How crowded the spectrum is in your area is going to be the biggest factor influencing your success. In environments where 802.11 has already been deployed densely your radios will be fighting for bandwidth with everyone else's radios. This will radically diminish the effective range of your access points. If you're serious about doing this you need to get some kind of spectrum analyzer to see how "noisy" the air is. (We don't do product recommendations here, but there are a variety of inexpensive spectrum analyzers out there that focus on the 2.4Ghz and 5.8Ghz ISM bands used by 802.11.)
Worse yet, you can't control how others use the air. What might be good spectrum today could turn bad tomorrow depending on what others deploy, and so long as they're not breaking the law w/ respect to transmit power, etc, there's nothing you can do.
The quality of your wireless access points (and, to an extent, their cost-- which serves as a bit of a proxy for quality) will be the biggest key to success or failure that you can control. Low-end access points will not support very many simultaneous clients, and will perform erratically.
Finally, the types of radios (and antennas) in your clients will make a big difference. You really should try the specific client devices and radios you intend to use in production as part of surveying the site.
I would echo the comment made by @thanasisk. You're going to have a lot closer to 80 wireless clients than 40 if you allow mobile phones to be joined. That's more than even a couple low-end access points can handle. You'll be graduating to more expensive APs right out of the gate.
In my opinion you'll spend less time and money supporting a wired network of that size versus wireless. The initial investment of time and money to wire such a small network will be more than eclipsed by the cost of quality wireless access points.
The cost of a 48-port 10/100/1000 Ethernet switch, likewise, will be a fraction of the cost of the access points. Low-end Ethernet switches, unlike low-end access points, actually work fairly well. You'll be sacrificing management and functionality in a low-end switch, but it'll still work well.
The reliability of a wired network, compared to wireless, is night and day. Even in deployments where I have very high-end access points I still see better bandwidth and reliability over the wired network.
Securing access to a wired network medium is also, in my opinion, easier than securing an equivalently-sized wireless network. Unless you're willing to move to 802.1x over shared-key encryption you run the risk that a former employee leaving the company will be carrying with them the credentials necessary to grant access to the wireless medium. Changing a shared-key in an environment with 40+ clients would be a nightmare. If you opt to go with 802.1x you will need some supporting infrastructure (RADIUS server, some type of authentication database) to support it.
If you had 5 - 10 clients I'd be more apt to suggest a wireless network, but I think that your situation is sized such that wired will end up being a "big win" over the long haul. It's small enough to make wiring fairly low-cost, but large enough that you're going to find wired to be a lot easier and cheaper to administer.