Signals in the Stream: Auditing Decisions Made in Slack and Teams

Team chat is where work happens, but it is also where decisions go to blur. A product manager posts a proposal in Slack, three colleagues react with a thumbs-up, one replies “fine by me” and the conversation moves on. Two months later, nobody can say who approved the spend, what exactly was approved or where the record lives.

The problem is not that chat is informal. The problem is that UK teams increasingly rely on informal signals for formal outcomes: releasing budget, publishing a policy change, switching a supplier or signing off a customer-facing message. A reaction is a signal. A decision needs a record.

Why chat feels like a decision

Chat tools are designed for flow, not custody. Slack and Microsoft Teams make it easy to respond quickly, and a reaction carries social meaning: approval, acknowledgement, agreement or simply “seen”. In a busy channel, that meaning can harden into assumed authority.

The risk grows when decisions are distributed. A thread may contain the proposal, a poll may contain the preference, a direct message may contain the caveat and a reaction may contain the approval. No single artefact tells the whole story. If the person who started the thread leaves, the context often leaves with them.

This is not a reason to ban informal agreement. It is a reason to audit where informal agreement is being used for decisions that need a named owner, a written record and a retention point.

The three-question test

Before adding another approval layer, apply a simple test to each recurring decision type. A chat decision needs escalation to a record if the answer to any of these questions is yes.

• Authority: Does the decision commit money, people, suppliers, customers or public statements?
• Specificity: Would a reasonable person need to know the exact scope, date or conditions to act on it later?
• Persistence: Will someone outside the original conversation need to find or rely on it in three, six or twelve months?

A quick “looks good” on a meeting-room booking may pass all three tests as a casual signal. A thumbs-up on a £15,000 software renewal does not. The test is not about hierarchy. It is about whether the decision needs to survive the conversation.

A half-day audit in four passes

A practical audit does not require a new platform. It requires one person with access to the main channels and a representative sample of recent decision-heavy conversations.

Pass one: inventory the decision types

List the recurring decisions that currently surface in chat. Typical examples include campaign approvals, spend under a threshold, policy exceptions, supplier changes, customer communications and changes to published content. Keep the list short and concrete. “Marketing decisions” is too broad; “approving a paid social budget increase” is auditable.

Pass two: classify the evidence

For each decision type, identify where the evidence currently sits. Is it a poll result, a thread, a pinned message, a reaction count or a direct message? Note whether the evidence includes the proposal, the approver, the date and the scope. A reaction count without the original proposal is not evidence of what was approved.

Pass three: trace one real example

Choose a recent decision of each type and try to reconstruct it from chat alone. Give yourself ten minutes. If you cannot answer who approved it, what they approved and when, the decision is not recorded. This is the most useful part of the audit because it replaces opinion with a concrete failure mode.

Pass four: assign a record owner

The owner is not always the most senior person in the thread. The owner is the person accountable for ensuring the decision reaches the record. For a spending decision, that may be the budget holder. For a published policy, it may be the content owner. The owner should be named in the record, not inferred from channel membership.

Turning a reaction into a record

You do not need a formal committee for every decision. You need a lightweight template that can be completed in the same channel and then moved to a durable location. A useful record has five fields:

• Decision: a one-sentence statement of what has been agreed.
• Owner: the named person accountable for the decision and its record.
• Scope: what is included, what is excluded and any conditions.
• Date and evidence: when the decision was made and where the discussion can be found.
• Review point: when the decision will be revisited or closed.

A simple message pinned in the channel can work if the channel has a retention policy and searchable access. A line in a shared decision log is stronger. The important discipline is that the record is created at the point of decision, not reconstructed after an incident.

Retention and access

Chat retention is not the same as decision retention. Most organisations set channel retention for operational convenience, not for governance. A decision made in a channel with a 30-day retention window may disappear long before the business needs to review it.

UK teams should align the record with the relevant retention schedule, whether that is financial, contractual or policy-related. The record should be accessible to the people who need it, without requiring a search through thousands of messages. In practice, that means a small number of durable decision logs rather than a sprawling archive of screenshots.

What good looks like after 30 days

A successful audit produces a short list of decision types that now have a named owner and a record template. It also produces a short list of decision types that deliberately remain informal. Both lists are useful. The point is not to formalise every conversation; it is to stop formal outcomes from depending on informal signals.

The thumbs-up can stay. It is a useful social cue. But when the question is “who agreed to this, and where is it written down?”, a reaction is not a signature. It is a starting point for a record that someone has to own.