First, un-confuse the product
Microsoft sells three things called Copilot, and this query is about the least understood of them. Microsoft 365 Copilot is the assistant your employees chat with inside Word and Outlook; its alternatives are graded in Microsoft 365 Copilot alternatives. Copilot Studio is the factory: a platform where you design agents, connect them to data and systems, set their actions, and publish them into Teams, SharePoint, websites, and other channels. Comparing Studio to ChatGPT is a category error; its rivals are other agent factories.
Studio's metering is the second thing to get straight, because every alternative meters differently and the comparison is meaningless until the units are on the table. Per Microsoft Learn, Copilot Studio capacity comes in packs: $200 per pack per month buys 25,000 Copilot Credits, and agent activity consumes credits at rates that depend on the action, with a pay-as-you-go option at $0.01 per credit. An agent that answers a question from its knowledge base costs a few credits; an agent that runs flows and calls premium connectors costs more. Your bill is therefore a function of agent behavior, not headcount, which is either liberating (idle agents are nearly free) or terrifying (a chatty agent is an unbounded meter), depending on your governance.
| Platform | Build model | Meter | Entry economics |
|---|---|---|---|
| Copilot Studio | Low-code studio in the Microsoft tenant | Copilot Credits | $200 per 25,000-credit pack per month, or $0.01 per credit PAYG |
| Salesforce Agentforce | Agent suite inside Salesforce | Per conversation and platform pricing | Quote-led; lives on Salesforce data |
| Dust | Per-seat agent builder for teams | Per user | $29 per user per month (vendor's own blog) |
| Zapier Agents | Agents on the Zapier automation base | Tasks and plan tiers | Rides Zapier plans; Professional from $19.99 |
| n8n | Workflow engine with AI nodes | Executions | Cloud from $20 per month yearly; self-hosted free |
| Botpress, Dify, Langflow | Open builder platforms | Cloud tiers or self-host | Self-host options; cloud prices vary (see notes) |
The enterprise suites: Agentforce, ServiceNow, Gemini Enterprise
If the reason you are pricing Studio is that your agents must live where your enterprise data and permissions already live, the honest rivals are the other platform vendors' agent suites. Salesforce Agentforce builds agents on Salesforce data and workflows; ServiceNow's AI Agents build on the ServiceNow platform's process data; Google's Gemini Enterprise is the agent layer for Google shops. Each inherits its parent platform's identity, security, and audit model, which is the entire argument for Studio inside a Microsoft tenant, mirrored. The selection rule is brutal and simple: agents belong to the platform whose data they act on. A Studio exit that lands your agents on a platform your data does not live in buys a second integration project disguised as a savings. These suites are quote-led and enterprise-sold; this dataset prints no figures for them rather than inventing any, and your procurement process, not a blog table, is where their prices will come from.
The per-seat builder: Dust
Dust is the cleanest conceptual alternative: a platform where teams build agents over their company knowledge (documents, tickets, wikis) and use them in a chat surface, priced per user at $29 per user per month on the vendor's own blog. The contrast with Studio is the meter: seats are forecastable in a way credits are not, and Dust's agents are knowledge agents first, strongest at answering and drafting from company sources, lighter on the deep transactional automation Studio reaches through Power Platform connectors. Choose Dust when the agents your company actually wants are knowledge agents for employees, and when a per-seat bill your finance team can predict beats a credit meter they cannot.
The automation platforms growing agents: Zapier and n8n
Zapier Agents and n8n approach agents from the automation side: the agent is a layer over workflows and integrations you may already run. Zapier's advantage is its integration catalog and a builder non-engineers can operate; its metering rides the task-based plans (Professional from $19.99 per month, scaling with volume). n8n's advantage is the execution meter (cloud from $20 per month billed annually for 2,500 executions; the self-hosted Community Edition is free) and an engineer's control over every step, with AI nodes and agent patterns built on top of ordinary workflows. Neither matches Studio's governance inside a Microsoft tenant, and neither needs a tenant at all. For agents whose job is to move data between SaaS tools and make small decisions on the way, this class is usually the pragmatic winner; for agents that must act inside Teams with Microsoft identity and compliance, Studio's home-field advantage is real.
The open builders: Botpress, Dify, Voiceflow, Langflow
The fourth class trades platform polish for ownership. Botpress is purpose-built for customer-facing chatbots and agents, with a visual studio and channel publishing; its cloud pricing could not be read in a verifiable form for this dataset, so its prices are recorded as honest nulls rather than guessed. Dify is an LLM application platform with RAG, workflows, and observability, self-hostable under a modified Apache 2.0 license whose extra conditions (including restrictions on multi-tenant service use) make it source-available rather than plain open source; its cloud prices are likewise carried as nulls. Voiceflow serves conversation designers building voice and chat agents; Langflow is the visual builder over the LangChain ecosystem for teams who want code-adjacent control. The shared argument for this class: your agents, your infrastructure, your model choices, and no credit meter. The shared cost: you own uptime, upgrades, and the security of a system that holds your customers' conversations.
The credit math, worked once
Because Studio's meter is the least intuitive, work it once with the published rates. A capacity pack is $200 for 25,000 credits, an effective $0.008 per credit; pay-as-you-go is $0.01 per credit. If a support agent's average conversation consumes, say, a handful of credits for knowledge answers and more when it triggers flows, a month of ten thousand conversations lands somewhere between a fraction of a pack and several packs; the honest statement is that nobody knows their credit consumption until a pilot agent runs against real traffic, which is why the checklist below starts with a metering pilot rather than a feature bake-off. Any vendor comparison that skips this step is comparing brochures.
