CEO Agent¶
Thinklio Built-in Agent Specification Version 0.2 | March 2026
1. Purpose and Problem Statement¶
The CEO Agent works with a chief executive, not instead of one. It is a coordinator, sounding board, and context agent — the highest-level agent in the Thinklio system, with delegated access to the full suite of specialist agents and broad visibility across the organisation's knowledge.
Its job is threefold:
Information discovery. The organisation's information is distributed across every agent in the workspace. A CEO needs to be able to ask any question — about the pipeline, the team, the finances, the relationships, the strategic landscape — and get a synthesised, executive-level answer without knowing which agent holds the relevant data or having to query each one manually. The CEO Agent knows where everything is and how to get it.
Workflow protection. A CEO's most valuable resource is strategic attention. Operational detail, routine decisions, and information that should be handled below them constantly intrude. The CEO Agent filters, routes, and delegates — protecting the time and attention that should be spent on genuinely strategic work.
Sounding board. A CEO's most important thinking often happens in dialogue: working through a strategic decision, pressure-testing an assumption, exploring a course of action before committing. The CEO Agent can hold that conversation — asking good questions, surfacing relevant considerations from across the organisation's knowledge, pushing back constructively, and being honest when the answer is genuinely uncertain.
The second purpose worth naming directly is organisational signal. A CEO who uses Thinklio seriously, and is seen to do so, changes their organisation's relationship with AI more effectively than any policy or training programme. The CEO Agent should be the kind of tool a CEO is comfortable using in front of their board, their team, and their investors — not because it makes them look impressive, but because it makes them demonstrably better prepared, better connected, and better informed.
2. Architecture: The CEO Agent as Coordinator Root¶
The CEO Agent is architecturally distinct from every other agent in the system. Where most agents are specialists, the CEO Agent is a coordinator with organisation-wide access. It does not hold data; it knows where data lives and how to retrieve and synthesise it for an executive audience.
CEO Agent (coordinator root)
├── Personal Assistant — schedule, tasks, communications execution
├── Briefing Agent — structured preparation documents
├── Research Agent — external intelligence
├── Customer Intelligence — account health, relationship signals
├── Project Coordinator — initiative status, blockers
├── Finance Agent — financial position, budget, anomalies
├── Enquiry Agent — inbound contact volume, escalations
├── Support Triage Agent — support health, recurring issues
├── Rolodex — relationship context and history
├── Monitor Agent — triggered alerts across configured watches
├── Digest Agent — external landscape signals
├── Knowledge Base Agent — organisational reference
└── Pulse — team communications health
When a CEO asks "what do I need to know today?", the CEO Agent queries the relevant specialist agents in parallel, receives their outputs, and synthesises a CEO-level answer. The CEO does not need to know which agent holds what — they ask one agent and get one answer.
This architecture means the CEO Agent becomes more valuable as more specialist agents are deployed. With only a few running, it has limited data to synthesise. With the full suite active, it has a complete picture of the organisation.
3. What This Agent Is Not¶
Not a decision-maker. The agent prepares, surfaces, and engages in dialogue. It does not decide.
Not a replacement for the Personal Assistant. The PA handles execution — scheduling, task management, communications logistics. The CEO Agent operates above that level. Most CEOs will run both. The PA is a delegate of the CEO Agent.
Not a dashboard. A dashboard shows "revenue is $2.3M." The CEO Agent says "revenue is tracking 8% below plan, the shortfall is concentrated in two accounts that both have open relationship issues unaddressed for three weeks, and the Customer Intelligence Agent has flagged both as at-risk."
Not a general chat assistant. The Chat Agent exists for that. This agent is configured for a specific person with knowledge of their organisation, their relationships, their priorities, and their decision-making context.
4. UI Structure¶
Designed for short, high-value sessions — every view useful in under three minutes, on a phone as much as a desktop.
Chat Tab¶
The primary surface. Used for:
- Sounding board conversations — "I'm thinking about whether to take Series A capital now or wait six months — talk me through the considerations" — the agent engages with strategic ambiguity, asks clarifying questions, surfaces relevant data, and helps the CEO think
- Pre-meeting briefs — "brief me on the 2pm board call" — delegates to Briefing Agent, Rolodex, Customer Intelligence as needed
- Discovery questions — "what are our three biggest relationship risks right now?" — queries across agents and synthesises
- Post-meeting capture — "just finished with Marcus, he's committed to closing by end of quarter, flag this if we haven't heard from him in two weeks" — transcribes, creates the follow-up task, logs the Rolodex interaction
- Routing and delegation — "pass this to the finance team with context" — prepares a brief and routes without the CEO context-switching into operational detail
Voice input is as important as text. A CEO walking between meetings should be able to capture a thought or request a brief by speaking.
Today Tab¶
A synthesised morning briefing assembled by querying the specialist agent set. Not a task list. A narrative:
- Upcoming moments — today's meetings and calls, each with a one-paragraph context note and a "brief me" link
- What needs attention — items surfaced overnight: escalated Items, at-risk relationships, triggered monitors, pending decisions
- Relationship signals — a key investor's company made an announcement, a board member has an unanswered message, a major customer had three escalations this week
- One thing — the single item most deserving the CEO's focus today, with brief reasoning
Readable in under three minutes, on a phone, with the first coffee.
Relationships Tab¶
The CEO's strategic relationship layer. Not the full Rolodex — a curated, prioritised view of board members, investors, major customers, strategic partners, key staff, and regulators.
For each relationship: - Health score (recency of CEO contact, open commitments, sentiment trend, strategic alignment signals) - What is overdue — commitments not delivered, contact that has lapsed - Recent signals — news, interactions by other team members, changes in their circumstances - Suggested action — reach out, follow up, congratulate, nothing needed
Commitment tracking is the highest-value function here. When the CEO makes a commitment in any conversation — "I'll get you the numbers by Friday," "let's speak next week" — the agent captures it, creates a task, and surfaces it before the next interaction with that person. Forgotten commitments are reputationally costly. The agent makes this structurally impossible to miss.
Horizon Tab¶
The 30/60/90-day forward view: - Upcoming decisions with suggested preparation timelines - Key dates: board meetings, investor calls, contract renewals, regulatory deadlines, milestones - Trends building: themes growing in significance across Monitor, Digest, and internal data - Strategic threads: ongoing initiatives and current status
5. The Sounding Board Capability¶
This is the most underspecced aspect of AI assistance for senior leaders, and the most valuable.
A CEO's hardest thinking happens in dialogue — not "give me information" but "help me think through this." The CEO Agent must hold this kind of conversation effectively. That requires:
Asking good questions. When the CEO presents a strategic question, the agent's first move is often a question back — not to stall, but to understand the actual decision. "Are you asking whether to enter the market at all, or whether now is the right time?"
Surfacing what the CEO might not have considered. The agent draws on organisational knowledge, external research, and the CEO's own decision history. "The Customer Intelligence data shows three existing clients are in adjacent markets — is this an expansion opportunity rather than a new market entry?"
Pushing back constructively. If the CEO has reached a conclusion the agent's analysis suggests may be incomplete, it says so — specifically, not vaguely. "The support triage data from the last six weeks suggests the product stability issue may be more significant than it appears in this framing — worth weighing before committing to an aggressive growth timeline."
Honesty about uncertainty. The agent does not project false confidence. When a question is genuinely uncertain, it says so and explains why. A CEO who starts relying on an agent that sounds confident but is poorly grounded is worse off than a CEO with no agent.
Knowing when to stop. When the CEO has reached a considered conclusion, the agent's job is to summarise the reasoning clearly and move to action — not to keep generating considerations.
The sounding board mode is configurable: supportive, balanced, or challenging. Most CEOs will want balanced by default — an agent that helps them think, not one that just agrees, and not one that argues for sport.
6. Organisational Pulse¶
The CEO Agent assembles a daily synthesis by querying the specialist agent set. Not a data dump — a synthesised narrative identifying patterns and surfacing what deserves the CEO's attention.
| Agent | What it contributes |
|---|---|
| Customer Intelligence | Account health, at-risk relationships, pipeline signals |
| Support Triage | Support volume, recurring issues, escalation patterns |
| Project Coordinator | Initiative status, blockers, milestone risks |
| Finance Agent | Budget position, spend anomalies, pending approvals |
| Enquiry Agent | Inbound contact volume and type trends |
| Monitor Agent | Triggered alerts from configured watches |
| Digest Agent | External landscape signals relevant to current priorities |
| Pulse | Team communications health |
| Taskmaster | Overdue items across the organisation (aggregate, not individual) |
The synthesis identifies cross-agent patterns no individual agent would surface. A support spike, a declining customer health score, and a stalled project in the same product area might be three separate signals — or one systemic problem. The CEO Agent's job is to notice the connection.
7. Privacy and Access Model¶
Read access: Broad, across all agents in the workspace, at an aggregate level. The CEO Agent cannot read individual users' private knowledge layers or personal conversations.
Write access: Narrow. The CEO Agent writes only to the CEO's personal knowledge layer, the strategic relationship layer, and explicitly created tasks and interactions. It does not write to team or org knowledge layers without explicit instruction.
CEO's own knowledge: The CEO's strategic notes, personal assessments of people, and relationship observations are the highest-privacy data in the system. Stricter than the Coach Agent — never accessible to org admins, never surfaced to other agents without explicit instruction.
Audit trail: Every action taken on the CEO's behalf is logged immutably.
8. Configuration¶
8.1 Initial Setup (guided)¶
- Identity and organisational context — who the CEO is, what the organisation does, current stage and priorities
- Agent connections — which specialist agents are active (defaults to all; can be scoped)
- Strategic relationship layer — initial population from Rolodex with strategic weighting
- Communication and dialogue style — how direct, how much challenge vs. support
- Privacy settings — what is strictly private, what the PA can access
- Sounding board mode — supportive / balanced / challenging
8.2 Ongoing Settings¶
| Setting | Description |
|---|---|
| Strategic priorities | Current focus areas; drive attention scoring across all views |
| Connected agents | Which specialist agents to draw from |
| Relationship categories | Who is in the strategic layer, with weighting |
| Brief depth | Default detail level for preparation briefs |
| Sounding board mode | Supportive / balanced / challenging |
| Capture preferences | Voice-first, text-first, or context-dependent |
| Daily synthesis time | When the Today tab is assembled |
9. Use Cases¶
UC-1: Strategic sounding board¶
A CEO is considering whether to accelerate the product roadmap to respond to a new competitor. "Talk me through whether we should pull forward the Q3 roadmap to respond to what Acme just announced." The agent asks two clarifying questions, then surfaces: current engineering capacity (from the Project Coordinator — two engineers are already overcommitted), a Customer Intelligence signal (three key accounts have mentioned the competitor in the last month), and a pattern from the CEO's own decision history. It does not recommend. It helps the CEO see the decision clearly.
UC-2: Morning synthesis¶
Four minutes with the Today tab. The agent has assembled overnight: one customer account moved to at-risk status, a Monitor alert that resolved itself, a board member with an outstanding email from 12 days ago, and a note that a requested document has not been started. Three things to handle. One brief to request. The CEO is oriented for the day.
UC-3: Commitment capture on the move¶
After a call with an investor: "Log call with David Chen — positive on the timeline, wants the updated financial model before end of month. Follow up if I haven't sent it by the 20th." The agent logs the Rolodex interaction, creates a Taskmaster task, and adds a commitment note to David's relationship record. The CEO is already in the next meeting.
UC-4: Routing an operational intrusion¶
An operational issue arrives at CEO level. The agent recognises it: "This looks like it belongs with the engineering lead — should I route it with context?" The CEO approves. The agent creates a task with a full brief and routes it. The CEO never context-switched.
UC-5: Board preparation¶
Board meeting in 48 hours. The agent delegates to the Briefing Agent with CEO-level scope, incorporating financial position, initiative status, account health, and a synthesis of the three items most likely to require board input. Nine minutes to read. The CEO is prepared.
10. Open Questions¶
- Delegated access across all agents requires careful privilege design. Does this require org-owner role, or a specific access tier above org-admin?
- Voice capture is important for mobile use cases. In scope for initial release?
- The decision log — records of significant decisions with context and reasoning. When a CEO leaves the organisation, does this stay with the organisation or leave with the person?
- The sounding board capability requires constructive pushback without sycophancy. How is this calibrated and tested?
- The organisational pulse degrades gracefully as fewer specialist agents are deployed. Is the degradation experience (partial data, clearly flagged) sufficient, or does the agent need a minimum viable agent set before it is useful enough to deploy?
The CEO Agent is worth building well. A CEO who uses it seriously is the platform's most credible reference — and its most visible demonstration of what AI-assisted work actually looks like.