If by ‘counted by hand’ one refers to a process wherein a ballot is visually examined by a human being, the votes seen are recorded on paper or manually typed into a computer, and those votes are added to the official totals without ever running through a tabulation machine, the answer is no, unless there is reliable cause found to doubt machine results (G.S. 163-182.2 (a)(6)).
All valid ballots, even Absentee by Mail and Provisional ballots, are eventually run through a machine for the official count. If a ballot never passes through a machine (specifically a tabulator, not the ExpressVote Ballot Marking Device), the votes on that ballot are not part of the results unless a full hand-eye recount of all results is ordered and subsequently deemed more accurate.
This is why provisional ballots that are determined to be valid are run through a machine rather than just being tallied up, and why damaged ballots returned through Absentee by Mail are duplicated rather than tallied up (for more on ballot duplication, see the Absentee by Mail Voting section). A machine count must be produced and relied upon if a certified voting system is being utilized in a particular county absent any major inaccuracies or valid protests resulting in further investigation.
Machines are tested for functional mistakes prior to an Election and sealed to prevent any access before use. Machines are incapable of deciding, of their own accord, to modify vote totals for malicious purposes, whereas a human being can never be guaranteed to produce no mistakes in counting and is capable of hiding malicious intent. Therefore, aside from write-in votes (which still must be registered by a machine as being a write-in vote), any manual human tallying of votes as an initial means of determining results is near-universally discouraged for jurisdictions over 1,000 voters.
As a helpful metaphor, this is the same reason why most of the duration of a commercial airliner flight is through autopilot. Only a handful of major air disasters have ever been tenuously connected to software malfunction, whereas tens of thousands have resulted from human error. In flight as well as in vote tabulation, it makes the most statistical and practical sense to cut as much potential for human error out of standard procedure as possible, saving hand counts for sample-based accuracy checks.
