Updated May 2026. Reading time: 7 minutes.
An AI fellow is a custom-built AI specialist that lives in your Slack or Microsoft Teams. It is trained for one role, like Sales or Paralegal or Support Engineer, and it serves your whole team. Here is what that actually means, how it differs from a general AI assistant, and how a fellow joins your team.
An AI fellow is a custom-built AI specialist trained to do one job. The Sales fellow does sales. The Paralegal fellow drafts legal work. The Support Engineer fellow handles tickets. Each fellow lives in your Slack or Microsoft Teams, and anyone on your team can ping it.
The word "fellow" is intentional. A fellow is not the same as an assistant. An assistant supports one person. A fellow is a specialist who works with the whole team. Think of an AI fellow the way you would think of a junior hire who joined to do one specific role.
This is different from a general-purpose AI tool. Tools like ChatGPT or Microsoft Copilot are generalists. They will help you with anything if you prompt them. They will not know your sales playbook, your firm's contract templates, or your customer escalation rules. A fellow is built around that knowledge.
Most AI tools today are generalists with a system prompt. They are built to be flexible, to handle any kind of question, to be useful in many contexts. That is great for a solo founder doing every job themselves.
It stops scaling the moment your team grows. The minute you have one person who runs sales, another who runs support, and a third who runs legal, the generalist AI cannot keep up with all three. It does not have the depth in any one role.
An AI fellow is the opposite bet. We pick one role and we build the fellow around it. The Sales fellow gets your ICP, your sales playbook, your CRM workflows, your tone of voice, your competitive landscape. The Paralegal fellow gets your contract templates, your jurisdiction rules, your client intake process. Each fellow goes deep instead of broad.
A fellow lives inside your Slack or Microsoft Teams workspace. There is no new app for your team to learn. Anyone on the team can ping the fellow in a channel or DM.
This matters because the friction kills value. If your team has to switch tools to talk to the AI, they will use it once and forget. If the AI lives in the same channel where they are already working, they will use it every day.
The fellow keeps context across the whole team. What your AE updated in HubSpot this morning is something your founder can ask about this afternoon. The fellow remembers because it is a shared team resource, not a one-person assistant.
It is easy to say "role-specific" and harder to show. Here is what it actually means in practice.
A Sales fellow knows your ICP. When a new lead drops in, the fellow checks the company size, industry, tech stack, and recent news. It scores the lead against the ideal customer profile you defined during setup. It tells you whether to chase or skip.
A Paralegal fellow knows your firm's templates. When a new client requests a service agreement, the fellow does not write a generic agreement from scratch. It pulls the right template from your library, fills in the client-specific details, and flags any clauses that need attorney review.
A Support Engineer fellow knows your product. When a tier-1 ticket comes in, the fellow recognizes the issue type, checks customer history, and either solves it directly or routes it to the right engineer with full context.
In each case, the fellow is built around the way that role works in your business. It is not a chatbot with a "be helpful" prompt. It is a specialist.
Building a fellow takes about a week. The first day is a scoping call. We sit down with you and walk through the role: what does success look like, what tools does this role use, what tone of voice does the team use, what playbook does the team follow.
Days 2 through 7 are custom build. We train the fellow on your context. Sample emails, your CRM setup, your ICP document, your sales playbook for a Sales fellow. Your contract templates, your jurisdiction rules for a Paralegal fellow. Your product docs, your common ticket categories, your escalation paths for a Support Engineer fellow.
By the end of week one, the fellow is in your Slack and starting to work. Days 8 through 14 are shadowing — the fellow watches and you give it small jobs to test the output. Weeks 3 and 4 are delegation, where the fellow takes on more autonomous work. Week 4 you measure the time saved and decide what to expand.
These terms get used loosely. Here is how we use them.
An AI assistant supports one person. ChatGPT is an AI assistant. Microsoft Copilot is an AI assistant. Their job is to help one user be more productive. They are general-purpose by design, because they need to be useful for whatever job the one user happens to be doing.
An AI coworker is a step up. The pitch is one AI that lives in your team's communication tool and does work for everyone. Viktor is an AI coworker. The bet is that breadth wins — one AI that does a little of every role.
An AI fellow is the next level of specialization. Like an AI coworker, it lives in Slack and serves the whole team. Unlike an AI coworker, it is built for one role. The bet is that depth wins. As your team grows, you do not stretch one AI thinner across more jobs. You bring on more fellows, each one specialized.
Tell us what you need. Sales? Paralegal? Support Engineer? Market Researcher? We start by understanding the role you want to fill.
We scope your tools, your playbook, your tone, and your context. We custom-train the fellow on that material. Setup takes about a week.
On day 8 the fellow is in your workspace and starting to work. Your whole team can interact with it. We refine and expand over the first 30 days.
Each fellow is on a custom annual plan based on the scope of the role. We do not charge per credit, per message, or per seat. Your whole team uses the fellow for one predictable annual price.
Annual pricing matters because credit-based pricing creates anxiety. With credits, you do not know what a workflow will cost until it runs. Annual is fixed in your budget. The fellow is yours regardless of how often the team uses it.
It is both, and that is fine. The category of role-specific AI agents is real and growing. We use "fellow" because it captures the specialist + team-member framing. Other companies use "AI agent", "AI coworker", "virtual employee", or "AI assistant". Our framing is the most accurate to what we actually build.
A chatbot answers questions. A fellow does the work end-to-end. The Sales fellow does not just suggest an outreach email — it pulls the lead from your CRM, qualifies it, drafts the email in your tone, and logs the activity back. You do not copy and paste.
Yes. The model is one fellow per role, then add fellows over time. A growing team might start with a Support Engineer fellow, add a Sales fellow six months later, and bring on a Paralegal fellow at year two.
It is closer to a force multiplier. The fellow handles the recurring work that slows humans down, so your humans stay on the work that requires real human judgment. We do not pitch AI fellows as "replace your team". We pitch them as "let your team do the work only humans can do".
The fellow learns continuously. As your playbook evolves, your tone changes, your tools shift, the fellow refines. We do quarterly check-ins to recalibrate based on what the role looks like now versus when the fellow was first built.
You can pause or end the engagement. We do not lock you into long contracts. If a fellow is not earning its keep, the right answer is to scope what would, or to step away. We would rather have happy customers than trapped ones.
Tell us the role. We will scope, build, and ship in about a week.