AI at Work
Fellow Review: Can One Workspace Handle Agendas, Action Items, and AI Meeting Notes?
Evaluate Fellow for recurring agendas, reviewed action items and AI meeting notes. A complete four-week one-on-one and corrected fictional recap show what people must verify before buying.
Fellow is a credible candidate when a team needs shared agendas, recurring meeting history and assigned follow-up alongside AI recaps. Its strongest documented fit is the connection between preparation and later accountability. Evaluate the paid plan against the history access, task controls, recording volume and permissions your team requires. A generated recap still needs people to confirm what was agreed, who accepted the work and which decisions remain open.
This LandOffer.ai review examines current public Fellow documentation on October 27, 2026. The four-week one-on-one and deliberately flawed recap below are original fictional examples. They are not recorded conversations, Fellow-generated output or an accuracy benchmark. No account, meeting, integration, sharing permission or subscription was changed.

What one workspace actually brings together
Fellow's Meetings documentation, dated March 9, describes calendar events linked to collaborative notes, templates, assigned discussion points and attached AI recaps. This gives preparation, conversation records and follow-up a common meeting context. It does not mean every item in a summary is approved work or that everyone in a company should see every conversation.
The useful distinction is between three kinds of note content. A talking point is something to discuss; an action item is follow-up work; an ordinary bullet preserves context. Fellow's content-type guide says talking points have no due dates, while action-item dates require a paid plan. A topic can be discussed without its associated work being finished. Checking off discussion should not falsely close the task.
The recording layer adds a transcript, summary and suggested tasks. Evaluate it separately from collaborative note-taking. A team can prepare and confirm a useful agenda manually even when recording is unsuitable. Conversely, paying for plentiful recordings does not resolve an unclear agenda or someone else's unaccepted deadline. Fellow brings the records together. People still need to distinguish a discussion, a proposed task and accepted work.
Price the ongoing workflow, not just the first recordings
The current pricing page, independently checked in its team view, displays these USD rates. Annual billing is a commitment; taxes and checkout were not tested.
| Team plan | Annual-billed monthly equivalent | Monthly billing | Relevant AI allowance |
|---|---|---|---|
| Free | $0 | $0 | Five AI notes and recordings per user, lifetime |
| Team | $7 per user | $11 per user | Ten AI notes per user per month |
| Business | $15 per user | $23 per user | Unlimited AI notes and recordings |
For six enabled paid users, Team's base annual calculation is 6 × $7 × 12 = $504, versus 6 × $11 = $66 for one month. Business is $1,080 annually or $138 for one month. These are alternative billing scenarios, not charges incurred. Confirm billable seats, eligibility, any minimum and the final quote before purchasing. Business's broader organizational controls are a separate requirement to evaluate.
The current price comparison lists maximum seats as Free five, Team twenty, Business fifty and Enterprise unlimited. February 18's billing-model guide instead says Free ten; treat that older help limit as conflicting, not a ten-seat entitlement. The guide also describes pooled Team credits and paid due dates, version history and exports. Confirm how recordings consume credits before multiplying by attendees.
Free is a limited ongoing plan, not an unlimited AI subscription. The dedicated Free-plan article describes five lifetime recording credits per user and inaccessible older notes, preserved rather than deleted. It says fourteen days of accessible history; the billing-model article says fifteen. That discrepancy remains unresolved. Four weekly meetings span twenty-one days from the first to the last. Both stated windows are shorter, so the example must not assume its first note stays accessible on Free.
The Meetings page also broadly calls recording a paid feature, while current pricing and Free help explicitly describe limited free AI allowances. Read those together as a free allowance versus sustained paid use, then verify the offered account entitlement. Do not turn the broad paid-only sentence into a claim that no free recordings exist, or confuse temporary trial access with permanent features.
Worked asset: one continuous four-week one-on-one
Marisol, a technical lead, and Benoit, an engineer, need a recurring conversation about a search-relevance evaluation protocol. Their fictional workspace has an approved paid plan with due dates and full history. Recording is merely an evaluation option subject to permission, not a condition of their working agreement. This entire sequence is authored for the article; none of the meetings occurred.
They use a consistent thirty-minute agenda: five minutes for the previous commitments, ten for Benoit's current obstacle, ten for a decision or coaching question and five for owner-and-date readback. Preparation is short: Benoit adds the example or blocker; Marisol adds the decision she can make. A related stakeholder's unanswered question stays visible without making that person an involuntary task owner.
| Meeting | Prepared question and material | Completed fictional outcome and accepted next action |
|---|---|---|
| Week 1, November 3 | Can two representative search queries expose the protocol's unclear judgment rules? Benoit supplies two query descriptions. | They choose a two-example draft, not a production evaluation. Benoit accepts preparing it by November 6. Marisol accepts asking reviewer Romy for clarification by November 5; Romy's response date is unknown. |
| Week 2, November 10 | The two examples are drafted, but an uncertain judgment could be mistaken for a definite negative result. | Benoit's drafting action closes after the examples are inspected. Marisol's clarification request was sent within the fiction, but the answer remains pending. The pilot stays on hold; no one commits Romy to Friday. |
| Week 3, November 17 | Romy's supplied clarification separates an uncertain judgment from a negative label. Can the examples express that distinction? | Marisol and Benoit accept the distinction for the draft protocol. Benoit accepts revising both examples by November 20. This is an internal review step, not launch approval. |
| Week 4, November 24 | Both revisions are available. Do they contain enough context to apply the clarification? | One example is accepted for the internal protocol; the other still lacks locale context. Benoit accepts adding that context by November 26. Broader protocol approval and pilot scheduling remain open. |
The recurring workspace is useful here because each conversation needs the preceding commitment and its actual status, rather than a fresh list with no history. The Week 2 agenda should show that writing was completed while clarification was not. Week 3 uses the received answer; it does not keep carrying an obsolete question. Week 4 preserves a mixed result rather than reducing two different review statuses to one completed checkbox.
The dates are fictional commitments, not product reminders demonstrated here. Each accepted task has a person, bounded deliverable and deadline. Romy's contribution is requested through Marisol; the table never promises Romy's time. That distinction prevents a tidy assignment list from silently exceeding the people present at the meeting.
Fellow's carry-forward guide, dated September 11, documents always, never and ask-first options separately for discussion points and actions. It also describes linked copies whose completion can update past notes. Choose deliberately: a repeated task is not a fresh commitment each week, and a changed checkbox is not an immutable historical approval record. If an owner agrees to a revised deadline, preserve the original date, new date and reason in the note; carrying the item forward alone does not explain the change.
Worked verification: an attractive recap is not an agreement
Use Week 2 to test the review procedure before evaluating any AI model. The following source fragment and simulated recap are written by the author. The errors are planted teaching examples; Fellow did not produce or miss these statements.
Authored source fragment: Benoit says, “Both query examples are drafted. I don't think we should start the pilot until the uncertain-versus-negative rule is clear.” Marisol replies, “Agreed. I asked Romy last week and will follow up tomorrow. I cannot promise when she will answer.” Benoit adds, “After that answer, we can decide what revision is needed.”
Deliberately flawed simulated recap: “Benoit will complete both examples by Friday. Romy will approve the protocol Friday. The pilot is ready. Marisol will follow up.”
The simulated recap reopens completed drafting work, assigns an absent reviewer a deadline, turns an explicit hold into readiness and omits the accepted follow-up date. Comparing it with the source produces this completed verification record:
| Proposed statement | Comparison with the authored source | Corrected human-reviewed record |
|---|---|---|
| Benoit completes examples Friday | Drafting was already finished; no new Friday commitment appears. | Close the existing draft task after inspection. Do not create a duplicate Friday task. |
| Romy approves Friday | Marisol explicitly cannot promise Romy's answer date. | Remove the invented commitment. Keep clarification pending with Marisol coordinating the request. |
| Pilot ready | Both speakers agree to wait for the rule. | Record pilot on hold pending clarification and later review. No rollout approval. |
| Marisol follows up | Accepted action, but the relative date was omitted. | Marisol confirms follow-up on November 11. The response itself remains unpromised. |
The final fictional recap therefore reads: “The two example drafts are complete. The pilot remains on hold because the judgment rule is unresolved. Marisol accepts following up with Romy on November 11. No reply deadline or protocol approval is recorded. Decide revisions after clarification.” This is the verified editorial record, not a claim that the underlying tasks were executed in Fellow.

Accepting an AI task is a separate step
Fellow's current AI-suggested action-items documentation makes an important state distinction. Unaccepted suggestions do not enter the inbox, synchronize to project tools or carry forward. Accepted items do. Suggestions remain in the action view for thirty days; regular items remain until completed or deleted. Editing wording or a date does not itself accept a suggestion; marking it complete does.
Auto-accept is documented as off by default and applies to the individual user's suggestions. When enabled, suggestions are accepted automatically and appear in the meeting note rather than the recap. Check this setting when testing review procedures. The example needs an explicit review gate, so an unnoticed automatic acceptance would defeat the proposed control.
The reviewer should correct the suggested work, confirm the owner's agreement and then accept only the valid item through the authorized workflow. In Week 2, only Marisol's dated follow-up is a new accepted task. Benoit's completed draft is not reopened, Romy's invented approval is removed and pilot readiness remains a decision status rather than a to-do. A checkbox cannot replace this comparison.
Access and export are part of the purchase decision
Fellow's visibility guidance describes attendee-based default note access and says workspace administrators do not automatically gain extra note visibility. Private notes are a separate personal area. External attendees do not automatically have internal note access; explicit sharing mechanisms apply. Check the actual invitees, shared links and destinations before treating a one-on-one series as a team archive.
For Marisol and Benoit, the fictional disposition is to keep their series restricted to its approved participants and prepare a separate minimal protocol update for others. Private reflections are not included. No real sharing or recipient-view check was performed, so the article cannot certify those permissions. A broader audience also needs an appropriate record, not merely access to the whole conversation.
The Word-export guide, dated September 11, describes exporting one meeting or all meetings in a series, optionally including AI notes, with assignee and due-date content. That is useful for evaluating record portability. It does not prove a live task connection, complete audit-history preservation or successful export of this example. Test the required output on approved nonsensitive material before relying on it.
Verdict: choose for continuity, verify accountability
Fellow is worth evaluating when shared preparation, recurring context and reviewed action items must live alongside AI notes. Team is a plausible starting point for the example's history and task controls; Business deserves consideration for greater recording volume or organizational needs. Free can introduce the workflow, but its finite AI allowance and short history make it a poor assumption for a durable four-week record.
If your existing approved document and task system already handles preparation and ownership, Fellow's additional workspace needs a clear reason to exist. This review establishes documented capabilities and a finished evaluation packet, not ease of use, capture accuracy, time savings or adoption results. Keep those questions open until an authorized test measures them.
Use the completed four-week packet and correction record to define that evaluation: preserve agreed work, distinguish suggestions from accepted tasks, confirm the intended audience and inspect the required export. Before authorizing payment, confirm the exact quote, required history access, task-review settings and approved sharing route. A documented feature list does not establish that the configured workspace meets them. LandOffer.ai publishes this workplace-tool review; it does not provide Fellow's meeting service.
Sources and Further Reading
- Fellow pricing — Current public rates and allowances, checked October 27, 2026.
- The Meetings Section — March 9, 2026.
- Action Items v. Talking Points v. Bullets — February 25, 2026.
- What is Fellow's pricing model? — February 18, 2026.
- Our Free Plan — February 17, 2026; short-history discrepancy retained.
- Carry Incomplete Items Forward — September 11, 2026.
- AI-Suggested Action Items — Current body checked October 27; relative update label only.
- Who can see my meeting notes? — March 9, 2026.
- Export Notes to Word Docs — September 11, 2026.


