Turn Work Conversations into Weekly Briefings and Handover Packs

Conversations are becoming work records

Slack now offers AI features that summarize and search messages stored in a company’s conversation space. It can summarize not only unread messages but also the past seven days or a period chosen by the user, and it shows the original messages and files used in each answer. Guide to AI features in Slack

Teams can register channels they care about and receive a daily digest of new conversations. For voice meetings, Slack turns key points and action items into a document, while notifying participants when recording starts and stops. Meeting participants can stop or delete the record. Slack huddle notes guide

Nansen said that up to 50 messages could accumulate in a single conversation when an operational issue occurred. The company said that, after using conversation summaries, it could draft incident-response documents 10 times faster. These are results published by Slack and its customer, however. Nansen case study

A sales leader at consulting company Plative said finding a customer’s situation took about 30 seconds, whereas it could previously require arranging a 30-minute call with a project manager. The company said the time needed to prepare customer-specific plans was cut in half. This is also a vendor customer story, so the same outcome cannot be assumed elsewhere. Plative case study

The same pattern is visible in Korea. Woowa Brothers (우아한형제들), the company behind the Baemin food-delivery platform, has said that about 1,700 people exchange more than 130,000 messages and over 6,000 files a day in Slack. At that scale, no person can read everything from a day, so channel structure, search, and retention practices become necessary alongside summaries. Woowa Brothers case study

How Monday morning changes at a small company

Imagine Kim Min-su, who runs a 12-person company that builds online stores. Every Monday, he opens six customer channels, email, meeting documents, and private messaging apps in sequence, then copies the previous week’s situation into a notes app. He may read the same conversation two or three times while trying to find why a schedule changed.

A change requested by a customer on Friday gets buried mid-conversation. A staff member’s statement that something “seems possible” is passed on as a confirmed promise. Work without an owner or deadline is only discovered in the next meeting. When a new employee joins, a long-serving colleague has to explain each customer’s history again.

This company could still fail after adding AI summaries. If decisions remain only in private conversations, channel names are inconsistent, or sentences such as “let’s do that” do not clearly identify what they refer to, the material being summarized is poor. Producing clean sentences is different from creating accurate work records.

The workflow can start simply. Assign one public work channel to each customer. From the previous seven days of conversation, AI finds “confirmed decisions,” “work in progress,” “answers the customer is waiting for,” and “owners and dates,” then arranges them with links to the source messages. Content without a source does not enter the confirmed-items list.

The first screen shows only three groups by customer name: red “Needs review,” yellow “In progress,” and green “Confirmed.” When Kim opens an item, he can immediately see who said what and when, then choose “Approve,” “Edit,” or “Not decided yet.” Only approved content enters the Monday briefing and customer-meeting preparation document.

After a meeting, new decisions are compared with existing items. If a delivery date changes, the old date is not deleted; the before-and-after dates and the reason remain together. Tasks without an owner trigger a request for confirmation from the relevant person. This is what turns a summary from disposable reading into a record that leads to the next action.

Many decisions still require human judgment. Kim must decide whether an ambiguous customer statement counts as a promise, who should first be told about a delayed schedule, and what to choose when two employees disagree. Private conversations about personnel or payroll should be excluded from the outset, and people should see content only within the access they already have.

In the end, summary quality alone does not determine success. More important are whether work conversations remain in public channels, whether decisions are separated from proposals, whether users can return to the source, and whether a responsible person approves the result. Without these four conditions, a quickly produced wrong answer can spread across the organization.

Abroad, the product is the workflow—not just the summary

In the United States, Anthropic collected conversations from multiple teams and created a daily briefing in a channel called #anthropic-times. Sales, research, product, and customer-facing staff reduced the need to check many channels one by one. Anthropic case study

The company’s actual cost has not been disclosed. Slack sells paid workspaces priced by active users. Slack and Anthropic presented figures on shorter deal cycles and annual cost savings, but these are their own calculations and need separate verification. The clear starting point was not a company-wide knowledge system, but one recurring daily summary.

Japanese data-analysis company BrainPad introduced morning summaries for managers who had to monitor many customer and project channels. One manager checked 30 of roughly 100 channels each day and moved about 40 lower-priority channels into the summary workflow. BrainPad case study

After three months, the company said the time needed to understand one channel’s situation fell from about three minutes to 20 seconds. That is a saving of two minutes and 40 seconds each time, which can add up to the time for a meeting for a manager viewing dozens of channels. This figure is also a customer’s own statement, and BrainPad later expanded Slack’s AI features across the company.

In the United States, HireVue used Guru, an internal knowledge service, for training and question handling for 30 customer-support employees. At first, new employees spent five weeks learning four product lines. The company then gathered training materials and verified answers in one place, so that a question asked in Slack could retrieve relevant documents. HireVue case study

Guru uses custom pricing that requires a separate inquiry. HireVue said new-hire training fell from five weeks to two, while Slack questions dropped 40% even as support requests increased by more than 500. The starting point in this case was not training on every conversation, but first deciding which official documents would answer recurring questions.

What could be built from this

1. Customer promise inbox

• What it does: Finds potential commitments about price, delivery dates, and revision scope in work conversations, then asks a responsible person to approve them.
• Who uses it: The owner of a design studio with around 10 employees producing advertising materials for several customers at once.
• Why now: Conversation summaries and source links can reduce the time people spend finding promise candidates from scratch.
• First screen: Places “Promises awaiting approval,” source statements, owners, and expected dates side by side for each customer.

2. Pre-departure handover pack

• What it does: Groups recurring work, ongoing work, and key contacts from a departing employee’s public work conversations and documents.
• Who uses it: A company with 20 or fewer employees where supplier work is concentrated with one accounting or operations employee.
• Why now: Past conversations can be found by period and topic, so a handover document does not have to start from a blank page.
• First screen: Shows “Work to do this month,” “Unfinished work,” and “People the successor will need to ask” first.

3. Missed customer-request alert

• What it does: Finds unanswered questions and requests without an owner in customer-support channels, then alerts the team.
• Who uses it: A small food-ingredient supplier that receives order changes by phone, messaging app, and group chat.
• Why now: Rather than perfectly organizing every conversation, it can narrowly check whether a reply exists and whether an owner has been assigned.
• First screen: Shows the longest-waiting requests first, with the customer name, elapsed time, last reply, and an owner-assignment button.

4. Decision-change record board

• What it does: When a decision changes in meetings or conversations, it keeps the previous decision, the reason for change, and the approver connected.
• Who uses it: A site manager at a window-installation contractor where construction schedules and material choices change often.
• Why now: It can go beyond summarizing the latest conclusion by comparing it with source material from earlier conversations to create a change history.
• First screen: Puts “Decisions changed this week” at the top, showing the before and after, the reason, and approval status on one screen.

Why this matters where you are

Check whether your team can find a final decision, an owner, and a deadline from one difficult week of work conversations within 10 minutes. The communication tools, privacy rules, and channels may differ in your market, but the need to distinguish proposals from approved commitments remains. Start with one repeated workflow, such as customer promises or unanswered requests, rather than trying to organize every conversation.

What to check today

Ask one potential customer to open their most complicated work conversation from the past week and try to find the final decision, owner, and deadline within 10 minutes. If they cannot find even one of the three, or different people give different answers, it may be a problem worth testing. If all three are immediately clear, look for another pain point before building a summary service.