Seg-wit is an opt-in soft-fork to miners, they can signal and protect users without supporting. It probably is the case that there are people who do not understand this subtle point, so I wanted to flag it. Non-segwit blocks remain valid after activation of segwit, though a miner should protect their node using a segwit aware border node, or upgrade their node but tweak it not to yet create segwit blocks until they get comfortable with it. And unlike previous soft-forks, segwit is more miner opt-in forgivving and wont penalise via block invalidity people who do not switch their block version after activation, and there is protection provided by other miners: even if a minority of miners dont upgrade nor border node protect their infrastructure and continue mining without defences, their blocks remain valid they just are exposed to someone wasting $13k to make an invalid segwit block, which they might temporarily build on until the majority orphans or rejects it, as other miners and ecosystems economic full nodes would be validating.
In addition seg-wit transactions themselves are opt-in and provide unilateral scaling to people who adopt (there is a long list of services and wallets etc that are ready https://bitcoincore.org/en/segwit_adoption/) and adoption also creates capacity for everyone by logically (but not physically) moving the witness (signatures) out of the 1MB block. (Physically segwit blocks are just larger blocks https://twitter.com/lopp/status/830129625196068865 it is only old nodes that have an alternate serialisation sent to them)
There is a list of segwit benefits https://bitcoincore.org/en/2016/01/26/segwit-benefits/ and costs/risks https://bitcoincore.org/en/2016/10/28/segwit-costs/ for a fuller discussion, just wanted to highlight that segwit is 3x opt-in. First 1) the ecosystem has decide if it wants it as a pragmatic tested incremental step to scaling (seems to be the case by organic node count, ecosystem support and readiness https://bitcoincore.org/en/segwit_adoption/ ) where-upon miners then should signal readiness (which is different from support) to allow those who want to opt-in to do so, and then 2) post activation it is opt-in for miners whether they generate segwit blocks (or wait a while to decide) and 3) it is up to users and services whether they upgrade and benefit directly from the scale (which benefits others via increased free space).
Finally there is some dangling confusion about "discounts" or "economic changes". Whether you call it a discount or an economic change, it is an unequivocal good that the perverse incentives to create UTXO bloat are reduced. Today it is literally cheaper for a wallet to split one coin into change vs spend change from two coins which reduces UTXO impact. UTXO is a scaling factor, it needs fast access (cache, memory) and accesses to it scales non-linearly, latency to it has been getting worse as use grows. Artificially bloating it helps no one. That the bloat is not worse is in part due to mildly altruistic or non-short-term cost-optimal change selection in wallets. The weight construction in segwit is technically necessary anyway to get scale without introducing a 2d optimisation problem, and reduces the incentive to bloat UTXO hurting scalability. So for people who look at that economically, the way to do so is that it is a beneficial reduction in a negative economic externality.
edit: add soft-fork security explanation from https://np.reddit.com/r/btc/comments/5uf6am/charlie_lee_people_dont_realize_what_segwit_is/ddutt8x/
A non-upgraded node will not see non-segwit transactions until they are included in a block, because they are non-standard, but valid (once mined) so they are not relayed to non-segwit nodes, nor between them (not locally accepted even if injected direct to node) and only appear skipping to 1-conf once a block is confirmed. The selection of a non-standard format was by design to reduce the issue of accepting as 0-conf transactions that are not valid.
In general about soft-fork upgrades and rogue miner attack: for people who do not upgrade they are more vulnerable to miner attacks post soft-fork. The statistics in the network today of non-soft-fork upgraded nodes are not great, so it's not a new problem, all soft-forks are equal basically for this kind of attack. The attack costs $13k to make an invalid block whether that is segwit post activation, or a CSV or even CLTV to people running old nodes. However even people who have upgraded are vulnerable to finney attack, double-spend etc at costs of $13k and below. So in general for high value transactions people should run uptodate fullnodes, or SPV wallets that cross check an uptodate and semi-trusted fullnode with p2p fullnodes and wait a few confirmations.