Part B · after “How DMA AI works”

The screens

A short guide to each screen of the DMA AI dashboard: what it's for, the tasks you can do there and how many clicks each takes, and what runs behind it. Every screen in the dashboard links here with “How this works”.

No.	Screen	What it's for
B.1	Today	The home screen: how much dormant revenue the AI found in the CRM, the ten accounts most worth reviving today with the reason why now, and what needs a person. Signing in lands here, so the answer takes no clicks.
B.2	Approvals	One queue for everything the AI wants to do that needs a person: outreach e-mails, CRM updates from a call, proposals, and changes to the AI's own settings. A second tab lists what the AI wrote to Nutshell on its own, with undo.
B.3	Escalations	Cases the AI handed to a person: a visitor asking for someone, a pricing or contract question, a hostile message, low confidence, or a hot lead nobody has contacted. Each has a short brief, an owner and a response deadline, and the AI holds that thread until a person takes it.
B.4	Live chats	The website conversations happening now. Anyone on the team can join one and answer the visitor themselves; the AI pauses for that visitor until the chat is handed back.
B.5	Accounts	Every company from Nutshell with its tier, estimated value, status and next step, and one click to its Account 360: the facts from old notes with their quotes, the people and whether each can be contacted, the deals, and the timeline.
B.6	Pipeline hygiene	What the data-quality rules found on the CRM copy: stuck deals, deals with no next step, records owned by people who left, contacts who left or bounced, and likely duplicates. People fix the records in Nutshell; a flag closes on the next run.
B.7	Agents	The ten AI workers, what each one does, its prompts, its model and the tools it may use. Any change is tested, then approved by someone other than its author before it goes live, and can be reverted.
B.8	Tools	Everything the AI can do, as a registry: 8 tool servers split by risk and the 40 narrow tools they offer, with which agents hold each one. New servers can be added; nothing they offer runs until someone else approves it.
B.9	Knowledge	The approved library of what the AI may say about DMA: service pages, FAQs, objection answers, case studies and battlecards. A document goes live only after a second person approves it, and the website chat only ever uses live public documents.
B.10	Taxonomy	The labels the AI sorts things into: services ("service lanes"), intents, objections, lost reasons, buyer types, record classes and consent bases. Renaming or adding an item makes a new version that someone else approves.
B.11	AI Lab	Where changes to the AI are tested before anyone approves them: try one question and see exactly what the model was given, run a test suite on saved cases, or compare two versions on the same cases.
B.12	Controls	The kill switches and the sending limits. Pause everything, or one capability at a time; nothing already queued is dropped, it waits until you resume.
B.13	Escalation roster	Who gets each kind of case first, their backup, how long before a case moves on, working hours, the after-hours rule, and who is on call right now. Changes wait for a second person's approval.
B.14	Rules	Everything the AI may and may not do, written as plain sentences: which actions each tool server may take, the send governor, the claims checker, the consent matrix, the circuit breakers, the escalation triggers and the sign-offs. Change a value, test it against recent decisions, submit it: someone else approves before it's live.
B.15	Fields	The field sheet: for each Nutshell field, whether the AI may change it, at what autonomy and with whose approval, and which fields stay people-only. DMA signs the sheet off; every change is four-eyes.
B.16	Observe	Everything the system does, as it happens: syncs, AI calls, decisions, writes to Nutshell, approvals and e-mails. Any step opens its full trace, so nothing the AI does is a black box.
B.17	Measurement	What the outreach actually earned, measured against a held-out group that got nothing, with the range of the result and a verdict in words. A second tab backtests the scoring on past deals.
B.18	Log	Every AI decision, with its inputs, the agent, the tool, the rule result, the model's tokens and the cost, and the AI spend this month against its budget.
B.19	Kickoff checklist	The 13 things the project needs from DMA, the day each is needed, what it unlocks and what moves if it's late, with an owner, a status and notes for each.
B.20	Settings	Integrations (Nutshell, and Slack as an option), the team and invitations, e-mail health (the sending domain, the warm-up and list verification), keyboard shortcuts and appearance.
B.21	Connect Nutshell	The first run: paste a Nutshell user e-mail and API key, and the system copies the CRM, scores it and shows the deals most worth reviving, in about a minute on the first results.