Akama - and more to the point Akama Prime (AP) - is a card that I have wanted to find interesting for the last year, but it has always been underwhelming. I'm hoping that by writing all my thoughts down I can begin to understand why, and then better understand card design too.
As a 3 mana 3/4 stealth, the first body is a solid card by all accounts, and certainly in the same league as the first bodies of the other prime cards. If anything it is better than most of the primes' first bodies because stealth denies the opponent the ability to remove the minion before it accomplishes something, and isn't situational like rush or spell damage.
The problem arises with the prime version of Akama. All the prime cards were designed to be bonkers strong, and something you couldn't print as a normal minion. Where the other prime cards achieve this with a hefty block of text detailing a fancy effect, Akama takes a more succinct and elegant approach by simply switching out 'stealth' for 'permanently stealthed', which we know from adventures can be very powerful. I have no doubt that if AP was a collectible card, it would indeed be bonkers strong.
So where's the problem? I think it is that the odds of drawing Akama twice by turn 6 are tiny, even in rogue. Statistics tell us you need to be ~75% of the way through your deck on average to draw AP, by which point you're probably long past turn 6. If you're against aggro or playing aggro yourself, the game is probably over before you see AP, so it is only worth evaluating against slow decks, and therein lies the issue: most slow decks have a way to deal with a 5 health stealthed minion.
Even when they cannot remove it immediately, AP still takes a whole turn before it can attack and show off the permanence of its stealth, and it probably has to attack at least twice before people it becomes interesting, making AP really, really slow. Compare that to the other primes, all of whom have much faster effects except Reliquary Prime, but at least that has a set of stats and abilities that make it more likely to achieve something significant right away.
The end result is that AP is in practice only marginally better than a 6 mana 6/5 stealth, a stat-line that can conveniently be compared to two cards that have been around in the Classic set since 2014: Stranglethorn Tiger and Ravenholdt Assassin. I think the tiger did see a tiny bit of play in druid at one point, but that's about all these two ever did in constructed. So AP sitting between them makes it seem like AP is actually just a bad card on its own. I'd say he's weaker than MSoG's attempt at this sort of effect (Lotus Assassin), and you can run 2 of those without any worries about shuffling them first.
Am I contradicting myself? First I say I think AP is a bonkers card, then a bad one? Again, timing is everything. If you could play AP on turn 6 and have him be part of a wider plan to pressure the opponent, then he's great. But if you have him on turn 14, when your opponent has stabilised or you are playing a deck that doesn't put on much pressure, then he doesn't do very much.
The (flawed) logic behind the design
It's quite easy to see how AP was made the way he was. Not only does it look cool and powerful at first glance (the community rated Akama quite highly), but we've seen first hand how powerful permanent stealth can be in adventures.
It's also notable that the prime mechanics (deathrattle and shuffle) are both things rogue had synergies for, and it stands to reason that rogue would have an easier time abusing primes than other classes. Stowaway in particular seems to alleviate the slow draw of AP. The trouble with the Stowaway solution is that you still need to have drawn both Akama and a Stowaway, and you have to be playing a fairly slow deck to justify him when you don't see Akama.
The fix
Fixing AP is simple: give him more health. I'd lean towards 7 to put him just out of range of modern Flamestrike +1 spell damage (which is so common and cheap in mage today that it's pretty much always going to be there). That way it makes 1 more class have to work to remove him, but makes almost no difference to classes that couldn't kill him to begin with, or who have un-targeted hard removal. It also gives him the opportunity to trade into a taunt minion without being reduced to piddly health that can be removed by small AoE.
The design lessons
We already know a card's power depends strongly on when it is drawn and played. This is doubly true for uncollectible cards that are shuffled into your deck. So the lesson is to ensure such a card is a good late game card at bare minimum. It doesn't matter how good it is as an early-to-mid game card if it's almost never going to be drawn then.
