I'm not willing to guess at capabilities, other than those hinted at by the WatchOS beta.
Strategically, I don't think Apple is quite ready to cut the Watch loose from the iPhone. There has to be a solid, compelling reason for someone whose only Apple product is a Watch to invest in other ecosystem devices later on. For now, the linkage to iPhone is mandatory, but when it becomes optional, the Watch should enhance the iPhone/iPad/Mac experience, and vice-versa. It's nice that my Watch automatically unlocks my Mac, but that's not enough. I don't think that will happen overnight with Series 3. I'd give it a couple of more years.
I'm thinking that one ecosystem benefit might be motion-sensing-related - interpretation of hand gestures, proximity sensing, and such - automatically unlock iPhone and iPad the way I can currently unlock my Mac, mouse-like/game controller input, maybe a silent "Hey Siri" wave to HomePod? Watch should be a better input device for HomeKit devices than an iPhone.
Emergency monitoring, the "I fell down and I can't get up" and automotive collision monitoring businesses, are ripe for disruption. The existing hardware leaves a lot to be desired, and those extra cellular connections are quite redundant. It'd be better if incorporated into the Apple ecosystem - HomeKit, Car Play, SOS on the Watch and iOS.... The monitoring companies continue providing the subscription-based service, with 15%-30% going to App Store. The monitoring biz makes most of its money on service, not hardware - without hardware they might be more profitable.
Further, I'm not ready to trust that my 82-year-old aunt or 88-year-old father will remember to trigger an SOS from either Watch or iPhone (I'm not sure I'd remember, either). They need a panic button/impact sensor, and a monitoring service that will assess the situation before summoning emergency services.
The recently-announced linkage to exercise machines hints of the same possibility for easy connection to medical monitoring equipment.
From the beginning, I've expected to see Watches appear on the wrists of every hospital patient, perhaps tethered to a HomePod in each room - replace those fingertip heartbeat monitors, full-time fever monitoring, motion sensing (unauthorized exits from bed, etc.), patient identification... with less attention required of the nursing staff, since the sensors are securely fastened to the patient. It's the perfect item for US-style, no-expense-spared hospital care - the Apple logo proof that Mom or Grandpa are getting first-class care.
While some may dream of a free-standing cellular capability for Watch, the cost of adding another device to the monthly cellular bill is likely to deter a fair number of users (compare number of wifi-only iPads to those with cellular). Continued tethering to a more powerful cellular radio, with greater battery power, bigger antenna, etc. makes sense economically and also as a peace-of-mind thing - if Dad is wearing a Watch as an emergency device, why take a risk on weaker cellular reception? First- and second-generation cellular-equipped Watches are going to be for early adopters, not worried family members.
This stuff would likely be part of a five-year roadmap, gradually rolling out, a few capabilities at a time. I don't want to predict which will be for Series 3, which for subsequent models.
There's one other thought, which is that each Series may be tied to a different market segment. Series 1 was the Fashion watch. Series 2 is Exercise/Fitness. Series 3... I don't want to guess. I think timing-wise, Series 4 is more likely for Medical, and maybe Series 5 will be robotics?
