Technology research
AI Assistant Market Map
As the AI assistant market moves beyond chatbots that simply answer questions, AI assistants can increasingly remember context, use tools, take actions, coordinate work, and eventually pursue goals with much less human intervention. This map is an attempt to understand what these products can actually do today, who they are built around, and where the capabilities of AI assistants may be heading.
Click on a dot for the full write-up.
PersonalWork / Business
Autonomous
agent
Agentic
assistant
Traditional
assistant
PersonalWork / Business
Assistants List
Compare who each assistant serves, what it can do, and how independently it can act.
Click on a name or swipe for more→
| Assistant | Who it's for | What it does (in their own words) | Agency / Autonomy | What's unique | Memory & Context | Privacy & Trust | How to access it | Pricing |
|---|---|---|---|---|---|---|---|---|
| What's uniqueNo new interface: interact with it like a person (via SMS and WhatsApp), while it operates your phone and computer for you. | Individuals | Instinct is a personal assistant that understands what you're working on and what's important to you. It connects to your applications and devices — email, messaging, screen, audio, location, and more. The interface is simple: there are no new interfaces. It's trained to use a phone and a computer. You can text or call it. It uses its devices in the same way that humans do. Instinct's core model is trained to understand and handle the deeply personal nuances of everyday life — following up on threads you've dropped, proactively calling or texting you, arranging a ride to the airport, booking a handyman for your apartment, and much more. | Autonomous — acts through connected devices and services and can initiate follow-ups. | No new interface: interact with it like a person (via SMS and WhatsApp), while it operates your phone and computer for you. | Designed around deeply personal context, but public documentation does not yet clearly describe the full memory system. | Deep access to personal data is central to the product; public privacy detail is still limited, so trust is harder to assess. | SMS and WhatsApp; currently private/invite-only access. | Pricing not publicly disclosed; private access group. |
| What's uniqueA personal agent with its own secure computer/browser that can keep working after you close the app. | Individuals | Muse can handle tasks, like sending an email or booking travel, and it can take on big audacious goals. Once a person shares a goal with Muse, it helps them develop a personalized plan and coordinate their time and resources, then advances the work on its own. It can open a browser, fill out forms, and negotiate on their behalf. Muse uses its own secure computer/browser that can keep working after you close the app. | Autonomous — builds plans, works in the background and executes multi-step tasks, with approval for critical actions. | A personal agent with its own secure computer/browser that can keep working after you close the app. | Learns your goals and preferences; stores editable memory and working files in its dedicated Secure VM. | Strong technical controls and user approvals, but Muse can access deeply personal connected data; Meta says conversations/VM data are not shared with its ad systems and training can be opted out. | Dedicated Muse app and WhatsApp; U.S. launch. | Free with usage limits; paid tiers reported at $20/mo and $100/mo. |
| What's uniqueMix of AI + human assistant: AI layer plans the approach — who to call, what to say, etc. Then real, background-checked humans execute it. | Individuals; Households | Duckbill handles the messy, complex, real-world part — phone calls, errands, hold queues, arguments — so things actually get finished. Examples: negotiate a lower bill, call about a denied insurance claim, compare quotes, find the best price on a mattress, reserve a restaurant. | Execute — AI plans and researches; background-checked human specialists execute the real-world work. | Mix of AI + human assistant: AI layer plans the approach — who to call, what to say, etc. Then real, background-checked humans execute it. | Primarily task-specific context and account/task history; not positioned as a deep personal-memory AI. | Human-in-the-loop creates a different trust model: background-checked operators handle execution, while AI handles research and planning. | App, text and email. | $49/mo Essentials; $99 Individual; $169 Household; $350 Household Plus. |
| What's uniqueOpen-source personal assistant with its own identity, memory and computer-like environment. You can choose where your data lives, including self-hosting. | Individuals; Professionals | Built for busy business owners: offload work and make more time for life. Vellum is designed for continuous daily task execution. It runs ambiently in the background, follows schedules, monitors feeds, and operates across all your apps without draining your monthly reasoning tokens. | Autonomous — runs continuously with schedules, heartbeats and multi-step actions, while sensitive actions can require approval. | Open-source personal assistant with its own identity, memory and computer-like environment. You can choose where your data lives, including self-hosting. | Persistent long-term memory, workspace files, identity/personality and conversation history; remembers you across conversations. | Strong user control: self-hosting keeps data on your machine; cloud assistants run in isolated environments; Vellum says it does not train on your data. | Web, Mac, Windows, iPhone/iPad, voice, phone, Telegram, Slack and email. | Free Base; paid plans start at $30/mo, with $100/mo and $200/mo tiers; usage credits are also available. |
| What's uniqueCombines a personal assistant with reusable routines, skills, person research and assistant-to-assistant collaboration. | Individuals; Teams; Businesses | Learns how you work, drafts and executes work, runs routines, handles inboxes, meetings and research, and can work across email, Slack, WhatsApp and Telegram. | Execute → Autonomous — runs routines and background work; external actions such as sending email require approval. | Combines a personal assistant with reusable routines, skills, person research and assistant-to-assistant collaboration. | Builds context from your work, routines, integrations, documents and history; designed around an evolving work wiki. | Strong public trust posture: Town says it does not sell/share/train on your data, encrypts data, and requires approval for external actions. | Web/app plus email, Slack, WhatsApp and Telegram. | Free; Starter $15/mo; Pro $49/mo; Power $99/mo; Power Plus $199/mo; team plans available. |
| What's uniqueBroadest general-purpose assistant ecosystem combining frontier models, tools, memory, research and agentic capabilities. | Individuals; Professionals; Teams; Businesses | General-purpose assistant for research, writing, analysis, coding, image generation, planning, files and increasingly agentic multi-step work. | Execute → Autonomous — can use tools, connected apps and browser workflows to complete multi-step tasks, with confirmations for higher-impact actions. | Broadest general-purpose assistant ecosystem combining frontier models, tools, memory, research and agentic capabilities. | Rich persistent memory and past-chat context; can remember preferences, projects and recurring context, with user controls. | Strong controls but consumer privacy is configurable rather than absolute: memory can be managed, connected apps are permissioned, and users can opt out of model improvement. | Web, iOS, Android, desktop apps and connected apps. | Free; Plus $20/mo; Pro $100/mo; Pro $200/mo tier exists but new signups/upgrades are temporarily paused. |
| What's uniqueStrong combination of long-context knowledge work, coding and careful, user-directed collaboration. | Individuals; Professionals; Teams; Developers | General-purpose assistant for writing, analysis, coding, research, documents, projects and multi-step knowledge work. | Assist → Execute — increasingly capable of tool use, research and connected workflows, but generally more user-directed than autonomous personal agents. | Strong combination of long-context knowledge work, coding and careful, user-directed collaboration. | Projects, preferences and past-chat search provide persistent context; memory is more workspace/conversation-oriented than a full personal-life model. | Strong privacy posture: consumer model-improvement use is opt-in; commercial data is not used for training by default; users can export/delete data. | Web, iOS, Android, desktop and API. | Free; Pro $20/mo; Max $100/mo or $200/mo; Team $30/user/mo monthly. |
| What's uniqueDeep integration with Google's ecosystem, multimodal capabilities and access to Google Search, Gmail, Docs and other services. | Individuals; Professionals; Google users; Businesses | General-purpose assistant for research, writing, multimodal analysis, coding, planning and tasks across Google products and connected apps. | Execute — can use connected Google apps and agentic features, but much of the experience remains conversational and user-directed. | Deep integration with Google's ecosystem, multimodal capabilities and access to Google Search, Gmail, Docs and other services. | Can use memory of past chats to understand you and your world; Google ecosystem context can extend the assistant's usefulness. | Trust is mixed by design: Google provides substantial privacy controls, but Gemini processes a broad range of personal data and activity; users can control activity and personalization settings. | Web, iOS, Android and deeply integrated across Google products. | Free; Google AI Pro $19.99/mo; Google AI Ultra $249.99/mo. |
| What's uniqueCombines answer engine + source transparency with increasingly agentic Computer capabilities. | Individuals; Professionals; Researchers | AI search and research assistant that searches the live web, synthesizes sources, performs deep research and increasingly uses Computer to act on tasks. | Assist → Execute — exceptional research automation; Computer adds browser-based task execution and agentic workflows. | Combines answer engine + source transparency with increasingly agentic Computer capabilities. | Persistent memory remembers preferences and past searches; its newer Brain system builds a working model of projects, people, files and corrections. | Good transparency around sources and memory controls; Perplexity says it does not sell personal data, but search data can be used to improve the service unless you opt out. | Web, iOS, Android, desktop/browser via Comet and connected tools. | Free; Pro $20/mo; Max $200/mo; Enterprise Pro $40/seat/mo; Enterprise Max $325/seat/mo. |
| What's uniqueFeels like an AI teammate rather than a chatbot: persistent context + skills + scheduled routines + broad tool access. | Professionals; Teams; Businesses | AI teammate that handles inboxes, meetings, research, reports, business systems and scheduled multi-step workflows across connected tools. | Execute — performs work in connected systems, with approvals for actions that have external impact. | Feels like an AI teammate rather than a chatbot: persistent context + skills + scheduled routines + broad tool access. | Persistent workspace context, editable memory and shared team knowledge; built around remembering how the team works. | Strong published trust posture: encrypted, not sold, not used to train models; approvals are built in and Enterprise adds stronger controls. | Web, Slack, iMessage, connected apps and browser/computer use. | Plus $29.99/user/mo; Pro $99.99; Max $199.99; Enterprise custom. |
| What's uniqueStrong autonomous execution: give it a goal and it can research, build and deploy rather than simply answer. | Individuals; Professionals; Teams | Agent for research, websites, slides, reports, coding, cloud computers and other complex tasks; can run multiple tasks and scheduled work. | Autonomous — plans and executes multi-step work with its own cloud computer and scheduled tasks. | Strong autonomous execution: give it a goal and it can research, build and deploy rather than simply answer. | Project/task context and persistent cloud environments provide continuity; personal-memory depth is less central than task execution. | Public privacy posture is less distinctive than self-hosted tools; Team plans add SSO and data-training controls. | Web app and cloud computer; team workspace. | Free; Pro from $20/mo; higher Pro tiers from $40/mo; Team from $20/seat/mo. |
| What's uniqueOne subscription acts as a multi-model AI production studio with many specialized agents and media tools. | Individuals; Professionals; Teams; Creators | All-in-one AI workspace for chat, research, slides, docs, spreadsheets, images, video, audio and agentic creation. | Execute → Autonomous — Super Agent can coordinate multi-step creation and research across tools and models. | One subscription acts as a multi-model AI production studio with many specialized agents and media tools. | Primarily workspace/project/files context rather than a deep personal-memory relationship. | Team users are automatically opted out of model training; admins control connectors and access, while individual privacy depends on plan/settings. | Web app and connected AI workspace. | Free; Plus $24.99/mo; Team $30/user/mo; Enterprise custom. |
| What's uniqueThe product is explicitly built around owning a persistent AI model of you rather than simply chatting with a general model. | Individuals | Builds a personalized AI around your own knowledge, experiences, personality and communication style; supports personal AI personas and messaging. | Assist → Execute — strong representation and personalization, with less emphasis on autonomous real-world action. | The product is explicitly built around owning a persistent AI model of you rather than simply chatting with a general model. | Deepest memory orientation in this group: a continuously growing personal Memory Stack built from information you provide and connected sources. | Strong privacy proposition: Personal AI says you own your Memory Stack and Personal Model and that your data is not shared or sold. | Web, iOS and connected messaging experiences. | Consumer pricing varies by offering; enterprise pricing is custom. |
| What's uniqueA lightweight personal knowledge + proactive assistant designed to reduce context switching and information overload. | Individuals; Knowledge workers; Creators | Personal knowledge and productivity assistant that organizes notes, email, calendar and tasks, suggests what to work on and connects external AI models. | Execute — can proactively organize information and recommend next actions, but does less autonomous real-world execution than agent-first products. | A lightweight personal knowledge + proactive assistant designed to reduce context switching and information overload. | Builds context from notes, tasks and synced email/calendar; designed to reduce the need to manually maintain a knowledge system. | Publicly emphasizes privacy, export and data security; detailed trust posture is less mature than larger enterprise-oriented products. | Web, Chrome extension and mobile app; connects Google Drive, email, Slack and Calendar. | Free; Starter $8/mo; Standard $16/mo monthly. |
| What's uniqueNarrowly focused on replacing executive-assistant coordination work rather than trying to be a general-purpose chatbot. | Professionals; Founders; Consultants; Small-firm partners | Always-on executive assistant for inbox triage, email drafting, calendar management, task extraction, follow-up tracking and daily briefings. | Execute → Autonomous — works in the background without prompting, but requires approval before sending email. | Narrowly focused on replacing executive-assistant coordination work rather than trying to be a general-purpose chatbot. | Builds a working profile from email/calendar, learned preferences and summaries of past conversations; remembers what matters to your work. | Strong, unusually explicit privacy posture: no selling or model training on email, OAuth instead of passwords, encrypted data and user-controlled deletion. | Web, iPhone, Gmail/Outlook and Google/Microsoft calendars; SMS support. | $29.99/mo; 7-day free trial. |
| What's uniqueA genuinely local, hackable personal agent with broad channel support and full control over models, tools, memory and data. | Developers; Power users; Teams | Open-source personal assistant that runs on your own hardware, connects to many chat channels, uses tools, manages files, sends messages and can run commands and scheduled/background work. | Autonomous — tool use, heartbeats, background work and persistent sessions can run with minimal intervention, depending on configuration. | A genuinely local, hackable personal agent with broad channel support and full control over models, tools, memory and data. | Local, durable memory via workspace files plus cross-channel session context; memory lives on your machine. | Exceptional control: self-hosted, open source, no hosted subscription, no telemetry by default, and your data stays on your infrastructure. | macOS, Windows and Linux; WhatsApp, Telegram, Discord, Slack, Signal, iMessage and many other channels. | Free/open source; no hosted subscription tier. |
| What's uniqueTreats agents as a marketplace: instead of learning another app, you describe the outcome and run a specialist agent built by someone else. | Individuals; Professionals; Teams | Agent marketplace that matches users with expert-built agents for inbox triage, lead follow-up, competitor tracking, content scheduling, errands, watchlists and recurring reports. | Execute → Autonomous — agents execute tasks and automations with variable levels of independence. | Treats agents as a marketplace: instead of learning another app, you describe the outcome and run a specialist agent built by someone else. | Agent-specific execution context; not yet positioned as a single deep personal-memory assistant. | Clear permissions and isolated execution are part of the trust model; builders do not receive identifiable user content unless required by the agent, while Gravity uses anonymized aggregate data for improvement. | Web platform; agents are selected and run through the marketplace. | Free starter usage; subscriptions from about $20/mo plus usage credits. |
| What's uniqueOne of the clearest examples of AI designed primarily as an emotionally intelligent personal companion rather than a productivity or execution agent. | Individuals (Consumer) | Conversational companion for thinking through decisions, difficult conversations, learning, brainstorming, recommendations, reminders and everyday life. | Assist | One of the clearest examples of AI designed primarily as an emotionally intelligent personal companion rather than a productivity or execution agent. | Builds understanding of the user through ongoing conversations and adapts to their preferences over time. | Personal AI positioned around supportive, human-centered interactions; privacy details governed by Inflection's policies. | Web, iOS, Android | Free |
| What's uniqueCombines a mass-market general assistant with real-time X/web information and increasingly autonomous computer-using agents. | Individuals (Consumer & Professional); Teams | General-purpose assistant for conversation, research, writing, reasoning, coding, image/video creation, file analysis and real-time web/X search; Grok Bot can execute work on a persistent computer. | Agentic → Autonomous | Combines a mass-market general assistant with real-time X/web information and increasingly autonomous computer-using agents. | Remembers preferences and past conversations; can connect to email, files and calendar. Grok Bot maintains context while completing delegated work. | xAI offers individual and team plans; connectors and business controls; enterprise features are evolving. | Web, iOS, Android; Grok Bot on supported plans | Free; SuperGrok $30/mo; SuperGrok Plus $100/mo |
| What's uniqueBridges the gap between an AI writing/research assistant and an agent that can actually operate the browser on your behalf. | Individuals (Consumer & Professional) | Research, writing, email management and browser-based execution including booking, ordering, filling forms and other multi-step web tasks. | Agentic | Bridges the gap between an AI writing/research assistant and an agent that can actually operate the browser on your behalf. | Uses context from conversations, connected accounts and browsing tasks to complete workflows. | User-controlled browser/account access; privacy depends on connected services and HyperWrite's policies. | Web; Chrome extension | Free to try; paid plans available |
| What's uniqueTreats AI as a workforce of specialized digital employees rather than as a conventional assistant or chatbot. | Professionals; Teams; Businesses | AI Employees automate HR, IT, finance, recruiting and other enterprise workflows; can build and execute multi-step processes across company systems. | Autonomous | Treats AI as a workforce of specialized digital employees rather than as a conventional assistant or chatbot. | Grounded in enterprise data, connected applications and organizational context; maintains workflow context across tasks. | Enterprise-grade governance, audit trails, RBAC/ABAC, PII protection and on-prem/air-gapped deployment. | Web; Teams, Slack and enterprise integrations | Usage/outcome-based; quote |
| What's uniqueAn AI employee explicitly designed to own an entire business function — outbound sales — rather than assist a human with individual sales tasks. | Professionals; Teams; Businesses | Autonomous B2B sales agent that finds and researches leads, writes personalized outreach, runs campaigns, handles replies, qualifies prospects and books meetings. | Autonomous | An AI employee explicitly designed to own an entire business function — outbound sales — rather than assist a human with individual sales tasks. | Uses CRM, prospect, campaign and intent-signal context to personalize and optimize outbound activity. | Enterprise controls including SOC 2, GDPR, SSO, role-based access, audit logs and escalation rules. | Web; CRM and sales-stack integrations | Usage/credit-based; contact sales |
| What's uniqueExtremely narrow but unusually complete delegation: you CC Clara and she takes the scheduling conversation through to a booked meeting. | Professionals; Teams; Businesses | Handles email-based meeting scheduling, including reading threads, proposing times, following up, negotiating changes and placing confirmed meetings on calendars. | Autonomous | Extremely narrow but unusually complete delegation: you CC Clara and she takes the scheduling conversation through to a booked meeting. | Uses email thread, calendar availability and user-defined scheduling preferences. | No human scheduling reviewer; supports Google and Microsoft environments and enterprise controls. | Email; Google Workspace; Outlook/Exchange | Free trial; $80/mo Standard; Enterprise |
| What's uniqueAn always-on executive assistant that talks to your network for you, rather than simply talking to you. | Executives; Founders; Professionals | Proactive executive assistance across scheduling, email, text, relationships, follow-ups, meeting preparation and reminders; can also make phone calls to handle reservations and appointments. | Autonomous | An always-on executive assistant that talks to your network for you, rather than simply talking to you. | Learns the user's scheduling preferences, relationships, meeting context and communication patterns; uses email, calendar and messaging context to act on the user's behalf. | Designed around delegated access to communications and calendar; external actions can occur automatically according to user preferences. | SMS; Email; WhatsApp; Voice/phone | Free 30-day trial; pricing not publicly established |
About this map
This is a conceptual view of the AI assistant and agent market, not a market share chart and not a ranking.
The AI assistant market is moving beyond chatbots that simply answer questions. Increasingly, AI can remember context, use tools, take actions, coordinate work, and eventually pursue goals with much less human intervention. This map is an attempt to understand that evolution conceptually: what these products can actually do today, who they are built around, and where the capabilities of AI assistants may be heading. It is less about ranking products and more about making the shape of the emerging market visible.
How are we thinking about the market?
One useful way to think about the evolution is as a progression from Traditional Assistant → Agentic Assistant → Autonomous Personal Agent. A traditional assistant primarily helps you think, research, write, recommend, or plan — you remain the driver. An agentic assistant can take actions and complete multi-step tasks when you ask it to. An autonomous agent goes further: it can plan, decide, execute, monitor, and follow up with limited intervention. These aren't rigid categories; they represent increasing levels of agency. At the same time, assistants are emerging along another dimension: some are primarily built around you as an individual, while others are designed around work, teams, and businesses.
Which products made the map?
The selection is intentionally narrower than a list of every product that has AI capabilities. The rule is simple: If you described the product to someone in one sentence, would “AI assistant” or “AI agent” be the noun — or would it be the feature? Products whose central proposition is being an AI assistant or agent made the cut. Products where AI is primarily a feature inside another type of product did not. For example, Granola is fundamentally an AI meeting notepad, Otter is fundamentally a meeting transcription and workflow product, and Notion is fundamentally a workspace that now includes powerful AI and agents. They may have impressive AI capabilities, but AI assistant/agent is not the core product category they are selling. The map is therefore deliberately focused on products where the assistant or agent itself is the product.
Visual map layout
This is a conceptual view of the AI assistant and agent market, not a market share chart and not a ranking.
The vertical dimension represents increasing agency: traditional assistant, to agentic assistant, to autonomous agent. A product placed higher can act more independently — it does not mean it is better.
The horizontal dimension represents whether a product is primarily oriented toward the individual or toward work and business. Most products serve both to some degree; the position reflects where the product is primarily built around.
Positions are editorial judgements, and the three agency bands are a continuum rather than rigid categories.
Last updated: September 15, 2026