1. LandOffer
  2. Blog
  3. All articles
  4. Shortwave Review for Work Email: AI Search, Inbox Sharing, and Plan Limits

Productivity & Workflows

Shortwave Review for Work Email: AI Search, Inbox Sharing, and Plan Limits

Five fictional email tasks establish source-grounded expected answers, then a three-person team weighs scoped AI searches, seat costs, and a completed access record without connecting a mailbox.

9 min readLandOffer team

Shortwave is worth considering for a Gmail-based work routine that needs source-grounded email retrieval and deliberate team collaboration. Check message age, the threads each AI request must cover, and the permitted audience before choosing a plan. A larger AI allowance does not establish that an answer is complete or that a teammate should receive the underlying thread.

This LandOffer.ai review examines public documentation and selected pricing views checked October 27, 2026. It includes a completed five-category synthetic retrieval assessment and a separate plan-and-access decision. We did not connect a mailbox, run Shortwave against messages, measure accuracy, share threads, or send email. The examples are editorial work, not application-test results.

An operations professional reviews email search evidence beside an organized reference shelf.

Confirm that your work account fits

Shortwave requires Gmail or Google Workspace for sign-in. Its other-provider guide describes forwarding arrangements for some providers, but lists Microsoft 365 and Exchange as unsupported. Forwarding mail into Gmail is a separate configuration and data-handling decision; it is not native Exchange support or an automatic exception to an employer's rules. Provider guide.

For this review, assume the organization already uses Google Workspace. That assumption removes a provider mismatch without granting permission to connect a third-party client. The mailbox owner and administrator still need to decide whether this client and its processing fit the intended data. Do not use a personal forwarding account to get around a work restriction.

Count people separately from the email addresses each person uses. The billing guide describes one paid feature entitlement across your own accounts signed in on the same device, while a team adds billing for other people. It requires all people on a team to use the same plan type. Two addresses belonging to one person should not automatically become two purchased team seats. Billing guide.

Read history, context, and usage as different limits

The public pricing table separates standard search from AI search. Free lists 90 days for both; Pro lists full standard history, three years of AI history, and 50 threads per AI search. Business lists five years and 50 threads; Premier unlimited AI history and 100 threads; Max unlimited AI history and 150 threads. These limits describe different constraints, not a single measure of search quality. Pricing.

The import guide says a new account initially imports 90 days and describes a separate full-history import for paid users. Buying a plan is therefore not evidence that the old source message is already present and ready to inspect. Check import completion and the source itself before interpreting an AI answer about an older discussion. History import.

AI usage is another boundary. The quota guide says consumption depends on intelligence level and conversation length; Basic does not count toward the quota, while other levels have different costs. It states a 24-hour reset and no mid-day top-up. A count of searches cannot be converted into a guaranteed day's capacity without knowing their actual context and selected level. Quota guide.

Check for the amendment that changes the answer; unlimited history still has a per-search context limit. Separate the questions: Is the source within the permitted history? Has it imported? Is it accessible to this person? Was it included in this request? Does the answer preserve its meaning? Paying for more context addresses only part of that chain.

Completed asset: five email tasks with known answers

Fictional operations coordinator Niko manages facilities requests for a small design company. The following invented source inventory is dated as of October 27, 2026. It contains no real email addresses or company data. Its expected answers are manually established from the text; every Shortwave retrieval result remains untested.

Category and source IDs Invented source text Required answer and supporting source Failure to catch
Old procedure, E01 December 12, 2022: “After-hours access requests need the facilities coordinator's written confirmation.” Written confirmation is required; cite E01, and check for a later replacement Claiming an old message is a current universal policy
Revised arrangement, E02 and E03 August 5, 2025: “Courier pickup at 16:00.” October 20, 2026: “For the November pilot only, pickup moves to 15:30.” Pilot pickup is 15:30; E03 is limited to November, while E02 supplies the older baseline Combining the messages into a permanent 15:30 schedule
Attachment, E04 October 21, 2026 email: “See loading-bay-map.pdf for Door C.” Invented attachment text: “Door C; visitor entrance remains Door A.” Loading bay is Door C, not visitor reception; inspect the attachment itself Answering from the filename or confusing two entrances
Conditional handoff, E05 and E06 October 22: “Can Lea review after the room list is checked?” October 23: Lea replies, “Yes, once you confirm the list; no date agreed yet.” Lea's review is conditional; list confirmation and date remain open Inventing a Friday deadline or completed room check
Restricted record, E07 October 24: private manager discussion about staffing, outside the facilities collaboration audience Exclude from the team retrieval exercise; do not summarize or share its contents Treating the ability to find a message as permission to disclose it

This inventory has seven messages across five task categories. Six messages belong in the proposed facilities exercise; the seventh is a deliberate access-boundary control. The proposed facilities test must exclude E07 from both search input and any shared answer. Reject an extra result when its disclosure exceeds the permitted purpose.

E01 is more than three but less than five years old on the review date. That makes Business the relevant documented AI-history candidate for this particular archive question; Pro's full standard search remains a different route. Neither statement says Shortwave actually found E01, that the import succeeded, or that a newer replacement does not exist.

For E02–E03, Niko's completed answer key reads: “Use 15:30 only for the November pilot. The earlier arrangement was 16:00. Ask the facilities owner which arrangement applies after the pilot.” It is a useful result precisely because it preserves the unresolved future period instead of choosing one time for every situation.

The attachment case requires more than seeing a PDF in a result list. The reviewer must open the returned source and compare its relevant text with the answer. No extraction test was performed here, so this article gives no PDF parsing success rate. Likewise, the conditional handoff is a meaning check: an answer can quote Lea's words correctly and still turn the condition into an unconditional commitment.

Niko's proposed search starts narrowly, with the relevant sender, subject, or date range. Shortwave documents operators including from:, subject:, filename:, and date filters, plus about: for AI topic search. These are available ways to specify scope; the example does not claim that one query will return an exhaustive archive. Search reference.

The finished assessment disposition is therefore: the source-answer checklist is ready, the restricted control is excluded, and product retrieval remains pending an authorized test. Success would require the expected sources, correct qualification, and no inappropriate disclosure. An attractive summary without its supporting message does not meet that requirement.

Separate labeled drawers for private and shared mail illustrate an explicit access boundary.

Completed asset: decide seats and sharing separately

Now add colleagues Lea and Tom. All three would join the same proposed Shortwave team. Their allowed facilities archive contains an invented 84-thread coordination collection from the last eighteen months, separate from the seven-message answer-key sample. They want to review that collection together, and they occasionally need the older procedure in E01. They have not connected accounts or bought seats.

Niko can partition the collection into two disjoint sets of 42 threads by a manually defined project boundary. Each set is numerically within Business's 50-thread AI-search limit; all 84 fit numerically within Premier's 100. That arithmetic does not establish that Shortwave will select precisely those threads or that a combined answer will cover every issue. Cross-set conclusions would need a separate reconciliation.

Their decision record compares two acceptable methods. Business is a candidate if the team accepts scoped searches and a manual combined review. Premier is a candidate if the workflow requires considering the whole collection in one AI context and testing supports that use. Max's larger limit does not answer an additional requirement in this case. No plan is selected merely to make a higher number appear safer.

Our native-browser readback of the selected Individual and Business pricing views showed Pro at $18 per seat per month, and Business, Premier, and Max at $30, $45, and $120. The page displayed dollar symbols; checkout currency and taxes were not verified. The billing guide says new subscriptions are monthly. One Pro seat is $216 over twelve unchanged monthly payments; three Business seats are $90 monthly or $1,080 over twelve months. Three Premier seats are $135 monthly or $1,620 over twelve months, an additional $45 monthly or $540 annually. These are base-price illustrations, not an annual discount or a quoted invoice.

Before deciding between those methods, they complete the following audience record. Every “permitted” entry is a fictional internal decision, not a change made in Shortwave.

Material Intended audience Disposition Boundary still requiring review
E02–E03 courier thread Niko, Lea, Tom Suitable for a proposed facilities-only shared thread after authorization Inspect historic attachments and future replies, not just the newest sentence
E04 map Same three Approved in the fictional plan for this work purpose Verify the actual returned file and audience before sharing
E05–E06 conditional review Same three Share the condition without inventing a due date Niko must still confirm the room list; no review date has been accepted
E07 staffing discussion Manager and original authorized participants only Withheld from the facilities exercise No shared label, comment mention, or copied summary
New supplier replies Audience depends on content Review before adding to the shared archive An old sharing decision may be too broad for later attachments

This last row matters because Shortwave's team guide describes live thread sharing: past messages, attachments, and future replies are included. It also says an @mention shares the entire thread, shared labels make labeled messages discoverable to teammates on that label, and a thread link works for people already given access. A private-looking comment or copied link is therefore not a substitute for an audience decision. Team collaboration guide.

Their completed recommendation is a three-person Business evaluation using the two scoped collections, with Premier reconsidered only if the required combined review cannot be completed acceptably that way. The facilities owner approves the proposed purpose in this fiction, while actual application authorization remains pending. E07 stays out regardless of the tier. They do not price one Premier member and two Business members within the same team when the billing guide requires a common plan.

Keep retrieval assistance separate from actions

The source-answer exercise requires read-only retrieval and review. It does not require automatic sending, deleting, forwarding, or creating events. Shortwave's AI guide describes capabilities that reach beyond search, including drafting and connected tools, and separately identifies Tasklet as a sister automation product. Evaluate those capabilities as additional workflows, not assumed benefits already included in a successful search exercise. AI Assistant guide.

There is also a practical reason to keep the test narrow. Shortwave says actions such as archiving, labeling, and deleting sync back to Gmail. A cleaner-looking client can therefore represent real mailbox changes. The organization must assess the processing and access involved before connecting a work mailbox; this review neither performed that assessment nor approved a particular data category. Security and privacy guide.

Use the documented limits to choose an evaluation, then record actual source retrieval, answer fidelity, and audience checks before accepting the workflow. Keep missing sources, incomplete import, unresolved conditions, and excluded audiences visible. For more work-tool decisions, return to LandOffer.ai with a concrete question and an answer key you can verify, rather than a feature list you hope will produce the right result.

Sources and Further Reading