAI, pricing: https://www.crewai.com/pricing.
15. Microsoft Learn, What is Microsoft Copilot Studio: https://learn.microsoft.com/en-us/microsoft-copilot-studio/fundamentals-what-is-copilot-studio.
16. Microsoft, Copilot Studio product and pricing page: https://www.microsoft.com/en-us/microsoft-copilot/microsoft-copilot-studio.
17. Microsoft Learn, Copilot Studio billing and licensing: https://learn.microsoft.com/en-us/microsoft-copilot-studio/billing-licensing.
18. Microsoft, Power Platform pricing: https://www.microsoft.com/en-us/power-platform/pricing/, returned HTTP 403 to our automated capture on 2026-07-30.
19. Salesforce, Agentforce Builder: https://www.salesforce.com/agentforce/agent-builder/.
20. Salesforce, Agentforce pricing: https://www.salesforce.com/agentforce/pricing/.
21. Relevance AI, homepage: https://relevanceai.com/.
22. Relevance AI, pricing: https://relevanceai.com/pricing.
23. Patricia, homepage: https://www.patricia.app/.
24. Patricia, pricing file: https://www.patricia.app/pricing.md.
25. Patricia, published agent skill: https://www.patricia.app/.well-known/agent-skills/patricia/SKILL.md.