For operators, founders, and team leads evaluating FellowHire.
From scoping call to live in your Slack in about a week. Here is the full picture, step by step.
Every fellow starts with a 30-minute scoping call. We are not pitching. We are asking enough to know if a fellow will work and what tier the role belongs in. Five questions cover most of it.
What role do you need? Sales fellow? Paralegal? Support engineer? Marketing? Each fellow is custom-built for one role. We map your need to a tier (Standard or Senior) based on the work, not your company size.
What does your stack look like? Slack or Microsoft Teams. CRM, ticket system, doc store, calendar, billing. We list the integrations the fellow will need on day one.
What does the work look like? What does a typical week ask of this role? What happens when things go sideways? What does "good" look like?
What is your voice? Style guide if you have one. Sample emails or docs if you do not. The fellow learns to write the way your team writes.
Who decides? Who on your side scopes the role, signs off on training data, reviews early drafts, and decides on production rollout?
By the end of the call, you have a tier, a price, and a rough integration list. You do not have a contract yet. That comes after the call, once the scope is clear on both sides.
Once the contract is signed, we build the fellow. This takes about a week for most roles. Here is what happens.
Training data ingest. You hand off the materials your fellow will learn from. Style guide, sample work, product docs, customer interviews if you have them, the playbook your senior person uses. We feed all of it into the fellow's training context.
Voice calibration. The fellow drafts a handful of sample tasks before we hand it over to you. We compare the drafts against your real work. We tighten the voice until the drafts read like you wrote them.
Integration setup. Each integration gets connected via OAuth or API key. We scope the read and write permissions you approve. Permissions are role-based, not workspace-wide. You can revoke any integration at any time.
Compliance review. For regulated industries (legal, accounting, healthcare), we walk through the compliance posture during setup so your risk team has what they need. Full security detail at /security.
First-week supervised tasks. The fellow ships its first work as drafts, not finals. You review, give feedback, and the fellow tightens. This is the tightest feedback loop you will have. Use it.
The fellow joins your workspace like a new hire. Same channels, same DMs, same naming pattern (Casey, Quinn, Patch, depending on the role). You can @mention the fellow. The fellow can post in channels it has been invited to.
Here is what to do on day one.
Introduce the fellow in #general. Tell the team what the fellow is for, which channels it is in, and how to ask it for things. Most teams treat this like onboarding any new colleague: a short post, a link to the role docs, and a "say hi" thread.
Run a few real tasks together. Not synthetic ones. Real work. Pick three tasks the fellow should be able to do and run them in a thread. Watch the output. Ask follow-ups. Get a feel for how the fellow responds.
Set the feedback channel. Where does the team report bugs, miscalibrations, or "the fellow did this weird"? A dedicated thread or channel keeps the feedback flowing into one place. We watch it during the calibration week.
Week one is feedback-heavy. By week four, the fellow should run on its own with light supervision. Here is what changes.
Voice tightens. Early drafts will sound 80% like you. Calibration tightens to 95%+. The fellow learns the phrases you actually use, the structure you prefer, the things you never say.
Edge cases get handled. The first time the fellow hits an edge case (an unusual matter type, a non-standard intake, an integration quirk), you tell it how to handle it. Next time, it handles it on its own.
Permissions sharpen. Some integrations are over-scoped on day one. Some are under-scoped. Calibration tunes the access scopes so the fellow can do its job without overreach.
The team learns to ask better. Your team gets faster at asking the fellow for what they actually need. Bad prompts get fewer. Specific asks get more common.
By the end of week four, you should be running production work through the fellow with light review, not deep review.
Production is where the fellow earns its keep. Here is what a typical week looks like, depending on the role.
Sales Fellow (Standard)
Casey
Watches the inbox and CRM. Qualifies new leads as they come in. Drafts outreach for the rep to review. Updates the pipeline after each call. Posts a Friday recap with deal movement and next steps.
Paralegal Fellow (Standard)
Quinn
Reads new client intake forms. Drafts matter summaries. Pulls hearing prep packets. Drafts client correspondence in your firm's tone for attorney review. Tracks deadlines and pings the channel when one is close.
Developer Fellow (Senior)
Kai
Reads Linear and Jira tickets. Drafts implementation plans. Reviews PRs with concrete feedback. Pairs on tricky bugs. Pulls context from your codebase and design docs.
Marketing Fellow (Standard)
Morgan
Drafts blog posts in your voice. Runs competitor research. Writes distribution copy for every launch. Ships weekly performance reports.
Every fellow lives in your Slack or Teams. Every fellow respects channel scope. Every fellow can be revoked at any time.
Browse all 15 fellows →Plain English on the technical guts. No jargon, no hand-waving.
Model providers. FellowHire uses OpenAI and Anthropic as the primary model providers. Standard fellows run on mid-tier models (Claude Haiku, GPT-4o-mini, Gemini Flash class). Senior fellows run on frontier models (Claude Sonnet or Opus, GPT-4o class). The model is matched to the work.
Isolated AWS Enclave. Every customer runs in their own isolated AWS environment. There is no shared compute. There is no shared storage. Your fellow does not have access to anything outside your environment.
No training on your data. This is the line we do not cross. Customer prompts, customer data, and customer outputs are never used to train models. We have enterprise no-training agreements with both OpenAI and Anthropic. This is contractually enforced.
Encryption. TLS 1.3 in transit. AES-256 at rest. Standard for serious work; non-negotiable for us.
Audit logs. Every fellow action and every integration call is logged. Logs are retained per SOC 2 requirements and available to you on request.
FellowHire is SOC 2 and ISO 27001 compliant. Encrypted in transit (TLS 1.3) and at rest (AES-256). No-training enterprise agreements with OpenAI and Anthropic.
Most teams start with one fellow. They add a second once the first one is humming. Here is what changes.
Per-fellow engagement. Each fellow is its own engagement, scoped and priced separately. Adding a Paralegal fellow on top of a Sales fellow is two contracts. We align them to a single annual renewal date so you do not have scattered review cycles.
Coordination across fellows. Fellows can hand off to each other when it makes sense. A Sales fellow can flag a deal that needs a Customer Success fellow follow-up. A Paralegal fellow can flag a contract that needs an Operations fellow review. Coordination happens through your Slack or Teams, the way humans coordinate.
Pricing scales linearly. A Standard fellow is $18,000 per year. Two Standard fellows is $36,000 per year. There are no team-of-fellows discounts at the moment. We are not running a SaaS gym membership. Each fellow is custom-built and priced accordingly.
Most teams add the second fellow in months three or four. Once the first fellow is calibrated and production-stable, you have the bandwidth and the trust to scope the next one. Trying to onboard two fellows at once usually splits attention and slows both down.
The honest version: a fellow is not a magic box. It needs four things from you to be great.
Feedback. Especially in the first month. Every "this draft is off" is a calibration signal. Teams that give heavy feedback in week one have fellows that run cleanly by week four. Teams that give light feedback have fellows that stay 80% calibrated forever.
Scope clarity. If the role drifts ("can you also do this random thing?"), the fellow's quality drops. Specialists beat generalists. Keep the role focused.
Voice samples. The more sample work you give, the closer the fellow gets to your real voice. A style guide is good. A folder of past blog posts plus the style guide is great.
Decision authority. Someone on your side has to be empowered to say "yes, ship that" or "no, redo it". Without a decider, the fellow stalls in review.
If you can supply those four, the fellow will return its annual cost in saved hours within the first quarter. If you cannot, hold off until you can.
AI fellow
Off-the-shelf chatbot
Hiring a contractor
Each option has its place. A chatbot is fine for one-off questions. A contractor is fine for a specific project with a clear end. A fellow is the right call when you have ongoing role-shaped work that a generic tool cannot handle and a contractor cannot stick around for.
About one week from contract signature to live fellow in your Slack or Teams. Faster for simpler scopes (one role, few integrations). Slower for multi-fellow rollouts.
Yes, within your isolated environment. The fellow refines on the feedback you give and the work it does inside your workspace. Your data never leaves your environment for model training.
Most customers do not. We work with sample emails, blog posts, customer-facing docs, anything that shows the fellow how your team writes. A style guide is nice but not required.
Yes. The fellow lives in your shared workspace. Anyone with channel access can ping the fellow. The fellow respects channel scope and your team's existing permissions.
You give feedback, the fellow learns, the next attempt is better. For high-stakes work (legal documents, financial reports, customer-facing copy), our recommended workflow is human review before send. The fellow drafts, you ship.
Annual contracts run for a year. At renewal, you renew or you do not. Customer data is deleted within 30 days on request or per the timeline in your DPA. Audit logs are retained per SOC 2 and then deleted.
Book a 30-minute walkthrough. We will show you a fellow doing the job you would hire it for. No slideware.