The problem with chat-based work instructions is not the chat app. It is a structure in which the outcome of the work and the decisions about it exist only in the chat. In a work design report we delivered to a law firm, three things were recommended to fix first: ① move work instructions, review and revisions into the matter record (Task); ② route the information questions among those reaching the owner to a system lookup; ③ connect consultation outcomes to follow-up work. None of the three is a feature to buy. Each is a change to how work is handled, and in the report’s roadmap all three sit in Phase 1, before any system is configured.
What this article covers
- Where chat-based work instructions become a cost in a law firm — observation records and outside research
- Fix 1: move instructions, review and revisions into the matter record
- Fix 2: sort the questions that reach the owner
- Fix 3: connect consultations to follow-up work — without presuming court data integration
- What to count after 30 days
Where chat-based work instructions become a cost in a law firm
The observation record in section 2.3 of our published work design report sample includes “work is shared, confirmed and revised through messenger and meetings” (Observed), “staff sometimes have to check each other’s progress separately” (Observed) and “some work information can only be learned by asking the principal directly” (Observed). Section 3.2 diagnoses that work requests go through an external messenger and that review and revision are separated from the work record, living in conversations and meetings. The report does not restrict using messenger for everyday communication (section 5.3). The problem arises because request, review, revision and completion are not recorded against the matter, so the next person has to find that conversation again and ask again.
Outside figures point the same way. According to Microsoft’s 2025 Work Trend Index special report “Breaking down the infinite workday”, the average Microsoft 365 user receives 153 Teams messages a day and is interrupted every two minutes during core working hours — 275 times a day — by meetings, emails or notifications. The cost of interruption shows up less as time than as state. In “The Cost of Interrupted Work: More Speed and Stress”, presented at CHI 2008 by Gloria Mark’s team at UC Irvine, participants actually finished interrupted tasks in less time, but reported significantly higher stress, frustration, time pressure and effort. For a firm that slots document review — work that needs sustained attention — between chat instructions, that result applies directly.
Section 3.6 of the sample splits messenger-related work in two. “Work requests by messenger” is classified Integrate (connect to Task-centred handling); “confirmation and revision by messenger” is Standardise / Integrate (into Task history). Neither is Automate. They are not candidates for automation; they are candidates for a different place of record.
Fix 1 — Move chat-based work instructions into the matter record
The rule in section 5.3 of the sample is one sentence: important communication about assigning, reviewing, revising and completing work is handled against the relevant Task, and the outcome of work or an important decision must never exist only in the messenger. Request → perform → review → revise → complete becomes one Task history.
The order of execution is in roadmap Phase 1 (section 7.1). Before the system is configured, seven standards are fixed: matters are managed as Projects (matter management), matter-related work attaches to that Project (work creation), instructions, revisions and review are Task-centred (communication), consultation information and follow-up work are structured (consultation records), information checks, approvals and professional judgement are distinguished (confirmation), meetings centre on problems and judgement rather than progress reports (meetings), and the reference location and recording rules for matter information are fixed (information management). Of these, two bear directly on chat-based work instructions: the communication standard and the confirmation standard.
On the ground it works like this. An instruction becomes a Task and only the Task link is sent in the chat. Revision requests go into that Task’s record and completion is marked in the Task. Meetings stop being a reading-out of progress already recorded in the system and deal instead with problems, delays, exceptions and judgement (section 6.6). Phase 3 (section 7.3) lists “work that drifts back to messenger” as a review item because the shift does not finish in one pass. Work that drifts back is tracked down and its cause checked on both the work design side and the system design side, rather than being put down to users failing to adapt.
Fix 2 — Sort the questions mixed in with chat-based work instructions
Instructions are not the only thing arriving by chat. Questions for the owner come through the same window. Section 3.3 of the sample sorts them into six types. Information (checking matter details), Status (checking progress) and Context (checking earlier consultations and decisions) are routed to a system lookup, Project/Task and Project history; only Approval, Professional Decision (legal judgement and matter strategy) and Exception (situations outside standard criteria) are judged by a person. The goal is not to reduce every question but to ensure that questions not requiring the owner’s professional judgement never reach the owner. How this standard is applied is covered separately in “Staff keep asking the owner”.
What the owner sees is cut to three columns (section 6.3): Action Required (work the owner must do), Decision Required (work needing professional judgement or approval) and Attention Required (delays, risks, key client issues). Work that is progressing normally and needs no judgement does not appear in those three columns. How review and decisions are handled is also fixed by type (section 6.5): simple completion by the assignee, compliance with set criteria by checklist, deliverable review by a named reviewer, legal judgement by an attorney, a change of matter strategy by the responsible attorney, exceptions by the person accountable. Once approval criteria are written down as sentences, the number of “can we go ahead like this?” messages falls by the same amount.
Fix 3 — Connect consultation outcomes to follow-up work
In a law firm, a large share of chat-based work instructions are generated right after consultations: the owner ends a call and sends “please prepare X for this matter” in the chat. Section 5.6 of the sample treats consultation not as a separate note-taking job but as part of the matter workflow: consultation → organise the notes → link to Customer → link to Project → record key facts and decisions → list action items → create and link Tasks. Section 6.4 adds assignee and deadline. The operating goal is not the recording itself but the consultation outcome flowing into the next piece of work without a break.
Court matter information is not presumed to be automatically integrated in this flow. The Korean court Judicial Information Sharing Portal provides basic matter information and progress history only through its linked API and states that an open API is “to be provided later”. Having confirmed this, section 3.6 of the sample excludes court matter information from custom automation, and section 5.7 rules that only the information needed for internal work is taken from the court system, recorded against the relevant Project and, where needed, linked to schedules and Tasks. External information is not rewritten in full; only the follow-up actions are structured.
Did chat-based work instructions fall? What to count after 30 days
The report does not judge the effect by whether things “feel easier” (section 8.1). In the 30-day review (section 8.4), the items that correspond to the three fixes in this article are:
- Messenger dependency — are work instructions and confirmations still happening outside the system?
- Questions to the owner — are things that can be read from the system still being asked of the owner?
- Consultation operation — are consultation records linked to follow-up Tasks?
As KPIs (section 8.2) you count the number of work instructions by messenger, the number of Information/Status questions reaching the owner, and the time from the end of a consultation to the creation of follow-up work. The target for all three is “decrease”, but no “x% decrease” is written. The report’s principle is that no savings rate is stated before quantitative data exists, and the first 30 days are for capturing the baseline.
The report these standards come from, and the audit
The standards above are part of a work design report actually delivered to a law firm; with only client-identifying details removed, the full text is published as a sample. Reading sections 3.3, 5.3, 5.6, 7.1 and 8.4 in sequence shows how diagnosis → design → operation → measurement connect. How to get the same audit is on the operating system audit page. The audit is KRW 800,000 for Basic (one work flow or one department, three days of on-site observation and five days for the report) and KRW 1,200,000 for Advanced (two or more work flows or the whole company, five days of observation and ten days for the report), excluding VAT, with a full refund before observation begins.
References
- Microsoft WorkLab — Work Trend Index Special Report: Breaking down the infinite workday (2025)
- Mark, Gudith & Klocke — The Cost of Interrupted Work: More Speed and Stress (CHI 2008, UC Irvine)
- Korean Court Judicial Information Sharing Portal — linked API guide (open API to be provided later)
Focus keyword: chat-based work instructions
SEO keywords: chat-based work instructions, law firm operations, matter-centred management, work design, operating system audit, founder dependency, consultation follow-up
자주 묻는 질문
Should a law firm stop using messenger altogether?
No. The report does not restrict messenger as an everyday communication tool. The standard is that work instructions, review, revisions, completion and important decisions must not exist only in the chat; they move into the matter's Task record.
Can court matter information be pulled in automatically?
Not at present. The Korean court Judicial Information Sharing Portal provides matter information only through its linked API and lists an open API as to be provided later. The design therefore does not presume integration; only the information needed for internal work is recorded against the matter Project.
How is the effect confirmed?
No savings rate is promised up front. After 30 days, count the number of work instructions by messenger, the number of Information/Status questions reaching the owner, and the time from the end of a consultation to the creation of the follow-up Task, and compare against the baseline.