I'll give you two solutions.
The first is to use back to back MX's.
Org1-MX1<------>Org1-MX2<->Org2-MX3
Org1 has two MXs, MX1 and MX2. MX1 is in their main DC. MX1 is the main AutoVPN hub for Org1.
Org2-MX3 is sitting in the main DC for Org2. MX3 is the main hub for Org3. Org1-MX2 and Org2-MX3 are both sitting right beside each other in the same rack.
Org1-MX1 and Org1-MX2 use AutoVPN between themsevles. Org1-MX2 has a static supernet route pointing to Org2-MX3. This static route gets redistributed into Org1 AutoVPN.
Org2-MX3 has a static supernet route pointing to Org1-MX2 which it redistributes into its own Org via AutoVPN.
Org1-MX2 and Org2-MX3 plug into each other using a plain patch cable.
You could also use BGP between MX2 and MX3 if they are both using VPN Concentartor mode.
You can use this approach to connect all of the divisions back to the mothership.
Note when sizing MX's for VPN Concentrator mode only pay attention to the number of VPN peers supported and the crypto throughput. Ignore all the other numbers.
Another similar approach is to specify that a specific DC will be a transit site. Every org puts an MX into this transit site. Their is a single subnet that all the MX's connect to, and they could all run BGP between themselves (or you could use static routing). You could also call this a peering point or a peering exchange.
The second solution is to put everything into a single org. If you ever sell a division you simply request an Org Split with support. Its a relatively simple procedure.
Using non-Meraki VPNs to join divisions together is painfull. It would only work on a very small scale.