How Your AI Agents Work as a Team (Delegation + @-Mentions)

Your AI team coordinates like a real team — and you stay in control. When you ask a question, the right agent answers it. When one agent needs another's expertise, it delegates. You see the full chain. Nothing happens in a black box.

The Roster

Your AI team is made up of specialist agents — each with a defined role, defined authority, and defined limits. The agents available to you depend on your plan tier.

• Strategy and marketing agent — Handles campaign direction, content briefs, audience analysis, and marketing performance questions. Think of this as your senior marketing voice.
• Operations agent — Handles supplier coordination, drafted orders, staffing schedule drafts, and cost notes. The agent that runs the daily operational work.
• Customer success agent — Manages customer follow-up, satisfaction tracking, and retention workflows. Keeps your existing clients from going quiet.
• Finance agent — Drafts revenue summaries, invoice reminders, and expense notes from the figures you share. If your accounting tool (QuickBooks, Xero) has a way in, the link is part of the Business Partnership plan.*
• Project and delivery agent — Tracks build milestones, task completion, and delivery timelines across active projects.
• Product and market agent — Supports positioning, competitive research, and market analysis for owners building or evolving their offer.

* Some tools take longer and need a workaround. Connecting your tools is part of the Business Partnership plan. Tool names belong to their owners; naming one means we can work with it, not that they endorse or certify us. See what we connect.

These roles are not job titles bolted onto a single AI. Each is a distinct agent with its own authority boundary — what it can access, what it can do, and what it must escalate to you.

How @-Mentions Work

When you're in a conversation and want a specific agent to handle something, you @-mention them directly — the same way you'd tag a person in a team chat.

Owner (Google Chat): @CMO I need a re-engagement campaign for clients we haven't heard from in 90 days.
               What's the approach?

CMO Agent:     Good brief. If you share your client list, I'll draft a 3-touch sequence:
               email, then SMS, then a personal outreach prompt for you.
               @OPS can you check the list for clients with no contact in 90+ days?

OPS Agent:     Done. Here is the list for you to review before anything goes out.

CMO Agent:     I'll have the campaign draft ready for your review. Want the first email
               in your brand voice or a fresh angle?

You sent one message. Two agents coordinated. You stayed in the loop throughout — you didn't manage the handoff; you just watched it happen.

What You See When Agents Hand Off

Delegation between agents is visible in the conversation thread. When your strategy agent routes a data request to your operations agent, you see both sides of that exchange. The chain is not hidden — it's part of the record.

This transparency is intentional. You should be able to trace every result back to every step that produced it. If the CMO agent drafted a campaign based on a segment your operations agent pulled, you can see both — the brief, the data request, the data returned, and the draft — in sequence.

When Delegation Needs Your Approval

Agent-to-agent delegation for information gathering and preparation does not require your approval — that's routine coordination. What always requires your explicit approval:

• Any external communication being sent (email, SMS, supplier message)
• Any financial commitment (purchase order, invoice, payment trigger)
• Any record modification that affects customer data
• Any irreversible action

The distinction is preparation versus execution. Agents can prepare anything. They execute only with your go-ahead.

Why This Matters

Expert humans set the direction. AI Agents handle the execution within that direction.

That's the model — not autonomous AI running your business, but a capable team that does the research, drafts the responses, coordinates across functions, and brings the results to you for review. You stay in the role of owner and decision-maker. Your agents handle the operational weight that currently costs you hours every week.

The coordination layer — the ability for agents to delegate to each other, trace the work, and surface it to you — is what separates this from a single-purpose AI tool. A chatbot answers one question. Your AI team handles multi-step work across your whole business and shows you every step.

Frequently Asked Questions

Can I disable @-mentions for specific agents?

Yes. During your build engagement, we configure which agents are active and which roles are included in your plan. You can also restrict agent-to-agent delegation for specific functions if you prefer tighter manual control over certain workflows.

What if two agents give conflicting answers?

Agents operate within defined authority boundaries and data scopes. Conflicts typically surface when the same question touches two domains — for example, a campaign budget question that spans marketing and finance. In that case, both agents contribute their view and the final decision is yours. They don't resolve conflicts autonomously.

Who decides the delegation chain?

The routing logic is configured during your build engagement based on your business structure and the agent roles included in your plan. You don't manage routing manually in normal operation — but you can see every routing decision in the audit trail.

Can I override a delegation mid-task?

Yes. If you see a delegation happening that you want to redirect, you can intervene in the conversation — address a specific agent directly or instruct the team to stop. Human direction takes priority over any in-progress agent coordination.

Is AI agent adoption still risky for SMBs?

Regulation is the most-cited concern. We mitigate the risk three ways: your data is governed by a deployment model we configure to your regulatory needs — so it is not used to train models under the business terms we hold with every provider, and stricter data-residency or isolation options are available where regulation requires — every consequential action waits for human approval (Human-in-the-Middle), and every decision is logged in a reviewable audit trail. You stay in control.

Governance, data, and why agents aren't chatbots

Connected Spaces AI Agents operate with scoped permissions — each agent can only access the specific systems and data it has been explicitly authorized to reach. How your data is handled is a deployment choice we configure to your regulatory needs: your business data is never used to train models, with Anthropic's commercial terms as one option and stricter data-residency or isolation models available where regulation requires. Every action is logged in a reviewable audit trail you can review at any time.

The governance model is also what makes AI agent adoption manageable rather than risky. We address the three most-cited concerns directly: data governance is a configured deployment choice, not a blanket promise; every consequential action requires your explicit approval before it executes (the Human-in-the-Middle model); and every decision is logged. You stay in control. Agents prepare the work. You decide what runs.

That governance model is also the clearest way to explain why an AI agent is not a chatbot. A chatbot answers questions. An AI agent takes action — it can work through the channels and files it has been given, prepare a purchase order, draft a campaign, and route work to another specialist agent, all within a defined authority boundary. Where a chatbot gives you a response, an AI agent gives you a traceable result waiting for your approval.

Related

• Who is on your AI team? — Specialist agents, each with a defined role, authority, and limits.
• Message your AI team over WhatsApp, Google Chat, SMS or email — No new app required; reach your agents on channels you already use.
• From chat to execution — How one message becomes a completed, traceable business task.
• How AI Agents connect to tools (MCP, explained) — MCP, the open standard, and how we connect your tools.
• See which plans include your AI team