diversoln, here is the straight skivvy.
Access is a small-shop database. Powerful as all Hell, pretty good for what it does, but small-shop mentality in its design. And WORKSTATION-oriented in its operational paradigm. Shared databases are an add-on and extension of something that originally was intended to be standalone.
When you have a shared database, the host for that database is a file server. It is NOT an application server. The Access application runs on the client machines. The problem is that when the client runs the app, the client is the machine that inspects each record for whether it is qualified to be a member of a particular query's answer (recordset). The down side of this is that the client machine must see EVERY RECORD in the originating tables to make this decision. So a count query or a select query or a sum query all have the same issue. They have to see a LOT of records.
I've read WayneRyan's answer and it seems misleading. Perhaps I'm reading it wrong. But I can tell you this - the designer's choices in an All-Access shop don't affect whether the client has to see all the underlying records. Access, when it runs its own queries, MUST ALWAYS see ALL records in ALL tables that contribute to the query.
If your app has a lot of records (by lot I'm talking many tens of thousands minimum) then your network person is spot-on in complaining. If your app has fewer records, you will pull less data across the net. So the first question is, when you run a query against your tables, how many records have to be EXAMINED to form the answer set?
The bit that Cable gave you about changing to SQL server with an Access front end is because when you do that, the SQL server examines the records locally and ONLY sends the answer set, not the tables from which it was derived. That is because SQL Server runs on the SERVER, not the client. If you had ORACLE or some other PC-based package that accepted ODBC already, and if you had the appropriate licenses for it, you could use something besides SQL Server. Actually, you would need licenses for THAT, too.
Another question that comes to mind is, how fast is your corporate backbone? Are we talking 10 Mbit Ethernet? 100 Mbit? 600 Mbit? That would also make a difference. How much competing traffic runs over your net? If you are running Ethernet with about 30% load, your collision rate (messages trying to step on each other) will start to rise. If you are above 40%, I know why your network person is having his/her apoplexy over yet another network-intensive app.
You can perhaps reduce the load a little by splitting your DB into a front end and back end, then distribute the front end. But sadly, that will only go a little ways towards load reduction.
Another possibility depends on how much data changes and whether you can distribute the database to multiple machines, then run a reconciliation. But if your system wasn't built for that up front, it would be hard as Hell to retrofit.