Skip to content

InitiativesPage by page

A spike

A spike’s page says what it must find out, what for, and what the agent found, question by question. Click a spike’s title on Spikes, or its key in the tabs at the top. Its address ends in its key (/spikes/SP1).

SP1 of ACM-290 asked whether the Access report can be limited to one manager’s team. Both its questions are answered.

SP1 of ACM-290: the tabs Spikes and SP1; SP1 Answered; the title Can the Access report be limited to one manager's team?; Goal, Context and Acceptance; Investigates, part A1 Team view; Opened for, the unknown Does the portal know which people a team manager manages?, Answered, with its answer.SP1 of ACM-290: the tabs Spikes and SP1; SP1 Answered; the title Can the Access report be limited to one manager's team?; Goal, Context and Acceptance; Investigates, part A1 Team view; Opened for, the unknown Does the portal know which people a team manager manages?, Answered, with its answer.
  1. SP1. The spike’s tab.
  2. The spike’s title, under its key and status.
  3. Goal. What the spike must find out.
  4. Context. Why it matters, and what nobody has checked yet.
  5. Acceptance. What an answer must show for the spike to be done.
  6. Investigates. The part of the approach it looks into: A1.
  7. Opened for. The unknown or risk it covers, where that stands, and the answer once there is one.

A spike answered while the brief is still a draft has no unknown or verdict to answer. What the agent finds goes into the draft instead.

Under Questions, each question has its key (SP1-Q1, SP1-Q2…), where it stands, and the answer once there is one. A question nobody has looked into yet says Not answered, with Answer beside it.

The questions of SP1, both Answered: SP1-Q1, Where does the portal keep who manages whom?, with a Refine button, then Answer, Confidence high 13 Oct 2026, In the identity service…, and 2 sources in 2 repositories; SP1-Q2, Does the Access report read access through one query we can narrow to a list of people?, Confidence high, and 1 source in 1 repository.The questions of SP1, both Answered: SP1-Q1, Where does the portal keep who manages whom?, with a Refine button, then Answer, Confidence high 13 Oct 2026, In the identity service…, and 2 sources in 2 repositories; SP1-Q2, Does the Access report read access through one query we can narrow to a list of people?, Confidence high, and 1 source in 1 repository.
  1. Answered. The agent’s answer to the question.
  2. Refine. Ask the agent to correct or complete the answer.
  3. Confidence high. How sure the agent is, high, medium or low, and when it answered.
  4. 2 sources in 2 repositories. Where the answer comes from. Click it to list them.

Click an answer’s sources to see where the agent read each thing. The code comes first, by repository, with the lines it read. Then documents in your connected tools, and topics and initiatives in Emargen by key. Web pages come with the day the agent read them. The sources of ACM-320’s SP1:

The questions of ACM-320's SP1, both answered: SP1-Q1's 3 sources, Knowledge · 2 documents and Emargen, ACM-256; SP1-Q2's 3 sources in 1 repository, identity-service · 2 files, with their lines, and Web · read 9 Oct, Microsoft Entra ID: group claims in sign-in tokens.The questions of ACM-320's SP1, both answered: SP1-Q1's 3 sources, Knowledge · 2 documents and Emargen, ACM-256; SP1-Q2's 3 sources in 1 repository, identity-service · 2 files, with their lines, and Web · read 9 Oct, Microsoft Entra ID: group claims in sign-in tokens.

Answer opens a box with an optional Note for the agent. Say where to look, what you already know or what to rule out, or leave it empty, and press Answer. The agent looks in your organisation’s knowledge, Emargen’s record, the code or the web, as the question needs. It answers, and says where it read each thing.

While it works, the question says Researching…, and the chat shows a box with how long the agent has run and its latest steps, such as what it searched for or which page it read. It usually takes minutes. The agent answers one question at a time.

If the agent stops on the way, the question says Didn’t finish, and nothing is saved. The chat says so too, with Open, which brings you back to the question. Press Answer again.

Refine asks you what to change:

The Refine SP1-Q1 dialog: the question, Where does the portal keep who manages whom?; Your note, What is missing, what looks wrong, or where to look…; The agent updates only what your note and its sources call for. And Refine.The Refine SP1-Q1 dialog: the question, Where does the portal keep who manages whom?; Your note, What is missing, what looks wrong, or where to look…; The agent updates only what your note and its sources call for. And Refine.

Say what is missing or looks wrong, and press Refine. It stays greyed until you write something. The agent updates only what your note and its sources call for. If it doesn’t finish, the answer stays as it was.

When the last question has its answer, the agent reads the answers against the spike’s Acceptance. If they meet it, the agent answers the unknown or risk it was opened for and marks the unknown Answered and the risk Mitigated. Then it re-checks verdicts linked to that part or blocked by the spike. If they don’t meet it, the agent adds a question to find out the rest. Refining an answer later leaves the verdict as it was.

Everyone who can open the initiative reads its spikes, their answers and sources. Answer and Refine are for the owner and the people who edit it.