Confidential · © 2026 Space-O Technologies Inc. · Prepared exclusively for DMA. Please do not copy, save or share.

Section 16 · Management Intelligence — Ask the Business Questions

16.1Ask DMA data in plain English

Built in the POC
No.
16.1
DMA's reference
Decision Book p.30
Phase
Phase 3
Autonomy
L1 · Recommend
DMA's decision
Yes

◆ Amended in v3 of the book.

What DMA asked for · Decision Book p.30

Let DMA ask its data questions in plain English, for example "show me every PPC opportunity over $20,000 a month not contacted in 7 days".

Why it matters to DMA: Management gets answers without a report request; DMA's amendment approves it as part of management intelligence.

In our words, from DMA's Product Decision Book v3, page 30.

How we do it

  • DMA's book schedules this for Phase 3; the POC already has it in the team workspace (LibreChat, branded DMA AI), for the questions its tools cover.
  • The workspace assistant answers through narrow, read-only tools, never raw database access: search accounts by tier, service, owner, status or country; the money list of dormant opportunities; an account brief; an account's timeline; a past conversation; a score and its reasons.
  • Numbers in the answer are copied from the tools (which compute them in code); the assistant is told never to recalculate them. Every call is logged with its cost.
  • Questions the tools can't answer yet (such as a value threshold combined with "not contacted in 7 days") get an honest "I can't filter on that yet"; Phase 3 adds those filters.

The flow

It plays on its own while it's on screen; hover or use the controls to pause or step through.

resultsTeam memberasks in plain EnglishWorkspace assistantpicks a toolRead-only toole.g. account briefBridge databasenumbers from codeAnswerwith the figures
Step 1 of 4

A question

"Brief me on Northside Dental."

  • Person
  • AI
  • Rule in code
  • Stored data
  • Screen

Where the data goes

The same six stages on every page. Nothing reaches Nutshell except through the write gate.

  1. 1Team workspace

    Source

    A question typed by someone at DMA.

  2. 2Sync

    Copies Nutshell changes into the bridge database

    Keeps the data the tools read current.

  3. 3Bridge database

    A copy of the CRM data, plus what the AI works out

    The accounts, scores, timelines and conversations the tools read.

  4. 4AI

    Claude models, only through DMA's own gateway

    Chooses the tool and writes the answer; numbers come from the tool.

  5. 5Write gate

    The only way back into Nutshell: checked, approved when needed, sent once

    Not involved: these tools only read.

  6. 6Nutshell

    The system of record

    Unchanged.

The tables behind it

Drawn from the POC's database catalogue: structure only, no data.

agent.mcp_toolsPKiduuidFKserver_keytext+ 11 more columnsagent.tool_grantsPKagent_version_iduuidPKtool_iduuid+ 3 more columnsai.account_twinPKaccount_iduuidPKtwin_versionintegerlifecycle_statetexttiertextnbatextIDprompt_iduuidgenerated_attimestamp+ 9 more columnsops.llm_callsPKiduuidPKoccurred_attimestampcaller_servicetextagent_keytextIDagent_version_iduuidIDrun_iduuidIDconversation_iduuidIDdecision_iduuidmodel_resolvedtextinput_tokensintegeroutput_tokensintegercost_usdnumeric+ 25 more columns

PK primary keyFK reference the database enforces (solid line)ID reference kept by id (dashed line)

TableWhat it holdsColumns
agent.mcp_toolsEvery tool an AI worker may call, with its risk and the lowest autonomy level it needs.13
agent.tool_grantsWhich AI worker may use which tool.5
ai.account_twinThe account profile the brief is built from.16
ops.llm_callsEvery AI call: who asked, model, tokens, cost.37
How the tables connect (8)
ColumnPoints toKept by
agent.mcp_tools.server_keyagent.mcp_serversThe database (foreign key)
agent.tool_grants.agent_version_idagent.agent_versionsThe database (foreign key)
agent.tool_grants.tool_idagent.mcp_toolsThe database (foreign key)
ai.account_twin.prompt_idagent.promptsThe application (by id)
ops.llm_calls.agent_version_idagent.agent_versionsThe application (by id)
ops.llm_calls.run_idagent.runsThe application (by id)
ops.llm_calls.conversation_idconv.conversationsThe application (by id)
ops.llm_calls.decision_idctrl.decision_logThe application (by id)

Worked example

Synthetic demo data: every name, business and number is made up.

One question in the team workspace

1 / 2

What you see

1.The question

Someone at DMA types a plain question.

localhost:13313/c/…
The team workspace (DMA AI): a plain-English question and the brief it returns.
Callout 1

What happens in the system

  • AI agentworkspace-assistant

    The workspace assistant reads the question; its instructions are a versioned, approved prompt.

Questions the workspace answers today, and the tool behind each (synthetic demo data).

Questions and tools

QuestionTool it uses
Brief me on Northside DentalAccount brief
Top 10 revivable deals todayDormant opportunities (money list)
Which tier-A accounts in legal does Sean own?Search accounts
What happened with Harbor & Finch last year?Account timeline

In the running POC

localhost:13313/c/…
The team workspace (DMA AI): a plain-English question and the brief it returns.

The team workspace (DMA AI): a plain-English question and the brief it returns.

  1. 1The question, as typed.
  2. 2The answer, built from the account's data.

Team workspace (LibreChat, branded DMA AI) · account brief. Captured from the running POC on September 25, 2026; synthetic data.

Status

Built in the POC

Built in the POC. This capability runs in the POC today, on synthetic data, end to end.

16.1 Ask DMA data in plain English · DMA AI proof