A year ago, “AI agent” was a lab topic. Today there are agents that answer your texts for you, agents that get their own cloud computer to fill in forms, agents inside your CRM, ERP and inbox, and specialist agents for contracts, medical records and bank reconciliation.
To get a sense of the scale, we built a map of more than 120 agents and platforms across 17 categories. The conclusion fits in one line: there are a lot of agents, and the number will only go up. Here is what we found and what it means if you want AI doing real work inside your operation.
Personal agents: one for every way of living
These are always-on agents with memory that work for one person. They alone account for almost 60 products on our map:
- From the AI labs. The companies behind the big models (OpenAI, Anthropic, Meta, xAI) shipping their own everyday agents.
- Open source. Agents you run on your own machine and connect to any model you like. OpenClaw, the best known, passed 370,000 GitHub stars and became the most-starred project on the platform in 2026.
- Lightweight and hardened versions. Dozens of slimmer or more secure spin-offs of the open agents, written in Rust, Go, Zig and C. A new one shows up almost every week.
- Text-first. You send a message on iMessage or WhatsApp and the agent takes care of it, with no new app to learn.
- Agents with their own computer. Agents that get a cloud machine to browse, fill in forms and use software the way a person would.
- By niche. Family and household, travel, creators, work life. Every niche gets its own.
ChatGPT dots
Grok Bot
Muse
Claude Duet
OpenClaw
Hermes Agent
Vellum
Comma
Pivra Desk
ZeroClaw
NanoClaw
nanobot
IronClaw
Moltis
PicoClaw
Instinct
Poke
Tab
Persona
folk
Hello Haven
Rene
Martin
Tomo
Pally
Iris
Wabi
Perplexity Computer
Cue
Genspark Claw
Lindy
Fo
szn
Eney
Hark Pro
Underdog
Skye
Ambi
Fambot
Ohai
Ollie
Orbits
Miso
Stanley
Town
What stands out most is the pace. Comparing two editions of the same landscape, one week apart, products came in, products dropped out and one lab swapped the agent it was leading with. The map you read today is out of date by Friday.
In the enterprise, the map is even bigger
On the business side, agents show up in several layers at once:
- Inside the software you already run. Microsoft Copilot Studio, Salesforce Agentforce, Google Gemini Enterprise, ServiceNow, SAP Joule, HubSpot. The agent arrives with the license you already pay for.
- Automation with an agent in the loop. UiPath, Automation Anywhere, Power Automate, Workato, Zapier and n8n now put an agent in charge of the decisions that used to stall a workflow.
- Infrastructure and frameworks. The clouds (AWS, Google, Azure, Cloudflare) and the toolkits for building from scratch (LangGraph, CrewAI, the OpenAI and Anthropic SDKs).
- Connections and oversight. Connectors, open protocols like MCP and tools that measure and monitor what agents do.
- Industry specialists. Legal, healthcare, finance, real estate, security. These are only examples: every industry with repetitive work is getting its own.
Microsoft Copilot Studio
Salesforce Agentforce
Gemini Enterprise
SAP Joule
IBM watsonx
HubSpot Breeze
UiPath
Zapier
n8n
AWS Bedrock
Google Vertex AI
Azure AI Foundry
Cloudflare Agents
LangGraph
CrewAI
LlamaIndex
Dify
Composio
Browserbase
MCP
Langfuse
Braintrust
Intercom Fin
Zendesk
ElevenLabs
Claude Code
Codex
GitHub Copilot
Cursor
Devin
Replit
The numbers back it up
Research firms never agree to the dollar, but they all point the same way:
- From under 5% to 40%. That is the share of enterprise apps with task-specific agents Gartner expects between 2025 and the end of 2026.
- $1 trillion by 2030. Gartner’s forecast for spending on agents embedded in enterprise software, eleven times the 2025 figure.
- More than 400,000 custom agents were already running in production on Microsoft Copilot Studio by May 2026, across 160,000 organizations.
- 29,000 deals for Salesforce Agentforce, with around $800 million in annual recurring revenue.
- $24 billion of venture capital went into agentic AI in 2025 alone.
- Only 13% of companies worldwide have built agents into their workflows, according to BCG. Even in leading markets like Brazil it is 18%. More than 80% of companies have yet to start.
Why it will keep growing
Four forces keep pushing the number of agents up, and none of them shows signs of slowing:
- Cost. Running a model gets cheaper every quarter. When an agent costs cents per task, it pays to automate work that never justified the effort before.
- Open source. Any team can take an open agent, adapt it and launch a new product. That is how dozens of variants appeared in a few months.
- Connections. Open standards let an agent talk to your CRM, ERP, spreadsheets and messaging without a hand-built integration for each system.
- Distribution. The big software vendors are building agents into their products. Plenty of companies will have agents running before they ever decide to buy one.
Finding an agent is easy. Getting it to run is the work.
With this many options, the question is no longer “is there an agent for this?” There almost always is. The question is which one to pick, how to connect it to the systems you already run and who looks after it once it is live.
That is exactly where projects stall. Gartner itself expects more than 40% of agentic AI projects to be canceled by the end of 2027 over high costs, unclear value or weak risk controls. The problem is rarely the technology. It is picking the wrong task, not connecting the agent to company data, or leaving it without an owner after launch.
If a task repeats, it can be automated. Start with the task, not the agent.
A simple way to avoid becoming a statistic:
- List what repeats. Finance, support, sales, purchasing and HR are just examples. Look at what your team does the same way every week.
- Choose without getting locked in. The map changes every week. Build the agent so you can swap the model or the tool later.
- Connect it to your systems. An agent that can’t read your CRM, ERP or inbox is just one more open tab.
- Look after it once it’s live. Measuring, fixing and improving every month is what separates the project that sticks from the one that gets canceled.
How BossaMotion helps
We don’t sell an off-the-shelf agent. We work inside your operation to pick, build and run the right agents:
- The AI Analyst maps your routine, finds what repeats and ranks it by return.
- The Forward Deployed Engineer builds the agent, connects it to your systems and puts it to work, with human sign-off where it matters.
- The team then reviews results every month and swaps components when something better shows up on the map.
Not sure where to start? Read the 4 levels of AI adoption and see where your company stands.
There will be no shortage of agents. What makes the difference is putting the right one to work inside your operation.