Most accounts start with one AI Agent. Many end up with two or three, each focused on a specific scenario. Multiple AI Agents are usually a better design than one trying to do everything.
When one is enough
Single property
A small set of guest channels (one or two)
Knowledge needs are similar across all channels (check-in, parking, breakfast, and so on)
Keep it simple. Add a second AI Agent only when the first one cannot do the job well.
When to add a second
Several patterns justify splitting:
Different channels with different needs. A WhatsApp-only check-in AI Agent can have a different tone, shorter responses, and a tighter knowledge section than an email reservations AI Agent.
Different properties. Each property has its own address, hours and parking arrangements, so giving each property a dedicated AI Agent keeps the knowledge clean.
Different teams behind the handovers. If WhatsApp handovers should go to the front desk and email handovers to reservations, splitting aligns with the handover routing.
Different levels of oversight. You may want one agent replying to guests directly and another running in draft mode, so an operator checks its answers before they send. That is a setting on the agent, so it means two agents.
Setting up the second agent: duplicate the first
The fastest way to create an agent that resembles one you already have is to copy it whole.
Go to Inbox Settings > Bots.
Find the agent you want to copy and click the ⋮ menu at the end of its row.
Click Duplicate.
You then open the copy and change what differs — usually the name, the channel filter and the assignment rule, plus whichever knowledge items are property-specific.
A duplicate is an independent copy. Once made, the two agents have no relationship. Editing one does not change the other, in either direction. That is deliberate — it is what lets you diverge them — but it does mean a policy that lives in both has to be updated in both.
Knowledge across several agents
Each AI Agent has its own knowledge section. There is no shared pool that several agents read from, and no live sync between them.
What has changed is the cost of putting the same knowledge in two places. You no longer retype it:
Copy the whole knowledge base by duplicating the agent, as above.
Copy just the knowledge into an agent that already exists: open its Knowledge tab, click Import, choose Existing bot, and pick the agent to copy from. See Importing knowledge into your AI Agent.
Both are snapshots. This is the thing to be clear about with your team. Copying your cancellation policy into four agents gives you four independent copies of it. Change it in the original later and the other three still carry the old wording, and will keep telling guests the old wording until someone updates them.
So the trade-off has shifted rather than disappeared. Setting up several agents is now quick. Keeping them consistent is the same work it always was.
Keeping several agents in step
Two habits make this manageable:
Know which items are shared. Tag them. If your cancellation policy, group booking rules and payment terms all carry a "shared" tag, then when one changes you can filter to that tag in each agent and see exactly what needs editing. Without tags you are reading the whole list in every agent.
Keep one agent as the reference. Decide which agent holds the canonical version of a shared policy, change it there first, then copy or re-enter it in the others. It is a convention rather than a feature, but it stops two agents drifting into two different answers.
Deciding not to split
Splitting is not always right. Before adding an agent, consider whether one agent with well-organised knowledge would serve you better.
Tags make a single larger knowledge base workable in a way it was not before — you can filter to one property, one topic or one channel's questions without splitting the agent. If the only reason you were going to split is that the list has become hard to navigate, tagging solves that problem without doubling your maintenance.
Split when the agent needs to behave differently — different tone, different channel, different handover team, different oversight. Do not split simply because the knowledge base is large.
How most teams structure it
The most common pattern: one arrival/check-in AI Agent on WhatsApp, plus one general enquiries AI Agent on email. Two agents, clear scopes, and a small set of shared policies to keep aligned between them.
Need Help?
Please contact us through the 'Talk to Us' option on the left menu in the platform, or through the Bookboost Support email at support@bookboost.io if you have questions or need additional support.