AI for marketing, sales and support: put role plugins to work

“Can your product send the weekly report to each store automatically?” Marketing sees a landing-page message. Sales sees a demo opportunity. Support sees a possible misunderstanding: exporting a report and sending it automatically are different capabilities.

Role plugins give recurring work a reusable set of inputs, steps and output expectations. We will follow a fictional reporting company, Fieldnote, through a campaign, a sales meeting and a support issue. The sample outputs are teaching references, not customer results or a promise of identical model responses.

The community selections adapt Anthropic’s Knowledge Work Plugins. Onevium maintains the packaging independently and includes the original attribution, Apache-2.0 license, pinned source revision and modification record. These are curated selections, not complete copies of the upstream plugins.

The packages install instructions without business-system connectors. Begin with supplied files and a working model connection. The plugin license does not include model usage or third-party subscriptions.

Fieldnote has one week to invite store managers to a demo. Supply this brief:

Audience: teams operating 5–20 stores.
Available: manual CSV export; weekly sales views by store.
Unavailable: scheduled emails; automatic messages to managers.
Campaign: one week, using the existing website and opted-in audience.
Budget: existing channels only; no advertising purchase.
Goal: relevant demo bookings; the team has not agreed a numerical target.
Voice: concrete and calm; no invented productivity claims.

Ask campaign-plan for the schedule, the purpose of each piece and missing decisions. Then use brand-review on a draft. “Reports arrive automatically in every manager’s inbox” should be flagged against the supplied product facts. A defensible alternative is “Review weekly sales by store and export a CSV to share.”

Check whether the plan invented past open rates or an approved booking target. Missing numbers should remain missing. The methods come from the marketing plugin; this campaign is our own fictional teaching case.

The inquiry reads: “We combine eight store spreadsheets every Friday and want less repetitive work.” Record ID L-17 belongs to fictional Pine Workshop. You have twenty minutes for tomorrow’s demo. You do not yet know whether the columns match, who combines the files or who uses the result.

Give call-prep this inquiry and the same product facts. Request known needs, three discovery questions, a relevant demo outline and capabilities that must not be promised.

Useful questions include: “Do the eight files use the same columns and reporting period?” “Who makes a decision from the combined report?” “Is collection, merging or distribution taking the most effort?” Each advances this conversation.

Use account-research for additional background when sources are available. Eight stores does not establish a budget or buying authority. Keep unsupported fields unknown. For the later task of recording a meeting, see our separate customer follow-up guide.

Fictional ticket T-104 reports blank CSV exports since 09:20. Two staff reproduced the problem in Chrome. Existing reports remain readable. No data loss has been reported. The customer wants a fix before noon; the team has confirmed neither the cause nor a deadline.

Use ticket-triage to prepare a draft for T-104.
Separate facts, unknowns, priority rationale, suggested owner and first reply.
No historical ticket register was supplied: do not claim a duplicate search.
A requested noon fix is not an approved commitment.
Produce a draft only; do not send or create a ticket.

The result should preserve the working report-viewing function, leave the root cause unconfirmed and explain its priority recommendation. Check that recommendation against your own support policy. The reply can acknowledge the issue and ask for diagnostic details without inventing a repair deadline.

After someone verifies the resolution, give the confirmed steps to kb-article. A plausible explanation alone is not enough to publish a help article.

One actual trial exposed a useful failure. The skill read the ticket and avoided promising a noon fix. Yet the reply draft said the issue had already been marked high priority and was being investigated. Our task only requested a draft; neither assignment nor investigation had happened. Before sending, replace that with: “We have received your report. We still need to confirm the affected scope and reproduction steps so the team can arrange an investigation.” It also turned “no data loss reported” into “no data loss found.” Restore the narrower statement.

That is a concrete review task: check every claim that something has already happened against the actions actually taken.

This exercise has a person call each skill, inspect the result and pass on the relevant material. Automatic coordination between packages would require separate triggers, data transfer and authorization.

Start with the handoff that repeatedly slows you down. If meeting preparation means searching the same records every time, begin there. Improve the process when you can identify a concrete omission in its output.

Open Plugins in Onevium’s left navigation. Find Marketing essentials, Sales essentials or Customer support essentials in Onevium picks. These community selections retain the upstream author and license; Onevium maintains the adaptation.

Select your practice project in Scope, then click Add beside the package. In the installation window, check its name, version, publisher, included skills and selected scope. Read the contents, then choose Install and enable. Back in the directory, confirm that the package is enabled for your project before starting a new conversation turn.

If the new packages are missing, open Sources → Check for updates and wait for a successful directory refresh. You do not need to download a ZIP, import JSON or run a preparation script.

Open the installed package details and find Use in conversation. Use the complete name shown there—for this article, /onevium-customer-support-essentials:ticket-triage—followed by the task and inputs above. You can also explicitly ask the AI to use that installed skill. Check that the run loads the intended skill and reads the supplied material, then review the answer using the criteria in this article.

The companies and figures are fictional. The support case includes observations from one actual invocation; the other examples provide reference answers, not evidence that every skill has been tested in business use. See the plugin guide and Skills guide for more controls. The community repository provides source code, licenses and adaptation records.

Replace the fictional inputs with the correct product facts, customer scope, dates and permitted actions for each new task. Do not carry a sample commitment into real customer work.

A useful result lets a colleague see the evidence, understand what has been confirmed and decide what to ask next. For numerical analysis and contractual review, continue with data, finance and legal workflows.