Think of an access point (abbreviated as "AP") as being like a cell tower.  Your mobile phone uses radio waves to talk to the antenna on the cell tower.  The cell tower forwards that call to the service provider, which might route the call onto the wired telephone network.  If the phone moves away from one cell tower and towards another, it will stop sending its signal through the far one and start sending its signal through the near one, which is known as roaming.

In the same way, a wireless client uses radio waves to talk to the antenna on an access point.  The access point is usually plugged into a wired Ethernet network (although it doesn't have to be), and it will forward packets onto the wireless network if necessary.  This means that access points can give wireless clients access to resources on the wired network, such as a server, or an Internet connection.

In general, to gain access to the Internet, a wireless user must be in range of an access point.  This influences how many access points will need to install and where those access points must be placed. Just as cellular signals can be blocked or have "dead zones" so to can wireless coverage.

An 802.11 Wi-Fi access point provides a connection between the electrical data path formed by an Ethernet cable and the RF signal data path formed by the Wi-Fi radios. In addition, access points provide a number of additional features and capabilities related to access control, data encryption, fault tolerance, network management. In essence, an access point is a translational bridge, converting the TCP/IP data packets from their 802.11 frame encapsulation format in the air to the 802.3 Ethernet frame format on the wired Ethernet network.