Craig Stanley
Home / Decisions / Decision records / A 30-second capture in Teams

A 30-second capture in Teams

A decision record only helps if people write it. A short Teams card that captures the essentials in 30 seconds, saved to a SharePoint list.

· 3 min read · Craig Stanley
In short, explained

When you make a choice, a little box pops up in Teams. You fill in a few words and how sure you are, press a button, and it's saved.

The full decision record has ten fields, which is too many in the moment. This version asks for five, on a card in Teams. The answers go into a SharePoint list, and the rest of the record can be filled in later. The aim is a habit that takes half a minute.

A Power Automate flow posts an adaptive card (wait for response) to the decision owner in Teams, capturing decision, options, choice, confidence and review date, and writes the response to a SharePoint list keyed to the full schema. Trigger it from an approval, a meeting tag or a manual button.

Why a short version

The decision record schema has ten fields. That's right for a review, but it's too much to ask in the moment, and the moment is when the record is most valuable. A record written afterwards is shaped by what happened, which is exactly the outcome bias the record is meant to avoid.

So I'd split it. Five fields at the time of the decision, in about 30 seconds. The rest later, when there's time.

The five fields

FieldExample
DecisionExtend the agent pilot to the north team
Options consideredExtend, wait a month, stop
ChoiceExtend
Confidence it achieves the goal70%
Review date15 January

Those five are the ones that can't be reconstructed fairly later. The goal, the evidence, the owner and any model involvement can be added afterwards from other records.

How it works in Teams

Power Automate's Teams connector can post an adaptive card to a person and wait for their response. Microsoft's documentation says the flow pauses until the recipient responds and then uses the response in later steps. So a flow can:

  1. Post a card to the decision owner with five input fields.
  2. Wait for them to submit it.
  3. Write the answers, the person and the time to a SharePoint list.
  4. Reply with a link to the list item, for anyone who wants to add more.

Microsoft notes that a card created with a "wait for a response" action can only be submitted once, and recommends setting an update message so the card shows it's been used. I'd set it to "Recorded. Add detail here", with the link.

What starts the flow

There are three triggers I'd consider, from least to most effort for the person.

The first is an approval. If a decision already goes through an approval flow, the capture card can follow the approval automatically. The second is a manual button in Teams, which anyone can use when they've just made a call. The third is a meeting: someone types a tag in the meeting chat, and the card goes to the meeting organiser afterwards. The first is the most reliable, because it doesn't depend on anyone remembering.

A worked example

This example is illustrative. A team lead approves a request to give a Copilot Studio agent access to a new knowledge source. The approval flow completes, then posts the capture card. She fills it in:

FieldEntry
DecisionAdd the 2025 policy archive to the HR agent
OptionsAdd all, add 2025 only, don't add
ChoiceAdd 2025 only
Confidence80% that answer quality improves
Review dateSix weeks

It takes about 30 seconds. Six weeks later, a scheduled flow reminds her to review it, with her 80% in front of her. That's the moment the record earns its keep, because she can see what she expected and compare it with what happened.

Making it stick

The capture has to feel almost as quick as skipping it. Keep the card to five fields, prefill anything the flow already knows, and never make a field compulsory except the choice. A record with three fields filled in is far more useful than no record.

What I'm still checking

Confidence is the field people are least comfortable with. Some will leave it blank or always write 50%. I'd offer five fixed options (50, 60, 70, 80, 90%) rather than free text, and see whether that helps. Once there are enough records, calibration can be checked.

Sources

Read next

A question to take awayWhich repeated decision would you trust a cheap model to score first, with a person checking the close calls?

About me

Craig Stanley

Microsoft AI consultant and technical architect, based in Whitley Bay. Over the last few years I've delivered Microsoft 365 Copilot, Copilot Studio agents, Microsoft Foundry (formerly Azure AI Foundry) work and governance for UK public sector and financial services organisations.

What interests me is the decision underneath the tool: what it costs, what it risks, and whether a small, transparent model can make it better. I write the methods up here and on Substack so anyone can use them.

I write this site to learn in public: explaining each idea simply is how I check I understand it. Why I write this site.

Find me