Nope; sorry, Wendy. First up, unfortunately, AncestryDNA can't be used for triangulation purposes, whether for WikiTree or for your own research...that is, not unless the raw data files are transferred to a different service provider, like GEDmatch, where shared segments can be compared. WikiTree does accept Ancestry DNA matching if the test-takers are 3rd cousins or closer and the AncestryDNA relationship estimate is in line with that, but Ancestry don't allow us to see any matching segment detail at all, so the benefits of using AncestryDNA for comparisons stops there.
As genetic genealogist Jim Bartlett frequently points out, triangulation groups are built around segments, not people. Even though there is no "industry" accepted standard or practice when it comes to autosomal DNA triangulation, one thing that everyone agrees on is that you absolutely positively need to see complete detail about those mutually-shared segments: the chromosome, the estimated size in centiMorgans, and the predicted start and stop points using the base pair number (also known as the reference cluster). "In common with," or the fact that you match a couple of other people--even if you know where on a chromosome--isn't good enough. If A matches B and C, you also then have to know exactly how B matches C: it isn't uncommon that A will match B and C only to find that B and C aren't a match at all on that segment.
It also helps greatly if the comparisons all use the same toolset and the same version of the human genome map (for example, if all test-takers involved are on GEDmatch and you can be assured you're comparing apples to apples). The Human Genome Consortium is currently on version 38 (technically GRCh38.p12), but to my knowledge only YSEQ in Germany has moved to that version; GEDmatch uses version 36, as do most major testing companies for the sake of compatibility. I personally also recommend that all matching segments be compared to known population pile-up regions before even starting triangulation work; can save some time by back-shelving those matches to a low or no priority. Pile-up regions are chromosomal areas that display much higher rates of commonality across large population bases, indicating they carry forward anthropologically, not with a genealogically useful reference.
Second, triangulation has to work in lockstep with the traditional genealogical paper trail. When you're working with test-takers that are 4th cousins or beyond, things get really tricky. In fact, with our current technology, it's believed that only 46% of your 4th cousins will share any detectable DNA with you at all. And at that degree of distance refining the data can get a bit squirrelly because that farther back you go, the more biological ancestors are in the mix (including possible if not likely pedigree collapse)...and the job then isn't just to accept the first line of inheritance that seems to fit, but to rule out lines of inheritance that might also fit. That last bit isn't in the WikiTree guidelines, but the underlying necessity for the paper trail is. It's awfully difficult to use autosomal triangulation to work backward to 3g-grandparents or farther without having a solid paper-trail hypothesis before you start.
Last up, "triangulation" was adopted from the land surveyor's lexicon, and was first applied to genetic genealogy for yDNA, I believe in 2002 or 2003. For autosomal DNA, the term is more than a bit unfortunate. Again, not a WikiTree guideline or policy, but autosomal triangulation should never mean three people tested and you're done. If you can get all five of those Ancestry hits over to GEDmatch for comparison, you'll have something goin' on. But in general, if you're trying to get to, say, 4g-grandparents and the 6g-grandparents were 2nd cousins, you may need to see test results from 6, 8, or more cousins before drawing any reasonably accurate conclusions. That's why the aforementioned Jim Bartlett talks about triangulation groups in terms of segments, not people. Jim wrote just last month that he now has over 385 triangulation groups consisting of over 10,000 shared segments...so he averages about 26 segments in each triangulation group. So...ABT: Always Be Triangulatin'!